AdAlta Ltd (ASX:1AD) CEO Dr Tim Oldham speaks with Proactive after announcing all healthy volunteers in its AD-214 Phase I extension study have successfully received three doses. The milestone allows pharmacokinetic and receptor engagement analysis to be ...

AdAlta making material progress on partnering talks

AdAlta Ltd (ASX:1AD) has reported positive outcomes from its AD-214 Phase 1 extension study. This study aims to assess the safety and availability of multiple 10 milligrams per kilogram intravenous doses of AD-214, the company's lead drug candidate.

AdAlta doses volunteers in Phase 1 extension study of AD-214, with promising safety results

AdAlta Ltd (ASX:1AD) has been selected for an extended oral presentation at the 11th Molecular Pharmacology of GPCRs meeting to be held in Melbourne from November 15-17, 2023.

AdAlta to present lead product AD-214 at MPGPCR 2023 pharmacology conference

Traders have been aggressively selling the three after Novo Nordisk appeared to have made a breakthrough in its wonder drug, Ozempic.

Check Up: Why Goldman Sachs won’t change its ratings on 3 ASX stocks despite Ozempic-driven selloff

AdAlta Ltd (ASX:1AD) has received its FY2023 Research and Development Tax Incentive (RDTI) refund of $2,350,940 from the Australian Government. The clinical-stage drug discovery company received the refund for 43.

AdAlta receives $2.3 million FY23 R&D refund; extends loan repayment terms

AdAlta Ltd (ASX:1AD) has received its FY2023 Research and Development Tax Incentive (RDTI) refund of $2,350,940 from the Australian Government.

AdAlta Ltd (ASX:1AD) has received its FY2023 Research and Development Tax Incentive (RDTI) refund of $2,350,940 from the Australian Government. The clinical-stage drug discovery company received the refund for 43.5% of eligible expenditure. The Research ...

AdAlta Ltd (ASX:1AD) has received its FY2023 Research and Development Tax Incentive (RDTI) refund of $2,350,940 from the Australian Government. The clinical-stage drug discovery company received ...

AdAlta Ltd (ASX:1AD) CEO Dr Tim Oldham speaks with Proactive after receiving the results of dose simulation studies that support the potential efficacy of its i-body drug candidate AD-214 in ...

Adalta Says New AD-214 Dosing Models Enhance Partnering Potential

AdAlta

adalta

What Are the Top Startups in Oceania?

Authors:
(1) Sanchit Sinha, University of Virginia (sanchit@virginia.edu);
(2) Guangzhi Xiong, University of Virginia (hhu4zu@virginia.edu);
(3) Aidong Zhang, University of Virginia (aidong@virginia.edu). Table of Links Abstract and 1 Introduction 2 Related Work 3 Methodology and 3.1 Representative Concept Extraction 3.2 Self-supervised Contrastive Concept Learning 3.3 Prototype-based Concept Grounding 3.4 End-to-end Composite Training 4 Experiments and 4.1 Datasets and Networks 4.2 Hyperparameter Settings 4.3 Evaluation Metrics and 4.4 Generalization Results 4.5 Concept Fidelity and 4.6 Qualitative Visualization 5 Conclusion and References Appendix 4.3 Evaluation Metrics We consider the following evaluation metrics to evaluate each component of the concept discovery framework. • Generalization: We start by quantitatively evaluating the quality of concepts learned by measuring how well the learned concepts can generalize to new domains. To achieve this, we compare our proposed method against the aforementioned baselines on domain adaptation settings. • Concept Fidelity: To evaluate consistency in the learned concepts, we compute the intersection over union of the concept sets associated with for two data points xi and xj from same class as defined in Equation 10: 4.4 Genenralization Results Tables 2, 3, and 4 report the domain adaptation results on the OfficeHome, DomainNet, VisDA and the Digit datasets, respectively. The notation X→Y represents models trained on X as the source domain (with abundant data) and Y as the target domain (with limited data) and evaluated on the test set of domain Y . The best statistically significant accuracy is reported in bold. The last three rows in all the tables list the performance of the RCE framework, RCE trained with regularization (RCE+PCG), and RCE trained with both regularization and contrastive learning paradigm (RCE+PCG+CCL). Comparision with baselines. The first row in each table lists the performance of a standard Neural Network trained using the setting described in [Yu and Lin, 2023] (S+T). As a standard NN is not inherently explainable, we consider this setting as a baseline to understand the upper bound of the performance-explainability tradeoff. The second and third rows in each table lists the performance of SENN and DiSENN respectively. SENN performs worse than S+T setting in almost all settings, except in a handful of settings where the performance matches S+T. This is expected, as SENN is formulated as an overparameterized version of a standard NN with regularization. Recall that DiSENN replaces the autoencoder in SENN with a VAE, and as such is not generalizable to bigger datasets without domain engineering. DiSENN performs the worst among all approaches for all datasets due to poor VAE generalization. Recall that UnsupervisedCBM is an improved version of SENN architecture with a discriminator in addition to the aggregation function. In most cases, it performs slightly better than SENN and is at par with S+T. However, in particular cases in OfficeHome data (R→A) and DomainNet (S→P), UnsupCBM performs the best. We attribute this result to two factors: first, the Art (A) and Sketch (S) domains are significantly different from Real (R) and Picture (P) domains due to both of the former being hand-drawn while the latter being photographed as mentioned in [Yu and Lin, 2023]. Second, the use of a discriminator as proposed in UnsupervisedCBM helps enforce domain invariance in those particular cases. BotCL explicitly attempts to improve concept fidelity and applies contrastive learning to discover concepts. However, the contrastive loss formulation is rather basic and they never focuses on domain invariance. BotCL’s performance is similar to S+T for the most part except in OfficeHome data (C→A), where it just outperforms all other approaches. One possible reason is that Clipart domain is significantly less noisy, and hence basic transformations in BotCL work well. As the last row demonstrates, our proposed framework RCE+PCG+CCL outperforms all baselines on a vast majority of the settings across all four datasets and is comparable to SOTA baselines in the other settings. Ablation studies. We also report the performance corresponding to various components of our proposed approach. We observe that the performance of RCE is almost identical to SENN, which is expected as there is very weak regularization in both cases. In almost all cases, adding prototype-based grounding regularization (RCE+PCG) improves performance over RCE while models trained with both PCG regularization and contrastive learning (RCE+PCG+CCL) outperform all approaches on a vast majority of settings across all datasets. Note that the setting RCE+CCL is not reported, as it defeats the fundamental motivation of maintaining concept fidelity. Effect of the number of concepts and dimensions. We observe that there are no significant differences in performance over a varying number of concepts or dimensions. For all results reported, the number of concepts is set to a number of classes in the dataset, and their dimension is set to 1. For results on varying numbers of concepts and dimensions - refer to Appendix. 4.5 Concept Fidelity As the RCE framework is explicitly regularized with a concept fidelity regularizer and grounded using prototypes, we would expect high fidelity scores. Table 5 lists the fidelity scores for the aforementioned baselines and our proposed method. Fidelity scores are averaged for each domain when taken as a target (e.g. for the domain (A) in DomainNet, the score is an average of C→A, P→A and R→A). As expected, our method and BotCL, both with specific fidelity regularization outperform all other baseline approaches. Our method outperforms BotCL in most settings, except when the target domains are Art in DomainNet and Clipart in OfficHome due to significant domain dissonance. 4.6 Qualitative Visualization Domain Alignment. We consider the extent to which the models trained using both concept grounding and contrastive learning maintain concept consistency not only within the source domain but also across the target domain as well. To understand what discriminative information is captured by a particular concept, Figure 5 shows the most important prototypes selected from the training set of both the source and target domains corresponding to five randomly selected concepts. We observe that prototypes explaining each concept are visually similar. For more results, refer Appendix. Explanation using prototypes. For a given input sample, we also plot the prototypes associated with the highest activated concept, i.e., the important concept. Figure 6 shows the prototypes associated with the concepts most responsible for prediction (highest relevance scores). As can be seen, the prototypes possess distinct features, for eg., they capture the round face of an alarm clock. More results are reported in Appendix. This paper is available on arxiv under CC BY 4.0 DEED license. Authors: (1) Sanchit Sinha, University of Virginia (sanchit@virginia.edu); (2) Guangzhi Xiong, University of Virginia (hhu4zu@virginia.edu); (3) Aidong Zhang, University of Virginia (aidong@virginia.edu). Authors: Authors: (1) Sanchit Sinha, University of Virginia (sanchit@virginia.edu); (2) Guangzhi Xiong, University of Virginia (hhu4zu@virginia.edu); (3) Aidong Zhang, University of Virginia (aidong@virginia.edu). Table of Links Abstract and 1 Introduction Abstract and 1 Introduction 2 Related Work 2 Related Work 3 Methodology and 3.1 Representative Concept Extraction 3 Methodology and 3.1 Representative Concept Extraction 3.2 Self-supervised Contrastive Concept Learning 3.2 Self-supervised Contrastive Concept Learning 3.3 Prototype-based Concept Grounding 3.3 Prototype-based Concept Grounding 3.4 End-to-end Composite Training 3.4 End-to-end Composite Training 4 Experiments and 4.1 Datasets and Networks 4 Experiments and 4.1 Datasets and Networks 4.2 Hyperparameter Settings 4.2 Hyperparameter Settings 4.3 Evaluation Metrics and 4.4 Generalization Results 4.3 Evaluation Metrics and 4.4 Generalization Results 4.5 Concept Fidelity and 4.6 Qualitative Visualization 4.5 Concept Fidelity and 4.6 Qualitative Visualization 5 Conclusion and References 5 Conclusion and References Appendix Appendix 4.3 Evaluation Metrics We consider the following evaluation metrics to evaluate each component of the concept discovery framework. • Generalization : We start by quantitatively evaluating the quality of concepts learned by measuring how well the learned concepts can generalize to new domains. To achieve this, we compare our proposed method against the aforementioned baselines on domain adaptation settings. Generalization • Concept Fidelity: To evaluate consistency in the learned concepts, we compute the intersection over union of the concept sets associated with for two data points xi and xj from same class as defined in Equation 10: • Concept Fidelity: 4.4 Genenralization Results Tables 2, 3, and 4 report the domain adaptation results on the OfficeHome, DomainNet, VisDA and the Digit datasets, respectively. The notation X→Y represents models trained on X as the source domain (with abundant data) and Y as the target domain (with limited data) and evaluated on the test set of domain Y . The best statistically significant accuracy is reported in bold. The last three rows in all the tables list the performance of the RCE framework, RCE trained with regularization (RCE+PCG), and RCE trained with both regularization and contrastive learning paradigm (RCE+PCG+CCL) . RCE RCE (RCE+PCG), RCE (RCE+PCG+CCL) Comparision with baselines. The first row in each table lists the performance of a standard Neural Network trained using the setting described in [Yu and Lin, 2023] (S+T). As a standard NN is not inherently explainable, we consider this setting as a baseline to understand the upper bound of the performance-explainability tradeoff. Comparision with baselines. The second and third rows in each table lists the performance of SENN and DiSENN respectively. SENN performs worse than S+T setting in almost all settings, except in a handful of settings where the performance matches S+T. This is expected, as SENN is formulated as an overparameterized version of a standard NN with regularization. Recall that DiSENN replaces the autoencoder in SENN with a VAE, and as such is not generalizable to bigger datasets without domain engineering. DiSENN performs the worst among all approaches for all datasets due to poor VAE generalization. Recall that UnsupervisedCBM is an improved version of SENN architecture with a discriminator in addition to the aggregation function. In most cases, it performs slightly better than SENN and is at par with S+T. However, in particular cases in OfficeHome data (R→A) and DomainNet (S→P), UnsupCBM performs the best. We attribute this result to two factors: first, the Art (A) and Sketch (S) domains are significantly different from Real (R) and Picture (P) domains due to both of the former being hand-drawn while the latter being photographed as mentioned in [Yu and Lin, 2023]. Second, the use of a discriminator as proposed in UnsupervisedCBM helps enforce domain invariance in those particular cases. BotCL explicitly attempts to improve concept fidelity and applies contrastive learning to discover concepts. However, the contrastive loss formulation is rather basic and they never focuses on domain invariance. BotCL’s performance is similar to S+T for the most part except in OfficeHome data (C→A), where it just outperforms all other approaches. One possible reason is that Clipart domain is significantly less noisy, and hence basic transformations in BotCL work well. As the last row demonstrates, our proposed framework RCE+PCG+CCL outperforms all baselines on a vast majority of the settings across all four datasets and is comparable to SOTA baselines in the other settings. Ablation studies. We also report the performance corresponding to various components of our proposed approach. We observe that the performance of RCE is almost identical to SENN, which is expected as there is very weak regularization in both cases. In almost all cases, adding prototype-based grounding regularization (RCE+PCG) improves performance over RCE while models trained with both PCG regularization and contrastive learning (RCE+PCG+CCL) outperform all approaches on a vast majority of settings across all datasets. Note that the setting RCE+CCL is not reported, as it defeats the fundamental motivation of maintaining concept fidelity. Ablation studies. Effect of the number of concepts and dimensions. We observe that there are no significant differences in performance over a varying number of concepts or dimensions. For all results reported, the number of concepts is set to a number of classes in the dataset, and their dimension is set to 1. For results on varying numbers of concepts and dimensions - refer to Appendix. Effect of the number of concepts and dimensions. 4.5 Concept Fidelity As the RCE framework is explicitly regularized with a concept fidelity regularizer and grounded using prototypes, we would expect high fidelity scores. Table 5 lists the fidelity scores for the aforementioned baselines and our proposed method. Fidelity scores are averaged for each domain when taken as a target (e.g. for the domain (A) in DomainNet, the score is an average of C→A, P→A and R→A). As expected, our method and BotCL, both with specific fidelity regularization outperform all other baseline approaches. Our method outperforms BotCL in most settings, except when the target domains are Art in DomainNet and Clipart in OfficHome due to significant domain dissonance. 4.6 Qualitative Visualization Domain Alignment. We consider the extent to which the models trained using both concept grounding and contrastive learning maintain concept consistency not only within the source domain but also across the target domain as well. To understand what discriminative information is captured by a particular concept, Figure 5 shows the most important prototypes selected from the training set of both the source and target domains corresponding to five randomly selected concepts. We observe that prototypes explaining each concept are visually similar. For more results, refer Appendix. Domain Alignment. Explanation using prototypes. For a given input sample, we also plot the prototypes associated with the highest activated concept, i.e., the important concept. Figure 6 shows the prototypes associated with the concepts most responsible for prediction (highest relevance scores). As can be seen, the prototypes possess distinct features, for eg., they capture the round face of an alarm clock. More results are reported in Appendix. Explanation using prototypes. This paper is available on arxiv under CC BY 4.0 DEED license. This paper is available on arxiv under CC BY 4.0 DEED license. available on arxiv

How Reliable Are AI Concepts in Real-World Domain Adaptation Tasks?

Table of Links Abstract and 1. Introduction Abstract and 1. Introduction 2 Related works 2 Related works 3 Methodology and 3.1 Causal language model as a classification model 3 Methodology and 3.1 Causal language model as a classification model 3.2 Functional token 3.2 Functional token 3.3 Dataset collection 3.3 Dataset collection 3.4 Model development and training 3.4 Model development and training 4 Experiments and 4.1 Android function calls 4 Experiments and 4.1 Android function calls 4.2 Extension to Vehicle, Yelp, and DoorDash function sets 4.2 Extension to Vehicle, Yelp, and DoorDash function sets 4.3 Full and partial training datasets and 4.4 Full training and LoRA training 4.3 Full and partial training datasets and 4.4 Full training and LoRA training 4.5 Parallel and nested function call and 4.6 Weighted loss function for special tokens 4.5 Parallel and nested function call and 4.6 Weighted loss function for special tokens 5 Discussion and future works and References 5 Discussion and future works and References Appendix Appendix A.1 Android function examples A.1 Android function examples A.2 Vehicle function examples A.2 Vehicle function examples 4.2 Extension to Vehicle, Yelp, and DoorDash function sets In addition to Android function calls, we expanded our evaluation to include 20 vehicle function calls, showcasing the algorithm’s adaptability to diverse use cases. For vehicle functions, we focused on essential control methods such as volume adjustment, air conditioning, and seat positioning. We conducted benchmarks for vehicle functions paralleling the Android function evaluation, observing consistent performance patterns. Details on vehicle functions are provided in the Appendix, enabling users to customize a new set of functional APIs for their specific needs. Furthermore, tests conducted with Yelp and DoorDash APIs confirmed a similar performance, underscoring our method’s versatility across various function sets. This paper is available on arxiv under CC BY-NC-SA 4.0 DEED license. This paper is available on arxiv under CC BY-NC-SA 4.0 DEED license. available on arxiv Authors:
(1) Wei Chen, Stanford University, with equal contribution and a corresponding author {weichen6}@stanford.edu;
(2) Zhiyuan Li, Stanford University and a corresponding author {zhiyuan8}@stanford.edu. Authors: Authors: (1) Wei Chen, Stanford University, with equal contribution and a corresponding author {weichen6}@stanford.edu; (2) Zhiyuan Li, Stanford University and a corresponding author {zhiyuan8}@stanford.edu.

Demonstrating Adaptability: Evaluating Function Calling on Vehicle, Yelp, and DoorDash APIs

Looking at nature, humankind’s survival through the ages has relied on one crucial trait: adaptability. While products are not living entities, adopting a "living vision" for your product can inspire adaptability, igniting its ability to thrive in a competitive landscape. Bringing successful products to life in today’s dynamic market demands more than just a great idea. It requires a clear, adaptable vision and a commitment to ongoing improvement. A strong product vision serves as a north star for teams, but its true power lies in its ability to evolve with shifting circumstances while staying aligned with core goals. This article breaks down five essential principles to help you build a strong, adaptable product vision that can withstand market shifts, meet customer needs, and unlock long-term success. 1. Set a Clear and Strong Foundation Your product vision is the foundation for your entire development strategy. A strong vision: Aligns the team around a shared purpose.
Clearly defines the problem your product solves.
Guides decisions for both short- and long-term growth with focus and clarity. Take the time to brainstorm, analyze, reflect, and refine. Gather diverse feedback, revisit the vision, and ensure it is solid before moving forward. Conduct thorough research to: Understand your audience and their needs.
Define the unique value your product brings.
Ensure your vision answers key questions:

What does success look like?
What unique value will this product bring?
How does it effectively address pain points? Engage stakeholders early to validate assumptions and secure team buy-in. A well-crafted vision is not only a guide but a safeguard against future obstacles. 2. Beyond Solving—Create an Exceptional Experience Listening to customers and responding to their needs is important, but true product success requires more. To truly stand out, you must: Solve existing problems effectively.


Anticipate future needs customers might not even be aware of. Even if customers are unhappy with an existing solution, switching comes at a cost—effort, time, and resources. According to Gartner, 77% of B2B buyers cite switching costs as a significant barrier to change. Your product must offer compelling features that make the switch worthwhile. Apple’s iPhone revolutionized the market by anticipating needs that consumers had not yet considered. By combining communication, productivity, and entertainment, Apple created an indispensable product that now dominates the smartphone market. 3. Ensure Adaptability in Your Vision Markets evolve, customer needs shift, and technology advances rapidly. A rigid product vision can quickly become obsolete. To mitigate this risk: Focus on outcomes rather than specific features.
Maintain modular flexibility to pivot when necessary.
Foster cross-functional collaboration to enhance innovation. Tesla serves as a prime example of adaptability. Initially focused on making electric cars aspirational, they expanded their vision to include renewable energy solutions, leading to billions in energy revenue in 2023. Their ability to pivot effectively ensured continued growth and market leadership. Preparing for uncertainty is essential. Scenario analysis can help businesses anticipate potential challenges and opportunities. Implement this by: Using “what-if” scenarios to test market shifts.
Conducting data-driven simulations to analyze risks
Regularly adjusting roadmaps based on scenario outcomes. 4. Revisit and Evolve Your Vision Frequently A product vision should not remain static. Regularly reassessing ensures alignment with current realities and emerging trends. To keep your vision relevant: Conduct quarterly or biannual reviews of your product vision.
Track progress using key performance indicators (KPIs).
Integrate customer and stakeholder feedback into revisions. A static roadmap can lead to stagnation, especially in industries driven by rapid technological advancements. Regular reviews ensure alignment and adaptability, preventing inefficiencies and misalignment across teams. 5. Balance Flexibility with Core Purpose A strong vision must be firm yet adaptable. Strike a balance by: Anchoring decisions to your core mission.
Defining non-negotiable elements while allowing room for change.
Setting clear deadlines and accountability structures. Spotify has maintained a clear vision of delivering personalized audio experiences. However, their adaptability enabled them to expand into podcasts, audiobooks, and live audio, significantly growing their market share. A product vision should function like a rubber band—firm enough to maintain its core shape yet flexible enough to stretch in response to change. Lessons from the Tech Industry: Success and Failure Successful Adaptations: Apple evolved towards sustainability, integrating recycled materials and targeting carbon neutrality.
Tesla expanded its vision beyond EVs into renewable energy, securing long-term growth. Cautionary Tales: Google Glass failed due to a lack of consideration for privacy, design, and practicality.
Crypto startups saw 80% failure rates due to issues with scalability, regulations, and trust. Key Takeaway: A static or incomplete vision can lead to irrelevance. Flexibility, adaptability, and attention to detail separate lasting success from fleeting hype. Final Thoughts A strong, adaptable product vision is the foundation of sustainable success. By: Setting a clear yet flexible vision,
Using scenario-based planning,
Regularly revisiting and evolving roadmaps,
Maintaining a balance between adaptability and focus ...you can navigate today’s uncertainties while delivering lasting impact. The best product development teams treat these fundamentals as an ongoing process, not a one-time checklist. With these in place, your product is well-positioned to thrive in competitive markets and lead future innovations. Looking at nature, humankind’s survival through the ages has relied on one crucial trait: adaptability . While products are not living entities, adopting a "living vision" for your product can inspire adaptability, igniting its ability to thrive in a competitive landscape. adaptability Bringing successful products to life in today’s dynamic market demands more than just a great idea. It requires a clear, adaptable vision and a commitment to ongoing improvement. A strong product vision serves as a north star for teams, but its true power lies in its ability to evolve with shifting circumstances while staying aligned with core goals. This article breaks down five essential principles to help you build a strong, adaptable product vision that can withstand market shifts, meet customer needs, and unlock long-term success. 1. Set a Clear and Strong Foundation Your product vision is the foundation for your entire development strategy. A strong vision: Aligns the team around a shared purpose. Clearly defines the problem your product solves. Guides decisions for both short- and long-term growth with focus and clarity. Aligns the team around a shared purpose. Clearly defines the problem your product solves. Guides decisions for both short- and long-term growth with focus and clarity. Take the time to brainstorm, analyze, reflect, and refine. Gather diverse feedback, revisit the vision, and ensure it is solid before moving forward. Conduct thorough research to: Understand your audience and their needs. Define the unique value your product brings. Ensure your vision answers key questions:

What does success look like?
What unique value will this product bring?
How does it effectively address pain points? Understand your audience and their needs. Define the unique value your product brings. Ensure your vision answers key questions: What does success look like?
What unique value will this product bring?
How does it effectively address pain points? What does success look like? What unique value will this product bring? How does it effectively address pain points? What does success look like? What unique value will this product bring? How does it effectively address pain points? Engage stakeholders early to validate assumptions and secure team buy-in. A well-crafted vision is not only a guide but a safeguard against future obstacles. 2. Beyond Solving—Create an Exceptional Experience Listening to customers and responding to their needs is important, but true product success requires more. To truly stand out, you must: Solve existing problems effectively. Anticipate future needs customers might not even be aware of. Solve existing problems effectively. Solve existing problems effectively. Anticipate future needs customers might not even be aware of. Anticipate future needs customers might not even be aware of. Even if customers are unhappy with an existing solution, switching comes at a cost—effort, time, and resources. According to Gartner, 77% of B2B buyers cite switching costs as a significant barrier to change. Your product must offer compelling features that make the switch worthwhile. Apple’s iPhone revolutionized the market by anticipating needs that consumers had not yet considered. By combining communication, productivity, and entertainment, Apple created an indispensable product that now dominates the smartphone market. 3. Ensure Adaptability in Your Vision Markets evolve, customer needs shift, and technology advances rapidly. A rigid product vision can quickly become obsolete. To mitigate this risk: Focus on outcomes rather than specific features. Maintain modular flexibility to pivot when necessary. Foster cross-functional collaboration to enhance innovation. Focus on outcomes rather than specific features. Maintain modular flexibility to pivot when necessary. Foster cross-functional collaboration to enhance innovation. Tesla serves as a prime example of adaptability. Initially focused on making electric cars aspirational, they expanded their vision to include renewable energy solutions, leading to billions in energy revenue in 2023. Their ability to pivot effectively ensured continued growth and market leadership. Preparing for uncertainty is essential. Scenario analysis can help businesses anticipate potential challenges and opportunities. Implement this by: Using “what-if” scenarios to test market shifts. Conducting data-driven simulations to analyze risks Regularly adjusting roadmaps based on scenario outcomes. Using “what-if” scenarios to test market shifts. Conducting data-driven simulations to analyze risks Regularly adjusting roadmaps based on scenario outcomes. 4. Revisit and Evolve Your Vision Frequently A product vision should not remain static. Regularly reassessing ensures alignment with current realities and emerging trends. To keep your vision relevant: Conduct quarterly or biannual reviews of your product vision. Track progress using key performance indicators (KPIs). Integrate customer and stakeholder feedback into revisions. Conduct quarterly or biannual reviews of your product vision. Track progress using key performance indicators (KPIs). Integrate customer and stakeholder feedback into revisions. A static roadmap can lead to stagnation, especially in industries driven by rapid technological advancements. Regular reviews ensure alignment and adaptability, preventing inefficiencies and misalignment across teams. 5. Balance Flexibility with Core Purpose A strong vision must be firm yet adaptable. Strike a balance by: Anchoring decisions to your core mission. Defining non-negotiable elements while allowing room for change. Setting clear deadlines and accountability structures. Anchoring decisions to your core mission. Defining non-negotiable elements while allowing room for change. Setting clear deadlines and accountability structures. Spotify has maintained a clear vision of delivering personalized audio experiences. However, their adaptability enabled them to expand into podcasts, audiobooks, and live audio, significantly growing their market share. A product vision should function like a rubber band—firm enough to maintain its core shape yet flexible enough to stretch in response to change. Lessons from the Tech Industry: Success and Failure Lessons from the Tech Industry: Success and Failure Successful Adaptations: Successful Adaptations: Apple evolved towards sustainability, integrating recycled materials and targeting carbon neutrality. Tesla expanded its vision beyond EVs into renewable energy, securing long-term growth. Apple evolved towards sustainability, integrating recycled materials and targeting carbon neutrality. Tesla expanded its vision beyond EVs into renewable energy, securing long-term growth. Cautionary Tales: Cautionary Tales: Google Glass failed due to a lack of consideration for privacy, design, and practicality. Crypto startups saw 80% failure rates due to issues with scalability, regulations, and trust. Google Glass failed due to a lack of consideration for privacy, design, and practicality. Crypto startups saw 80% failure rates due to issues with scalability, regulations, and trust. Key Takeaway: Key Takeaway: A static or incomplete vision can lead to irrelevance. Flexibility, adaptability, and attention to detail separate lasting success from fleeting hype. Final Thoughts Final Thoughts A strong, adaptable product vision is the foundation of sustainable success. By: Setting a clear yet flexible vision, Using scenario-based planning, Regularly revisiting and evolving roadmaps, Maintaining a balance between adaptability and focus Setting a clear yet flexible vision, Using scenario-based planning, Regularly revisiting and evolving roadmaps, Maintaining a balance between adaptability and focus ...you can navigate today’s uncertainties while delivering lasting impact. The best product development teams treat these fundamentals as an ongoing process, not a one-time checklist. With these in place, your product is well-positioned to thrive in competitive markets and lead future innovations.

This Five-step Framework Can Help You Build a Strong, Adaptable Product

This is Part 7 of a 11-part series based on the research paper “Reinforcement Learning In Agent-based Market Simulation: Unveiling Realistic Stylized Facts And Behavior”. Use the table of links below to navigate to the next part. Table of Links Part 1: Abstract & Introduction Part 2: Important Concepts Part 3: System Description Part 4: Agents & Simulation Details Part 5: Experiment Design Part 6: Continual Learning Part 7: Experiment Results Part 8: Market and Agent Responsiveness to External Events Part 9: Conclusion & References Part 10: Additional Simulation Results Part 11: Simulation Configuration 5 Experiment Results 5.1 Statistical Characteristics of Observed Asset Prices We use Google, Apple, and Amazon’s tick level limit order book data on 2012-06-21 from LOBSTER [28]. We are presenting graphs for Group A only due to page limitations. The continual learning Group A is the most interesting one to analyze. Heavy tails and kurtosis decay. We measure leptokurtic behavior using excess kurtosis which should be significantly larger than 0. Decay means that kurtosis decreases as the sampling frequency of returns decreases (seconds, minutes, hours, etc.). Figure 3 shows the Quantile-Quantile plots of simulations for the continual training RL agents (Group A) and ZI agents respectively, compared with real data. The real data, as well as data obtained using RL and ZI agents, all exhibit strong fat tail return and price distributions. The results from the ZI agents simulation show a milder tail. Results from RL agents closely align with the real data. The average kurtosis from the 10 simulations is presented in Table 1. The table shows the kurtosis decays as data is sampled less frequently. Absence of auto-correlations. [1] describes this property as small insignificant values for auto-correlation of the return time series unless within a very short time interval. Figure 3 shows a strong negative correlation in the first lag and that the auto-correlation function decays to 0 for increasing lags. The tighter boxplots in 3 imply the ZI agents generate a very consistent market, while the real data and RL agents exhibit more variability in this analysis. The author of [1] cites the bounce of market orders between bid and ask prices as the reason why the first lag shows stronger negative auto-correlation. Slow decay of auto-correlation for absolute returns. The auto-correlation function of (1-second) absolute returns decays slowly as a function of the time lags, this implies a long-range time dependency of return magnitude. Figure 4 shows auto-correlation of absolute returns from the simulations using RL agents (Figure 4a) and the ones using ZI agents (Figure 4b). The blue lines in both plots indicate the autocorrelation of the absolute returns of the real data. It can be seen that the auto-correlations from all three methods are decreasing slowly with increasing lags. Figure 4b shows again tighter boxes indicating the simulation using ZI agents is more consistent. Comparing both plots, we can see that RL agents simulate a market in which the return magnitude has a larger autocorrelation than the market simulated by ZI agents. Volatility clustering. Cont et al. [1] use autocorrelation of squared return time series to measure the volatility clustering effect. Figure 5 shows decreasing auto-correlations in both types of simulation. This indicates that they all exhibit volatility clustering effects. Similar to the auto-correlations of absolute returns, the auto-correlation in the ZI-agent simulated market shown in Figure 5b is significantly lower than that of the real data, indicating a less pronounced clustering effect in the ZI-agent simulated market than the RL-agent simulated market. To assess realistic behavior, the MM agents should control their inventory and rely on the Profit and Loss (PnL) from spread (providing liquidity) rather than from inventory (long-term investments). Figure 6 shows the average inventory evolution of all MM agents in the three groups over time. The continual training group has the smallest variation in the inventory, while the untrained group has the largest variation. We further separate the MM’s PnL into profit from spread and profit from holding inventory. The MMs in the continual training group profit on average $1.26 million from spread and lose $0.22 million from inventory. The results for MM agents from other groups are similar. Authors:
(1) Zhiyuan Yao, Stevens Institute of Technology, Hoboken, New Jersey, USA (zyao9@stevens.edu);
(2) Zheng Li, Stevens Institute of Technology, Hoboken, New Jersey, USA (zli149@stevens.edu);
(3) Matthew Thomas, Stevens Institute of Technology, Hoboken, New Jersey, USA (mthomas3@stevens.edu);
(4) Ionut Florescu, Stevens Institute of Technology, Hoboken, New Jersey, USA (ifloresc@stevens.edu). This paper is available on arxiv under CC BY-NC-SA 4.0 DEED license. This is Part 7 of a 11-part series based on the research paper “Reinforcement Learning In Agent-based Market Simulation: Unveiling Realistic Stylized Facts And Behavior”. Use the table of links below to navigate to the next part. This is Part 7 of a 11-part series based on the research paper “ Reinforcement Learning In Agent-based Market Simulation: Unveiling Realistic Stylized Facts And Behavior ”. Use the table of links below to navigate to the next part. Part 7 Reinforcement Learning In Agent-based Market Simulation: Unveiling Realistic Stylized Facts And Behavior Table of Links Part 1: Abstract & Introduction Part 1: Abstract & Introduction Part 2: Important Concepts Part 2: Important Concepts Part 3: System Description Part 3: System Description Part 4: Agents & Simulation Details Part 4: Agents & Simulation Details Part 5: Experiment Design Part 5: Experiment Design Part 6: Continual Learning Part 6: Continual Learning Part 7: Experiment Results Part 7: Experiment Results Part 8: Market and Agent Responsiveness to External Events Part 8: Market and Agent Responsiveness to External Events Part 9: Conclusion & References Part 9: Conclusion & References Part 10: Additional Simulation Results Part 10: Additional Simulation Results Part 11: Simulation Configuration Part 11: Simulation Configuration 5 Experiment Results 5.1 Statistical Characteristics of Observed Asset Prices We use Google, Apple, and Amazon’s tick level limit order book data on 2012-06-21 from LOBSTER [28]. We are presenting graphs for Group A only due to page limitations. The continual learning Group A is the most interesting one to analyze. Heavy tails and kurtosis decay. We measure leptokurtic behavior using excess kurtosis which should be significantly larger than 0. Decay means that kurtosis decreases as the sampling frequency of returns decreases (seconds, minutes, hours, etc.). Figure 3 shows the Quantile-Quantile plots of simulations for the continual training RL agents (Group A) and ZI agents respectively, compared with real data. The real data, as well as data obtained using RL and ZI agents, all exhibit strong fat tail return and price distributions. The results from the ZI agents simulation show a milder tail. Results from RL agents closely align with the real data. The average kurtosis from the 10 simulations is presented in Table 1. The table shows the kurtosis decays as data is sampled less frequently. Heavy tails and kurtosis decay. Absence of auto-correlations. [1] describes this property as small insignificant values for auto-correlation of the return time series unless within a very short time interval. Figure 3 shows a strong negative correlation in the first lag and that the auto-correlation function decays to 0 for increasing lags. The tighter boxplots in 3 imply the ZI agents generate a very consistent market, while the real data and RL agents exhibit more variability in this analysis. The author of [1] cites the bounce of market orders between bid and ask prices as the reason why the first lag shows stronger negative auto-correlation. Absence of auto-correlations. Slow decay of auto-correlation for absolute returns. The auto-correlation function of (1-second) absolute returns decays slowly as a function of the time lags, this implies a long-range time dependency of return magnitude. Figure 4 shows auto-correlation of absolute returns from the simulations using RL agents (Figure 4a) and the ones using ZI agents (Figure 4b). The blue lines in both plots indicate the autocorrelation of the absolute returns of the real data. It can be seen that the auto-correlations from all three methods are decreasing slowly with increasing lags. Figure 4b shows again tighter boxes indicating the simulation using ZI agents is more consistent. Comparing both plots, we can see that RL agents simulate a market in which the return magnitude has a larger autocorrelation than the market simulated by ZI agents. Slow decay of auto-correlation for absolute returns. Volatility clustering. Cont et al. [1] use autocorrelation of squared return time series to measure the volatility clustering effect. Figure 5 shows decreasing auto-correlations in both types of simulation. This indicates that they all exhibit volatility clustering effects. Similar to the auto-correlations of absolute returns, the auto-correlation in the ZI-agent simulated market shown in Figure 5b is significantly lower than that of the real data, indicating a less pronounced clustering effect in the ZI-agent simulated market than the RL-agent simulated market. Volatility clustering. To assess realistic behavior, the MM agents should control their inventory and rely on the Profit and Loss (PnL) from spread (providing liquidity) rather than from inventory (long-term investments). Figure 6 shows the average inventory evolution of all MM agents in the three groups over time. The continual training group has the smallest variation in the inventory, while the untrained group has the largest variation. We further separate the MM’s PnL into profit from spread and profit from holding inventory. The MMs in the continual training group profit on average $1.26 million from spread and lose $0.22 million from inventory. The results for MM agents from other groups are similar. Authors: (1) Zhiyuan Yao, Stevens Institute of Technology, Hoboken, New Jersey, USA (zyao9@stevens.edu); (2) Zheng Li, Stevens Institute of Technology, Hoboken, New Jersey, USA (zli149@stevens.edu); (3) Matthew Thomas, Stevens Institute of Technology, Hoboken, New Jersey, USA (mthomas3@stevens.edu); (4) Ionut Florescu, Stevens Institute of Technology, Hoboken, New Jersey, USA (ifloresc@stevens.edu). Authors: Authors: (1) Zhiyuan Yao, Stevens Institute of Technology, Hoboken, New Jersey, USA (zyao9@stevens.edu); (2) Zheng Li, Stevens Institute of Technology, Hoboken, New Jersey, USA (zli149@stevens.edu); (3) Matthew Thomas, Stevens Institute of Technology, Hoboken, New Jersey, USA (mthomas3@stevens.edu); (4) Ionut Florescu, Stevens Institute of Technology, Hoboken, New Jersey, USA (ifloresc@stevens.edu). This paper is available on arxiv under CC BY-NC-SA 4.0 DEED license. This paper is available on arxiv under CC BY-NC-SA 4.0 DEED license. available on arxiv

RL Agents Show Superior Realism and Adaptability Over Zero-Intelligence Agents

This is Part 3 of a 11-part series based on the research paper “Reinforcement Learning In Agent-based Market Simulation: Unveiling Realistic Stylized Facts And Behavior”. Use the table of links below to navigate to the next part. Table of Links Part 1: Abstract & Introduction Part 2: Important Concepts Part 3: System Description Part 4: Agents & Simulation Details Part 5: Experiment Design Part 6: Continual Learning Part 7: Experiment Results Part 8: Market and Agent Responsiveness to External Events Part 9: Conclusion & References Part 10: Additional Simulation Results Part 11: Simulation Configuration 3. System and Agents 3.1 System Description The system contains a machine engine that organizes LOBs and settles trades, as well as a brokerage center that keeps track of each agent’s account, including the agent’s buying power and assets. All agents place market and limit orders to the matching engine through their brokerage accounts. The matching engine runs a CDA market model. The engine updates the latest LOB information and streams its state to each trading agent in real time. The agents in this system are of two types: liquidity-taking (LT) agents and market-making (MM) agents. Each instance of these agents is formulated as an RL agent, each with its own parameters and reward function. Each agent observes the system independently, selects actions, receives feedback, and optimizes its own strategy. Each agent learns to adapt its strategy through actions (orders submitted) and feedback received (reward). The formulation of rewards is different for each agent, we provide details in the next section. We highlight two aspects of our work which we think helped improve the realism of the simulation compared to prior work. First, all agents run in their own respective threads, thus all threads run in parallel and are not waiting for any other thread once they are launched. Second, all agents are heterogeneous. Even though some agents belong to the same category, they use different sets of hyperparameters, and this results in significantly different behavior for each agent. Authors:
(1) Zhiyuan Yao, Stevens Institute of Technology, Hoboken, New Jersey, USA (zyao9@stevens.edu);
(2) Zheng Li, Stevens Institute of Technology, Hoboken, New Jersey, USA (zli149@stevens.edu);
(3) Matthew Thomas, Stevens Institute of Technology, Hoboken, New Jersey, USA (mthomas3@stevens.edu);
(4) Ionut Florescu, Stevens Institute of Technology, Hoboken, New Jersey, USA (ifloresc@stevens.edu). This paper is available on arxiv under CC BY-NC-SA 4.0 DEED license. This is Part 3 of a 11-part series based on the research paper “Reinforcement Learning In Agent-based Market Simulation: Unveiling Realistic Stylized Facts And Behavior”. Use the table of links below to navigate to the next part. This is Part 3 of a 11-part series based on the research paper “ Reinforcement Learning In Agent-based Market Simulation: Unveiling Realistic Stylized Facts And Behavior ”. Use the table of links below to navigate to the next part. Part 3 Reinforcement Learning In Agent-based Market Simulation: Unveiling Realistic Stylized Facts And Behavior Table of Links Part 1: Abstract & Introduction Part 1: Abstract & Introduction Part 2: Important Concepts Part 2: Important Concepts Part 3: System Description Part 3: System Description Part 4: Agents & Simulation Details Part 4: Agents & Simulation Details Part 5: Experiment Design Part 5: Experiment Design Part 6: Continual Learning Part 6: Continual Learning Part 7: Experiment Results Part 7: Experiment Results Part 8: Market and Agent Responsiveness to External Events Part 8: Market and Agent Responsiveness to External Events Part 9: Conclusion & References Part 9: Conclusion & References Part 10: Additional Simulation Results Part 10: Additional Simulation Results Part 11: Simulation Configuration Part 11: Simulation Configuration 3. System and Agents 3.1 System Description The system contains a machine engine that organizes LOBs and settles trades, as well as a brokerage center that keeps track of each agent’s account, including the agent’s buying power and assets. All agents place market and limit orders to the matching engine through their brokerage accounts. The matching engine runs a CDA market model. The engine updates the latest LOB information and streams its state to each trading agent in real time. The agents in this system are of two types: liquidity-taking (LT) agents and market-making (MM) agents. Each instance of these agents is formulated as an RL agent, each with its own parameters and reward function. Each agent observes the system independently, selects actions, receives feedback, and optimizes its own strategy. Each agent learns to adapt its strategy through actions (orders submitted) and feedback received (reward). The formulation of rewards is different for each agent, we provide details in the next section. We highlight two aspects of our work which we think helped improve the realism of the simulation compared to prior work. First, all agents run in their own respective threads, thus all threads run in parallel and are not waiting for any other thread once they are launched. Second, all agents are heterogeneous. Even though some agents belong to the same category, they use different sets of hyperparameters, and this results in significantly different behavior for each agent. highlight two aspects of our work Authors: (1) Zhiyuan Yao, Stevens Institute of Technology, Hoboken, New Jersey, USA (zyao9@stevens.edu); (2) Zheng Li, Stevens Institute of Technology, Hoboken, New Jersey, USA (zli149@stevens.edu); (3) Matthew Thomas, Stevens Institute of Technology, Hoboken, New Jersey, USA (mthomas3@stevens.edu); (4) Ionut Florescu, Stevens Institute of Technology, Hoboken, New Jersey, USA (ifloresc@stevens.edu). Authors: Authors: (1) Zhiyuan Yao, Stevens Institute of Technology, Hoboken, New Jersey, USA (zyao9@stevens.edu); (2) Zheng Li, Stevens Institute of Technology, Hoboken, New Jersey, USA (zli149@stevens.edu); (3) Matthew Thomas, Stevens Institute of Technology, Hoboken, New Jersey, USA (mthomas3@stevens.edu); (4) Ionut Florescu, Stevens Institute of Technology, Hoboken, New Jersey, USA (ifloresc@stevens.edu). This paper is available on arxiv under CC BY-NC-SA 4.0 DEED license. This paper is available on arxiv under CC BY-NC-SA 4.0 DEED license. available on arxiv

Reinforcement Learning Simulation Features Realism and Adaptability 

This is Part 10 of a 12-part series based on the research paper “Human-Machine Social Systems.” Use the table of links below to navigate to the next part. Table of Links Abstract and Introduction Human-machine interactions Collective outcomes Box 1: Competition in high-frequency trading markets Box 2: Contagion on Twitter Box 3: Cooperation and coordination on Wikipedia Box 4: Cooperation and contagion on Reddit Discussion Implications for research Implications for design Implications for policy Conclusion, Acknowledgments, References, and Competing interests Implications for Design Humans are resilient and successful when they exhibit high levels of efficient communication, context awareness, emotional recognition and response, and ethical and cultural sensitivities. These features should be encouraged when designing social machines to build trust, ensure legal compliance, promote social harmony, enhance user satisfaction, and achieve long-term sustainability [231,232]. Since H-H, H-M, and M-M interactions differ, machines should be specifically designed for each scenario, with separate training for each interaction. This could avoid market underperformance when bargaining algorithms trained on human-only markets adapt poorly to human-machine negotiations[64], or traffic jams when driverless vehicles trained on human driving fail to properly interact with each other [233]. AI design should also adopt a hierarchy of behavioral rules and conventions guiding H-M and MM interactions in the context of H-H interactions. Isaac Asimov’s famous Three Laws of Robotics [234] which regulate M-H interactions and self-preservation—1) not harming humans, 2) obeying humans, and 3) protecting own existence, with priority given to higher order rules—could be adapted for M-M interactions, considering specific contexts, implications, and unintended consequences [235]. Further, cultural context is crucial for AI design. People’s perceptions of machines vary by age, environment, personality, and geography [236,60]. Machines reflect their developers’ culture and operate in settings with specific organizational and community norms [36]. Thus, AI design for self-driving cars and assistant bots should consider local driving culture and attitudes toward domestic assistants during training. Finally, the systems under review exhibit complexity in that their emergent behavior transcends a mere aggregation of individual components, displaying what is known as “network effects.” While natural complex systems often demonstrate remarkable resilience and adaptivity [237], the human-designed complex systems discussed here, albeit to varying degrees, are not inherently adaptive. To enhance resilience and robustness, AI designers should incorporate complex adaptive system principles, like negative feedback, modularity, and hierarchical organization. For example, dense networks lacking diversity and modularity are susceptible to systemic failures [238,239,240] and are easy to control [241]. Introducing bots that increase network diversity, introduce resistance, build resilience, or incorporate negative feedback could enhance adaptability and stability. Such configurations have improved group outcomes in forecasting [242], exploration/exploitation[243], and general knowledge tasks [151]. Literature on monitoring and steering complex systems can help predict critical transitions [244,245,15], designing safer human-machine systems across different environments and perturbations. Authors:
(1) Milena Tsvetkova, Department of Methodology, London School of Economics and Political Science, London, United Kingdom;
(2) Taha Yasseri, School of Sociology, University College Dublin, Dublin, Ireland and Geary Institute for Public Policy, University College Dublin, Dublin, Ireland;
(3) Niccolo Pescetelli, Collective Intelligence Lab, New Jersey Institute of Technology, Newark, New Jersey, USA;
(4) Tobias Werner, Center for Humans and Machines, Max Planck Institute for Human Development, Berlin, Germany. This paper is available on arxiv under CC BY 4.0 DEED license. This is Part 10 of a 12-part series based on the research paper “Human-Machine Social Systems.” Use the table of links below to navigate to the next part. This is Part 10 of a 12-part series based on the research paper “Human-Machine Social Systems.” Use the table of links below to navigate to the next part. “ Human-Machine Social Systems .” Human-Machine Social Systems Table of Links Abstract and Introduction Abstract and Introduction Human-machine interactions Human-machine interactions Collective outcomes Collective outcomes Box 1: Competition in high-frequency trading markets Box 1: Competition in high-frequency trading markets Box 2: Contagion on Twitter Box 2: Contagion on Twitter Box 3: Cooperation and coordination on Wikipedia Box 3: Cooperation and coordination on Wikipedia Box 4: Cooperation and contagion on Reddit Box 4: Cooperation and contagion on Reddit Discussion Discussion Implications for research Implications for research Implications for design Implications for design Implications for policy Implications for policy Conclusion, Acknowledgments, References, and Competing interests Conclusion, Acknowledgments, References, and Competing interests Implications for Design Humans are resilient and successful when they exhibit high levels of efficient communication, context awareness, emotional recognition and response, and ethical and cultural sensitivities. These features should be encouraged when designing social machines to build trust, ensure legal compliance, promote social harmony, enhance user satisfaction, and achieve long-term sustainability [231,232]. Since H-H, H-M, and M-M interactions differ, machines should be specifically designed for each scenario, with separate training for each interaction. This could avoid market underperformance when bargaining algorithms trained on human-only markets adapt poorly to human-machine negotiations[64], or traffic jams when driverless vehicles trained on human driving fail to properly interact with each other [233]. AI design should also adopt a hierarchy of behavioral rules and conventions guiding H-M and MM interactions in the context of H-H interactions. Isaac Asimov’s famous Three Laws of Robotics [234] which regulate M-H interactions and self-preservation—1) not harming humans, 2) obeying humans, and 3) protecting own existence, with priority given to higher order rules—could be adapted for M-M interactions, considering specific contexts, implications, and unintended consequences [235]. Further, cultural context is crucial for AI design. People’s perceptions of machines vary by age, environment, personality, and geography [236,60]. Machines reflect their developers’ culture and operate in settings with specific organizational and community norms [36]. Thus, AI design for self-driving cars and assistant bots should consider local driving culture and attitudes toward domestic assistants during training. Finally, the systems under review exhibit complexity in that their emergent behavior transcends a mere aggregation of individual components, displaying what is known as “network effects.” While natural complex systems often demonstrate remarkable resilience and adaptivity [237], the human-designed complex systems discussed here, albeit to varying degrees, are not inherently adaptive. To enhance resilience and robustness, AI designers should incorporate complex adaptive system principles, like negative feedback, modularity, and hierarchical organization. For example, dense networks lacking diversity and modularity are susceptible to systemic failures [238,239,240] and are easy to control [241]. Introducing bots that increase network diversity, introduce resistance, build resilience, or incorporate negative feedback could enhance adaptability and stability. Such configurations have improved group outcomes in forecasting [242], exploration/exploitation[243], and general knowledge tasks [151]. Literature on monitoring and steering complex systems can help predict critical transitions [244,245,15], designing safer human-machine systems across different environments and perturbations. Authors: (1) Milena Tsvetkova, Department of Methodology, London School of Economics and Political Science, London, United Kingdom; (2) Taha Yasseri, School of Sociology, University College Dublin, Dublin, Ireland and Geary Institute for Public Policy, University College Dublin, Dublin, Ireland; (3) Niccolo Pescetelli, Collective Intelligence Lab, New Jersey Institute of Technology, Newark, New Jersey, USA; (4) Tobias Werner, Center for Humans and Machines, Max Planck Institute for Human Development, Berlin, Germany. Authors: Authors: (1) Milena Tsvetkova, Department of Methodology, London School of Economics and Political Science, London, United Kingdom; (2) Taha Yasseri, School of Sociology, University College Dublin, Dublin, Ireland and Geary Institute for Public Policy, University College Dublin, Dublin, Ireland; (3) Niccolo Pescetelli, Collective Intelligence Lab, New Jersey Institute of Technology, Newark, New Jersey, USA; (4) Tobias Werner, Center for Humans and Machines, Max Planck Institute for Human Development, Berlin, Germany. This paper is available on arxiv under CC BY 4.0 DEED license. This paper is available on arxiv under CC BY 4.0 DEED license. available on arxiv

AI Design Must Prioritize Adaptation to Human Culture, Trust, and Ethics 

The robotics industry is experiencing a profound transformation, fueled by artificial intelligence (AI). Traditional robots, which rely on rigid programming, are evolving into adaptive systems capable of responding to dynamic environments. I spoke with Ismail Karankin, the founder of Oli Robotics, a startup developing reinforcement learning and LLM-based software for industrial robot manipulators. Their mission: to revolutionize workplace safety by delegating hazardous tasks to robots. How did you start implementing AI in manufacturing? What inspired the creation of Oli Robotics? I founded Oli Robotics because I saw immense potential at the intersection of AI and robotics. The falling costs of AI technologies, driven by advancements in GPUs and tensor processing units (TPUs), have made robotics more accessible. At the same time, industrial robot prices are dropping, with over 200 brands of robotic manipulators emerging in China over the past five years, intensifying competition. However, software for these manipulators is still stuck in the past. Most rely on trajectory planning and sensor data. Programming a six-axis manipulator involves calibrating its coordinates, setting angular constraints, and designing motion trajectories—an inflexible process suited only to static environments. Tesla encountered this issue in 2018 when scaling production: minor changes in manufacturing required costly reprogramming of robotic systems. With AI, robots can adapt in real-time, adjusting their actions based on sensor feedback. Instead of pre-calculating fixed trajectories, our system dynamically recalibrates, improving efficiency and minimizing human intervention. Over time, robots learn to optimize their actions, improving assembly quality without manual reprogramming. At Oli Robotics, we are creating adaptive software that reacts to environmental changes, enabling robots to adjust their movements on the fly. While this was once cost-prohibitive due to computational demands, the decreasing costs of LLMs and AI infrastructure are turning this vision into reality. What opportunities does AI bring to manufacturers? AI is already revolutionizing industries by predicting equipment failures and identifying defects in production lines. But the most significant disruption, I believe, will come from humanoid robots. Unlike specialized robots, humanoids are designed to interact with existing tools and environments, making them highly versatile. Their human-like form enables them to work in conditions where human skills are typically required. Beyond manufacturing, humanoids can integrate into almost every aspect of human activity. The cost of humanoid robots is also becoming more accessible. For instance, Unitree’s G1 humanoid robot starts at $20,000, making them a feasible investment for many industries. You previously founded a drone delivery startup. What lessons did you learn from that venture? In 2016, I launched Airpost, a startup focused on delivering small parcels using drones. We even made it into Startupbootcamp as one of the top 1,100 projects. The idea was simple: drones could deliver packages up to 4 kg faster by avoiding traffic and eliminating the need for a pilot. However, we faced significant regulatory hurdles. At the time, drones weighing over 450 grams required registration, and fully autonomous drones were prohibited. For example, a client—a biolab in a smaller town in Eastern Europe—wanted to use drones to quickly transport samples to nearby capitol city. But the legal landscape made such operations impractical. One amusing yet sobering incident involved a request for pinpoint delivery accuracy within one meter. The caller eventually revealed they intended to use the drones to smuggle phones into a restricted area. Naturally, we declined the offer. After a year of challenges, we decided to close the startup. Do you see risks in the increasing use of AI technologies? One common concern is that humanoid robots could replace human jobs. However, the reality has often been the opposite. In the U.S., for example, companies are hiring more people to prepare training data for AI models. The real bottleneck in AI development is the availability of high-quality data. While the internet seems vast, the amount of suitable training data is finite. Companies are now investing heavily in creating and curating datasets to fuel the next generation of AI models. What challenges and milestones have you faced so far with Oli Robotics? The biggest challenge is convincing manufacturers to adopt adaptive robotics. Traditional robots excel in static, predictable environments, but they struggle with variability. With our software, robots learn and improve in real-time, making them suitable for dynamic tasks. One of our most promising collaborations is with a bakery in San Jose. Employees currently face high risks of burns when loading and unloading bread due to extreme temperatures. Our adaptive robot manipulators are designed to handle such tasks, ensuring both safety and efficiency. We’ve also been recognized by the Silicon Valley Residency accelerator as one of the top 30 startups from Central Asia. This validation, along with interest from five potential clients, reinforces our belief in the transformative power of our technology. What do you think is the future of robotics in the workplace? I believe we’re on the brink of a robotics revolution. Adaptive AI-driven robots will replace humans in hazardous roles, reducing workplace injuries. At the same time, advancements in humanoid robots will expand their use cases, making them invaluable in industries ranging from healthcare to logistics. Ultimately, AI and robotics aren’t just about replacing humans—they’re about augmenting human capabilities, creating safer workplaces, and enabling industries to operate more efficiently. The robotics industry is experiencing a profound transformation, fueled by artificial intelligence (AI). Traditional robots, which rely on rigid programming, are evolving into adaptive systems capable of responding to dynamic environments. I spoke with Ismail Karankin, the founder of Oli Robotics , a startup developing reinforcement learning and LLM-based software for industrial robot manipulators. Their mission: to revolutionize workplace safety by delegating hazardous tasks to robots. Oli Robotics How did you start implementing AI in manufacturing? What inspired the creation of Oli Robotics? How did you start implementing AI in manufacturing? What inspired the creation of Oli Robotics? I founded Oli Robotics because I saw immense potential at the intersection of AI and robotics. The falling costs of AI technologies, driven by advancements in GPUs and tensor processing units (TPUs), have made robotics more accessible. At the same time, industrial robot prices are dropping, with over 200 brands of robotic manipulators emerging in China over the past five years, intensifying competition. However, software for these manipulators is still stuck in the past. Most rely on trajectory planning and sensor data. Programming a six-axis manipulator involves calibrating its coordinates, setting angular constraints, and designing motion trajectories—an inflexible process suited only to static environments. Tesla encountered this issue in 2018 when scaling production: minor changes in manufacturing required costly reprogramming of robotic systems. With AI, robots can adapt in real-time, adjusting their actions based on sensor feedback. Instead of pre-calculating fixed trajectories, our system dynamically recalibrates, improving efficiency and minimizing human intervention. Over time, robots learn to optimize their actions, improving assembly quality without manual reprogramming. At Oli Robotics, we are creating adaptive software that reacts to environmental changes, enabling robots to adjust their movements on the fly. While this was once cost-prohibitive due to computational demands, the decreasing costs of LLMs and AI infrastructure are turning this vision into reality. What opportunities does AI bring to manufacturers? What opportunities does AI bring to manufacturers? AI is already revolutionizing industries by predicting equipment failures and identifying defects in production lines. But the most significant disruption, I believe, will come from humanoid robots. Unlike specialized robots, humanoids are designed to interact with existing tools and environments, making them highly versatile. Their human-like form enables them to work in conditions where human skills are typically required. Beyond manufacturing, humanoids can integrate into almost every aspect of human activity. The cost of humanoid robots is also becoming more accessible. For instance, Unitree’s G1 humanoid robot starts at $20,000, making them a feasible investment for many industries. You previously founded a drone delivery startup. What lessons did you learn from that venture? You previously founded a drone delivery startup. What lessons did you learn from that venture? In 2016, I launched Airpost, a startup focused on delivering small parcels using drones. We even made it into Startupbootcamp as one of the top 1,100 projects. The idea was simple: drones could deliver packages up to 4 kg faster by avoiding traffic and eliminating the need for a pilot. However, we faced significant regulatory hurdles. At the time, drones weighing over 450 grams required registration, and fully autonomous drones were prohibited. For example, a client—a biolab in a smaller town in Eastern Europe—wanted to use drones to quickly transport samples to nearby capitol city. But the legal landscape made such operations impractical. One amusing yet sobering incident involved a request for pinpoint delivery accuracy within one meter. The caller eventually revealed they intended to use the drones to smuggle phones into a restricted area. Naturally, we declined the offer. After a year of challenges, we decided to close the startup. Do you see risks in the increasing use of AI technologies? Do you see risks in the increasing use of AI technologies? One common concern is that humanoid robots could replace human jobs. However, the reality has often been the opposite. In the U.S., for example, companies are hiring more people to prepare training data for AI models. The real bottleneck in AI development is the availability of high-quality data. While the internet seems vast, the amount of suitable training data is finite. Companies are now investing heavily in creating and curating datasets to fuel the next generation of AI models. What challenges and milestones have you faced so far with Oli Robotics? What challenges and milestones have you faced so far with Oli Robotics? The biggest challenge is convincing manufacturers to adopt adaptive robotics. Traditional robots excel in static, predictable environments, but they struggle with variability. With our software, robots learn and improve in real-time, making them suitable for dynamic tasks. One of our most promising collaborations is with a bakery in San Jose. Employees currently face high risks of burns when loading and unloading bread due to extreme temperatures. Our adaptive robot manipulators are designed to handle such tasks, ensuring both safety and efficiency. We’ve also been recognized by the Silicon Valley Residency accelerator as one of the top 30 startups from Central Asia. This validation, along with interest from five potential clients, reinforces our belief in the transformative power of our technology. Silicon Valley Residency accelerator What do you think is the future of robotics in the workplace? What do you think is the future of robotics in the workplace? I believe we’re on the brink of a robotics revolution. Adaptive AI-driven robots will replace humans in hazardous roles, reducing workplace injuries. At the same time, advancements in humanoid robots will expand their use cases, making them invaluable in industries ranging from healthcare to logistics. Ultimately, AI and robotics aren’t just about replacing humans—they’re about augmenting human capabilities, creating safer workplaces, and enabling industries to operate more efficiently.

The Future of Robotics: AI-Powered Adaptation for Safer Workplaces

Table of Links Abstract and 1. Introduction Why ELIZA?
The Intelligence Engineers
Newell, Shaw, and Simon’s IPL Logic Theorist: The First True AIs
From IPL to SLIP and Lisp
A Critical Tangent into Gomoku
Interpretation is the Core of Intelligence
The Threads Come Together: Interpretation, Language, Lists, Graphs, and Recursion
Finally ELIZA: A Platform, Not a Chat Bot!
A Perfect Irony: A Lisp ELIZA Escapes and is Misinterpreted by the AI Community
Another Wave: A BASIC ELIZA turns the PC Generation on to AI
Conclusion: A certain danger lurks there
Acknowledgements and References 10 A Perfect Irony: A Lisp ELIZA Escapes and is Misinterpreted by the AI Community Our thesis is that Joseph Weizenbaum, afraid of people misinterpreting computers and AI, developed ELIZA not as an AI, but as a platform to support research about people’s interpretive process, perhaps more specifically their interaction with AIs. Weizenbaum, of course, had to provide an example – the DOCTOR script – to demonstrate the platform, but, as above, he had no illusion that it was actually intelligent, nor even very easily interpreted as intelligent; Much more “intelligent” programs already existing in various game playing programs (even his own, from the Datamation paper!), in Colby’s Parry, and in other programs that solved math word problems[17], did database lookups from natural language queries[29], and others. Although some of the sort of research envisioned by Weizenbaum was carried out by the MGH team, a specific event thwarted this use case for ELIZA, and, in almost the deepest possible irony, led to exactly the conflation of ELIZA with AI that Weizenbaum set out to study, not to bring into being. Weizenbaum built his ELIZA in MAD-SLIP on the IBM 7090, which was the primary machine at MIT’s Project MAC, on the 5th-through-9th floors of Tech Square. Almost immediately after ELIZA’s publication in 1966, Bernie Cosell created a Lisp knock-off of ELIZA, based on the algorithm and DOCTOR script in Weizenbaum’s paper. Cosell worked at Bolt Beranek and Newman (BBN), a large RAND-like consulting company situated nearby MIT, that commonly hired MIT graduates. Coincidentally, at almost the same time, BBN was a founding site of the ARPAnet[14], and Cosell’s Lisp ELIZA diffused rapidly through that network and across the soon-to-be Lisp-centered world of academic AI. As a result, Cosell’s Lisp ELIZA rapidly became the dominant strain, and Weizenbaum’s MAD-SLIP version, inaccessible on the ARPAnet, was almost instantly forgotten. Through interesting and complex criss-crossing histories, described in detail in [45], Danny Bobrow, a recent MIT AI graduate who headed up BBN’s AI program, brought McCarthy, and thus Lisp, to BBN, where Cosell was exposed to it, and built an ELIZA knock-off in Lisp as a side-project.[15] Cosell reports: “When I was working on the PDP-1 time-sharing system [...] I thought I would learn Lisp. That spring, Joe Weizenbaum had written an article for Communications of the ACM on ELIZA. I thought that was way cool. [...] He described how ELIZA works and I said, “I bet I could write something to do that.” And so I started writing a Lisp program on [the] PDP-1 system at BBN.”[41, p. 540] He continues, “I wrote that program and got it up and working. Playing with it was an all-BBN project. [...] It was written, at first, in the PDP-1 Lisp. But they were building a Lisp on the PDP-6 at that point—or maybe the PDP-10. But it was the Lisp that had spread across the ARPANet. So [ELIZA] went along with it [...].”[41, p. 541] Once Cosell’s Lisp ELIZA hit the academic world via the rapid spread of the ARPANet, Weizenbaum’s MAD-SLIP version was no longer relevant, and the name “ELIZA” (and the “DOCTOR” script), as well as the concept, was, from that point forward, associated with Cosell’s Lisp version, although its origin was still correctly attributed to Weizenbaum via the CACM paper, leading to a 50- year-long, community-wide misapprehension that ELIZA had been written in Lisp. In addition to being promulgated by Cosell’s knock-off being the one that was most easily available, via the ARPANet, it was a natural confusion because Lisp was rapidly becoming the go-to language of AI. Once Lisp came onto the scene no one thought much again about SLIP, or, for that matter, IPL-V. Author:
(1) Jeff Shrager, Blue Dot Change and Stanford University Symbolic Systems Program (Adjunct)( jshrager@stanford.edu). This paper is available on arxiv under CC BY 4.0 license. [14] At the time what became the internet, and then the web, was called the ARPAnet, and BBN built its core hardware and developed the programs that ran it.[11]) [15] Coincidentally, Bobrow was also the author of STUDENT, the program mentioned above that solved math word problems.[17] Table of Links Abstract and 1. Introduction Abstract and 1. Introduction Why ELIZA? The Intelligence Engineers Newell, Shaw, and Simon’s IPL Logic Theorist: The First True AIs From IPL to SLIP and Lisp A Critical Tangent into Gomoku Interpretation is the Core of Intelligence The Threads Come Together: Interpretation, Language, Lists, Graphs, and Recursion Finally ELIZA: A Platform, Not a Chat Bot! A Perfect Irony: A Lisp ELIZA Escapes and is Misinterpreted by the AI Community Another Wave: A BASIC ELIZA turns the PC Generation on to AI Conclusion: A certain danger lurks there Acknowledgements and References Why ELIZA? Why ELIZA? The Intelligence Engineers The Intelligence Engineers Newell, Shaw, and Simon’s IPL Logic Theorist: The First True AIs Newell, Shaw, and Simon’s IPL Logic Theorist: The First True AIs From IPL to SLIP and Lisp From IPL to SLIP and Lisp A Critical Tangent into Gomoku A Critical Tangent into Gomoku Interpretation is the Core of Intelligence Interpretation is the Core of Intelligence The Threads Come Together: Interpretation, Language, Lists, Graphs, and Recursion The Threads Come Together: Interpretation, Language, Lists, Graphs, and Recursion Finally ELIZA: A Platform, Not a Chat Bot! Finally ELIZA: A Platform, Not a Chat Bot! A Perfect Irony: A Lisp ELIZA Escapes and is Misinterpreted by the AI Community A Perfect Irony: A Lisp ELIZA Escapes and is Misinterpreted by the AI Community Another Wave: A BASIC ELIZA turns the PC Generation on to AI Another Wave: A BASIC ELIZA turns the PC Generation on to AI Conclusion: A certain danger lurks there Conclusion: A certain danger lurks there Acknowledgements and References Acknowledgements and References 10 A Perfect Irony: A Lisp ELIZA Escapes and is Misinterpreted by the AI Community Our thesis is that Joseph Weizenbaum, afraid of people misinterpreting computers and AI, developed ELIZA not as an AI, but as a platform to support research about people’s interpretive process, perhaps more specifically their interaction with AIs. Weizenbaum, of course, had to provide an example – the DOCTOR script – to demonstrate the platform, but, as above, he had no illusion that it was actually intelligent, nor even very easily interpreted as intelligent; Much more “intelligent” programs already existing in various game playing programs (even his own, from the Datamation paper!), in Colby’s Parry, and in other programs that solved math word problems[17], did database lookups from natural language queries[29], and others. Although some of the sort of research envisioned by Weizenbaum was carried out by the MGH team, a specific event thwarted this use case for ELIZA, and, in almost the deepest possible irony, led to exactly the conflation of ELIZA with AI that Weizenbaum set out to study, not to bring into being. Weizenbaum built his ELIZA in MAD-SLIP on the IBM 7090, which was the primary machine at MIT’s Project MAC, on the 5th-through-9th floors of Tech Square. Almost immediately after ELIZA’s publication in 1966, Bernie Cosell created a Lisp knock-off of ELIZA, based on the algorithm and DOCTOR script in Weizenbaum’s paper. Cosell worked at Bolt Beranek and Newman (BBN), a large RAND-like consulting company situated nearby MIT, that commonly hired MIT graduates. Coincidentally, at almost the same time, BBN was a founding site of the ARPAnet[14], and Cosell’s Lisp ELIZA diffused rapidly through that network and across the soon-to-be Lisp-centered world of academic AI. As a result, Cosell’s Lisp ELIZA rapidly became the dominant strain, and Weizenbaum’s MAD-SLIP version, inaccessible on the ARPAnet, was almost instantly forgotten. Through interesting and complex criss-crossing histories, described in detail in [45], Danny Bobrow, a recent MIT AI graduate who headed up BBN’s AI program, brought McCarthy, and thus Lisp, to BBN, where Cosell was exposed to it, and built an ELIZA knock-off in Lisp as a side-project.[15] Cosell reports: “When I was working on the PDP-1 time-sharing system [...] I thought I would learn Lisp. That spring, Joe Weizenbaum had written an article for Communications of the ACM on ELIZA. I thought that was way cool. [...] He described how ELIZA works and I said, “I bet I could write something to do that.” And so I started writing a Lisp program on [the] PDP-1 system at BBN.”[41, p. 540] He continues, “I wrote that program and got it up and working. Playing with it was an all-BBN project. [...] It was written, at first, in the PDP-1 Lisp. But they were building a Lisp on the PDP-6 at that point—or maybe the PDP-10. But it was the Lisp that had spread across the ARPANet. So [ELIZA] went along with it [...].”[41, p. 541] Once Cosell’s Lisp ELIZA hit the academic world via the rapid spread of the ARPANet, Weizenbaum’s MAD-SLIP version was no longer relevant, and the name “ELIZA” (and the “DOCTOR” script), as well as the concept, was, from that point forward, associated with Cosell’s Lisp version, although its origin was still correctly attributed to Weizenbaum via the CACM paper, leading to a 50- year-long, community-wide misapprehension that ELIZA had been written in Lisp. In addition to being promulgated by Cosell’s knock-off being the one that was most easily available, via the ARPANet, it was a natural confusion because Lisp was rapidly becoming the go-to language of AI. Once Lisp came onto the scene no one thought much again about SLIP, or, for that matter, IPL-V. Author: (1) Jeff Shrager, Blue Dot Change and Stanford University Symbolic Systems Program (Adjunct)( jshrager@stanford.edu). Author: Author: (1) Jeff Shrager, Blue Dot Change and Stanford University Symbolic Systems Program (Adjunct)( jshrager@stanford.edu). This paper is available on arxiv under CC BY 4.0 license. This paper is available on arxiv under CC BY 4.0 license. available on arxiv available on arxiv [14] At the time what became the internet, and then the web, was called the ARPAnet, and BBN built its core hardware and developed the programs that ran it.[11]) [15] Coincidentally, Bobrow was also the author of STUDENT, the program mentioned above that solved math word problems.[17]

The Accidental AI: How ELIZA's Lisp Adaptation Derailed Its Original Research Intent

The HackerNoon editorial team has launched this interview series with women in tech to celebrate their achievements and share their struggles. We need more women in technology, and by sharing stories, we can encourage many girls to follow their dreams. Share your story today! Tell us about yourself! Hey-hey! I’m Evgeniia Megrian, currently working as a Sales Hunter at Semrush, and previously at SLSBMB. My primary responsibility is searching for and hiring global sales talents for IT/Tech solutions. Over the past 3 years, I’ve successfully hired over 60 senior B2B sales professionals from tech giants like Google, Snap, and Salesforce. My journey began with an Accelerator program at uni, where I managed an EdTech SaaS designed to simplify learning complex subjects. After a successful launch but an unsuccessful work-life balance, I shifted to business development as an employee. Eventually, I transitioned to a sales hunter role, applying the same strategic approach to recruiting top talent and forming successful partnerships. Why did you choose this field in the first place? Early in my career, I joined an Accelerator program that introduced me to the startup world and the IT/Tech fields. This was my first real exposure to this dynamic industry. I loved the fast-paced environment, where we could test ideas quickly, be flexible, and create new processes instead of following strict rules. This approach matched really well with my values. I appreciated the opportunity to grow rapidly, try a dozen things, and see quick results. Through these experiences, I realized this was the field I wanted to develop in, and it also helped me advance quickly in my career. What tech are you most excited/passionate about right now and why? Currently, I’m super excited about fully implementing outbound strategies for hiring global talent. This approach allows me to reach not only candidates who apply but also those who are potentially relevant, greatly expanding our pool of prospects. It makes the hiring process more controlled and predictable. For instance, if we know exactly how many candidates we need to reach out to in order to hire a professional within a certain timeframe, the entire process becomes much more manageable. Additionally, I’m passionate about using automation solutions to create more effective outreach campaigns and build recruitment funnels. These technologies streamline the hiring process, making it more efficient and scalable. What tech are you most worried about right now and why? I'm kind of worried about social media tools that contribute to problems with concentration and potentially provoke ADHD. These days, it’s becoming even harder to focus on any task for more than 30 minutes. The constant notifications and the urge to check social media can significantly disrupt our concentration cycles, making it difficult to maintain productivity and deep focus. What are your hobbies and interests outside of tech? I'm a real travel nerd. In the past six months, I've been to 12 countries, including the Netherlands, Germany, Belgium, Singapore, Vietnam, and Indonesia. I love constantly learning and sharing knowledge with others. Recently, I became a judge for Technovation, where I provide feedback to a global community of girls in STEM. Let's talk about breaking the glass ceiling. What were the biggest challenges you faced as a woman in tech, and how did you deal with them? As a woman in tech, one of the biggest challenges I faced was staying consistent and true to my vision despite being doubted at every turn by others. At the beginning of my career, it was especially challenging to move forward without having strong results to build my confidence. What really helped me was building a healthy and supportive network. This network not only offered support and assistance but also provided a space where I could make mistakes without facing harsh criticism. For me, this supportive environment was the Accelerator Hub. There, making mistakes and trying new things was part of the process, and my mentor was a tremendous support in every way possible. So, I highly recommend ensuring you are surrounded by good people who support you. It makes a world of difference. Any questionable misogynistic story/situation you faced/handled, and you want to share with the HackerNoon Fam? Honestly, I was fortunate enough to work in companies where I received support from my colleagues and had ample opportunities for development. While I know that many women in tech face challenging situations, I am grateful that my experiences have been positive and supportive. What was the biggest setback/failure that you faced, and how did you manage it? I think one of my biggest career setbacks, which also turned into a huge push forward, was when I joined my first project as a business development manager. In the first few weeks, I brought in literally zero calls. At just 20 years old, I lacked the hardcore experience of international outbound B2B sales. Seeing my colleagues with a minimum of seven years of experience, hundreds of thousands of dollars in closed deals, and impressive backgrounds in world-class IT companies, I lost faith in my ability to keep up with them and perform at their level. However, the patience of my manager and my acceptance that I wasn't the smartest in the room, combined with a readiness to constantly learn and work hard, paid off. I regularly had calls with experienced managers, learned about their approaches, and asked for feedback as often as possible. Within a couple of months, I started showing results and eventually switched to sales hunting. What's your biggest achievement that you're really proud of? I'm proud of staying consistent with my career development. I started working hard when I was 17, and after starting university, it was extremely difficult to juggle studies, a regular job, and full-time work in a startup. For some reason, I thought it was a good idea to add regular sports into the mix, which resulted in getting only 4-5 hours of sleep for over six months. However, it was during this challenging time that my perseverance paid off, and I achieved significant milestones in my career. In your opinion, why do we see this huge gender gap in the tech industry, and how can we reduce it? I believe the huge gender gap in the tech industry is largely due to the lack of adequate infrastructure and opportunities for women to balance a well-established career and family life. For some reason, it often feels like a choice between the two, a choice that has predominantly been influenced by cultural norms. To reduce this gap, it’s crucial to provide work conditions that support a healthy work-life balance for women. Additionally, fostering a culture that emphasizes the importance of women's personal and professional development is key. By creating an environment that values and supports women's growth as individuals, we can help bridge the gender gap in the tech industry. Who is your tech idol? Why? It's hard to highlight just one specific person, but I won't be lying when I say that any woman who works hard and also manages to have a fulfilling life and a family outside of work is a hero to me. These women truly inspire me and earn my deepest respect. Do you have any advice for aspiring girls who want to join the field? My advice for aspiring girls who want to join the field is: don't be afraid to try! In the startup industry, flawless CVs and a list of completed courses aren't as important as your willingness to commit and work hard. If you want to try your hand at a project but aren't sure if you have enough experience, give it a shot anyway. Your dedication and eagerness to learn can make a big difference. The HackerNoon editorial team has launched this interview series with women in tech to celebrate their achievements and share their struggles. We need more women in technology, and by sharing stories, we can encourage many girls to follow their dreams. Share your story today! The HackerNoon editorial team has launched this interview series with women in tech to celebrate their achievements and share their struggles. We need more women in technology, and by sharing stories, we can encourage many girls to follow their dreams. Share your story today ! this interview series with women in tech Share your story today Tell us about yourself! Hey-hey! I’m Evgeniia Megrian, currently working as a Sales Hunter at Semrush, and previously at SLSBMB. My primary responsibility is searching for and hiring global sales talents for IT/Tech solutions. Over the past 3 years, I’ve successfully hired over 60 senior B2B sales professionals from tech giants like Google , Snap, and Salesforce . Google Salesforce My journey began with an Accelerator program at uni, where I managed an EdTech SaaS designed to simplify learning complex subjects. After a successful launch but an unsuccessful work-life balance, I shifted to business development as an employee. Eventually, I transitioned to a sales hunter role, applying the same strategic approach to recruiting top talent and forming successful partnerships. EdTech Why did you choose this field in the first place? Why did you choose this field in the first place? Early in my career, I joined an Accelerator program that introduced me to the startup world and the IT/Tech fields. This was my first real exposure to this dynamic industry. I loved the fast-paced environment, where we could test ideas quickly, be flexible, and create new processes instead of following strict rules. This approach matched really well with my values. I appreciated the opportunity to grow rapidly, try a dozen things, and see quick results. Through these experiences, I realized this was the field I wanted to develop in, and it also helped me advance quickly in my career. What tech are you most excited/passionate about right now and why? Currently, I’m super excited about fully implementing outbound strategies for hiring global talent. This approach allows me to reach not only candidates who apply but also those who are potentially relevant, greatly expanding our pool of prospects. It makes the hiring process more controlled and predictable. For instance, if we know exactly how many candidates we need to reach out to in order to hire a professional within a certain timeframe, the entire process becomes much more manageable. Additionally, I’m passionate about using automation solutions to create more effective outreach campaigns and build recruitment funnels. These technologies streamline the hiring process, making it more efficient and scalable. What tech are you most worried about right now and why? What tech are you most worried about right now and why? I'm kind of worried about social media tools that contribute to problems with concentration and potentially provoke ADHD. These days, it’s becoming even harder to focus on any task for more than 30 minutes. The constant notifications and the urge to check social media can significantly disrupt our concentration cycles, making it difficult to maintain productivity and deep focus. What are your hobbies and interests outside of tech? I'm a real travel nerd. In the past six months, I've been to 12 countries, including the Netherlands, Germany, Belgium, Singapore, Vietnam, and Indonesia. I love constantly learning and sharing knowledge with others. Recently, I became a judge for Technovation, where I provide feedback to a global community of girls in STEM. Let's talk about breaking the glass ceiling. What were the biggest challenges you faced as a woman in tech, and how did you deal with them? As a woman in tech, one of the biggest challenges I faced was staying consistent and true to my vision despite being doubted at every turn by others. At the beginning of my career, it was especially challenging to move forward without having strong results to build my confidence. What really helped me was building a healthy and supportive network. This network not only offered support and assistance but also provided a space where I could make mistakes without facing harsh criticism. For me, this supportive environment was the Accelerator Hub. There, making mistakes and trying new things was part of the process, and my mentor was a tremendous support in every way possible. So, I highly recommend ensuring you are surrounded by good people who support you. It makes a world of difference. Any questionable misogynistic story/situation you faced/handled, and you want to share with the HackerNoon Fam? Any questionable misogynistic story/situation you faced/handled, and you want to share with the HackerNoon Fam? Honestly, I was fortunate enough to work in companies where I received support from my colleagues and had ample opportunities for development. While I know that many women in tech face challenging situations, I am grateful that my experiences have been positive and supportive. What was the biggest setback/failure that you faced, and how did you manage it? What was the biggest setback/failure that you faced, and how did you manage it? I think one of my biggest career setbacks, which also turned into a huge push forward, was when I joined my first project as a business development manager. In the first few weeks, I brought in literally zero calls. At just 20 years old, I lacked the hardcore experience of international outbound B2B sales. Seeing my colleagues with a minimum of seven years of experience, hundreds of thousands of dollars in closed deals, and impressive backgrounds in world-class IT companies, I lost faith in my ability to keep up with them and perform at their level. However, the patience of my manager and my acceptance that I wasn't the smartest in the room, combined with a readiness to constantly learn and work hard, paid off. I regularly had calls with experienced managers, learned about their approaches, and asked for feedback as often as possible. Within a couple of months, I started showing results and eventually switched to sales hunting. What's your biggest achievement that you're really proud of? What's your biggest achievement that you're really proud of? I'm proud of staying consistent with my career development. I started working hard when I was 17, and after starting university, it was extremely difficult to juggle studies, a regular job, and full-time work in a startup. For some reason, I thought it was a good idea to add regular sports into the mix, which resulted in getting only 4-5 hours of sleep for over six months. However, it was during this challenging time that my perseverance paid off, and I achieved significant milestones in my career. In your opinion, why do we see this huge gender gap in the tech industry, and how can we reduce it? In your opinion, why do we see this huge gender gap in the tech industry, and how can we reduce it? I believe the huge gender gap in the tech industry is largely due to the lack of adequate infrastructure and opportunities for women to balance a well-established career and family life. For some reason, it often feels like a choice between the two, a choice that has predominantly been influenced by cultural norms. To reduce this gap, it’s crucial to provide work conditions that support a healthy work-life balance for women. Additionally, fostering a culture that emphasizes the importance of women's personal and professional development is key. By creating an environment that values and supports women's growth as individuals, we can help bridge the gender gap in the tech industry. Who is your tech idol? Why? It's hard to highlight just one specific person, but I won't be lying when I say that any woman who works hard and also manages to have a fulfilling life and a family outside of work is a hero to me. These women truly inspire me and earn my deepest respect. Do you have any advice for aspiring girls who want to join the field? My advice for aspiring girls who want to join the field is: don't be afraid to try! In the startup industry, flawless CVs and a list of completed courses aren't as important as your willingness to commit and work hard. If you want to try your hand at a project but aren't sure if you have enough experience, give it a shot anyway. Your dedication and eagerness to learn can make a big difference.

Stories, News & Mentions about ADALTA

Categories

Trending Topics

Stories, News & Mentions about ADALTA

Adalta’s Stories on HackerNoon

Adalta's News & Mentions

Categories

Trending Topics