Venture capital was pouring into new generative artificial intelligence technology before President Biden issued an order that could cool the rate of funds from flowing into the sector. Experts say that's not necessarily a bad thing for bank use of AI.

Should banks fear — or favor — a slowdown in AI investment?

Embedded financial services have proven ... entering the marketplace for the first time. “Software providers, in particular, are offering embedded finance services by leveraging external customer data, next-generation decisioning models and API-enabled ...

Embedded lending: helping consumers through a rough economy

If adopted correctly, financial services ... Officer at First Abu Dhabi Bank, continues: “Outside of risk management, quantum supercomputing will lead to a range of banking capabilities, such as analysing large areas of unstructured data to make ...

How quantum computing could transform the banking sector

Welcome to the digital revolution in finance! Today, we dive into a transformative phenomenon that has taken the financial world by storm: financial technology startups. With their game-changing innovations and disruptive approaches,

Investing in Tomorrow’s Finance: Why Financial Technology Startups are a Lucrative Opportunity

Yes, there are stringent regulations that the banks are worried about. So, the use cases where they've adopted AI have, until now, been in places that auditors have scored as low risk. Areas like customer service, where they've used things like text analytics and simplistic AI. But now generative AI is coming into the picture.

AI safety use cases - signing off on Financial Services’ ‘black box’

The family office technology space has grown and evolved to offer many efficient solutions that embrace the latest emerging technology to solve the most common challenges

The 2023 Family Office Software Roundup

First Savings Financial (FSFG) came out with quarterly earnings of $0.41 per share, beating the Zacks Consensus Estimate of $0.40 per share. This compares to earnings of $0.35 per share a year ago. These figures are adjusted for non-recurring items.

First Savings Financial (FSFG) Surpasses Q4 Earnings Estimates

JEFFERSONVILLE, Ind. (AP) — JEFFERSONVILLE, Ind. (AP) — First Savings Financial Group Inc. (FSFG) on Tuesday reported a loss of $747,000 in its fiscal fourth quarter. The Jeffersonville, Indiana-based company said it had a loss of 11 cents per share. Earnings, adjusted for non-recurring costs, came to 41 cents per share.

First Savings Financial: Fiscal Q4 Earnings Snapshot

Stocks in Toronto look set to snap an eight-session run lower as they track a bounce on Wall Street as investors look ahead to this week's Federal Reserve decision and the release of key economic data. The technology services and finance sectors were ...

Toronto Stocks Advance, Led by Tech and Financial Sectors

Gartner are right to point to bot-on-bot action for financial services moving forward. But even they may be underestimating the changes they will bring to money itself.

When The Bank’s Customers Are Replaced By Custobots

First Data

firstdata

When we think about data networks and systems that use mechanical mechanisms to send data across a distance, we think about this concept as a pretty modern one that only existed as a practical idea in the late 1900s. However, what if I told you that one of the earliest data networks to exist predated the modern internet as we know it? Heck, this ancient data network predates modern digital computers and analog computer systems. The modern electric telegraph is still younger than this particular early data network and most mechanical computers have not been around as long as this old-school data network. And when I say old-school, I mean truly old-school, as this data network was around before the Medieval Times and was developed by one of the oldest civilizations in written history. There is an ancient device known as the telegraph. This particular device was created to use water and devices in order to send information over long distances. The earliest version of this creation was produced back around 350 BCE by an early inventor and ancient Greek writer by the name of Aeneas Tacticus. It was produced by Tacticus as a device to be used for some long-form data communications during wartime & includes a very complicated system. The telegram worked by having identical containers on separate hills that are not connected to one another with each container being filled with water & having a vertical rod floating within both containers. The rods would have a bunch of predetermined codes put on them at various points along these poles. There would be a spigot at the bottom of these containers and water could be drained out of the device until a desired code was reached. Using a torch or signal, the person sending the code would be able to indicate how long it takes for the water to drain out to the desired code, allowing the receiver to know what signal/code is being communicated. hydraulic mechanical How these particular telegraph devices work is even explained by the Greek historian Polybius in his book : The Histories Then we are to fill the vessels with water and put on the corks with the rods in them and allow the water to flow through the two apertures. When this is done it is evident that, the conditions being precisely similar, in proportion as the water escapes the two corks will sink and the rods will disappear into the vessels. When by experiment it is seen that the rapidity of escape is in both cases the same, the vessels are to be conveyed to the places in which both parties are to look after the signals and deposited there. Now whenever any of the contingencies written on the rods occurs he tells us to raise a torch and to wait until the corresponding party raises another. When both the torches are clearly visible the signaler is to lower his torch and at once allow the water to escape through the aperture. Whenever, as the corks sink, the contingency you wish to communicate reaches the mouth of the vessel he tells the signaler to raise his torch and the receivers of the signal are to stop the aperture at once and to note which of the messages written on the rods is at the mouth of the vessel. This will be the message delivered, if the apparatus works at the same pace in both cases. What is so important about these primitive methods of sending data? Well, while it was limited in codes, it was a great early attempt at long-range telegraphic communication. It came as an early telegraph system that even comes before mechanical optical telegraphs and even inspired a British version that, while never made commercially available, was in 1838 by British civil engineer Francis Whishaw using pipes and more modern mechanical pressure systems. All of this came before and helped to inspire, the electric telegraphs that would -eventually - become the basis for the electric systems of communication we use today for the modern internet. So, to put a long story short, you better thank the Greeks and their system of sending codes long distances using containers and water for the modern communications that we get to enjoy today! developed

The Hydraulic Telegraph: The First Data Network All The Way From Ancient Greece!

Recently, while immersed in a game called "Factorio," I noticed striking parallels to the challenges encountered as the pioneer member of a data analytics team in a startup. In "Factorio," the game narrative revolves around an astronaut stranded on an alien planet with only raw materials at his disposal, tasked to build a spaceship for his escape. This scenario felt reminiscent of the real-world process of establishing an analytics team from scratch in a burgeoning startup. Let's visualize this scenario. You're just appointed the inaugural data and analytics team member in a fresh, promising startup. In your arsenal, you have an innovative product, abundant resources, and a horizon teeming with potential priorities. The questions then arise - where should you begin? What should be your initial thought process? How do you sequentially develop your strategy to construct an effective analytical team from scratch, which would eventually serve as a launchpad for the startup's trajectory? This article aims to delve into these intricate dynamics. : While this article draws parallels with the game Factorio, prior knowledge of the game isn't necessary to understand the insights shared; and for those familiar with Factorio, please pardon any oversimplification of the game mechanics for the sake of analogy. Disclaimer If you've ever wondered what it feels like to crash land on an alien planet with only your raw resources, congratulations – you're in a startup! Remember, a spaceship crash is just a pretty dramatic product pivot. Table of Contents Phase 1.  Manual/foundation Phase 2. Initiating automation Phase 3. Scaling Phase 4. Flight Conclusion As the inaugural member of the team in a startup, your journey is reminiscent of the first moments in "Factorio." Much like the astronaut, you're venturing into an alien planet, equipped with raw resources and the tools salvaged from your spaceship. These tools symbolize your wealth of knowledge, skills, and experience. Your mission? Transform these primary resources into a functioning rocket ship - a successful, scalable data analytics team. data analytics Phase 1. Manual/foundation In the initial phase of this mission, you're in exploration mode, much like the astronaut surveying his alien environment. You observe the data landscape, discover data sources (similar to mines of raw materials), and map out potential obstacles. This stage is all about understanding the topography of your new environment. It is a stage marked by high personal involvement, heightened curiosity, and hands-on work. You begin by manually gathering data from multiple sources, comparable to mining resources from different points in the game. You then create initial reports and analyses to better understand the product and business context. These reports serve as the first bricks in your rocket ship factory, offering valuable insights into your product's performance and customer behavior. Simultaneously, you forge relationships with various stakeholders in the startup - product managers, software engineers, marketers, and leadership. You communicate your findings, propose data-driven recommendations for product improvements, and even suggest simple A/B tests to validate your hypotheses. Your active involvement in these cross-functional dialogues enhances your understanding of the business and positions data as a crucial component in decision-making processes. Being the first data scientist in a startup is a bit like being the first one at the party - you're not quite sure where to put your coat, and you're a bit wary of double-dipping in the guacamole. However, this exploration and manual operation is challenging. This stage is labor-intensive and demands high levels of personal involvement. Analytics at this stage is manual, like a blacksmith forging tools by hand. Reporting involves connecting directly to data sources, stirring, and shaking data to understand what's happening. An ad-hoc analysis is frequent, which helps to understand unusual instances or random questions. Your nascent analytics infrastructure may also encounter hiccups that necessitate your manual intervention, sometimes leading to moments of realization about a forgotten piece that isn't working as it should. This stage can feel like a one-person show, where you're running around, juggling various tasks, and ensuring everything operate smoothly. While this phase may not be flexible or sustainable in the long run, it serves a vital purpose. It equips you with a deep understanding of the data landscape, the intricacies of the product, and the broader business context. You're intimately involved in shaping the foundation of your rocket ship factory, which imparts a rewarding sense of ownership and accomplishment. However, it's crucial to recognize when to transition from this manual, labor-intensive phase to a more automated, extendable model to avoid stagnation and burnout. This realization marks the end of stage one and the beginning of stage two: initiating automation. It's a vital turning point in your journey, keeping the evolution from a solitary explorer to a strategic architect aiming to build a robust, scalable data analytics team. Practical actions you can take at this stage Understand its sources, the relationships between different types of data, and the overall data architecture in your startup. Get to know your data. Start with developing essential reports that provide vital insights into your product or service. These can include user engagement metrics, performance metrics, sales, and revenue metrics, etc. Establish baseline reporting. These are custom, one-time pieces of research that are done on demand. Be prepared to perform ad-hoc analyses. Work closely with your colleagues, particularly those in product management and software engineering. Understand their needs, how they use data, and how you can help them make data-informed decisions. Collaborate with stakeholders. Phase 2. Initiating automation The initial stage's manual labor and high personal involvement serve a crucial purpose – they provide an intimate understanding of the data, business, and product context. This understanding is comparable to a detailed map of the alien planet in "Factorio," highlighting the rich mines of data, potential obstacles, and areas for exploration. However, a shift towards automation is necessary to ensure scalability and efficiency. As you transition into this phase, your role morphs from a hands-on miner to an architect. You begin to design processes that optimize the transformation of raw data into valuable insights, like setting up machines to automate the mining and refining of resources in the game. One of your primary tasks in this phase involves identifying bottlenecks in your processes – repeated frequently, essential for smooth operation, consuming a significant amount of your time, and, most importantly, can be automated. In the game of Factorio, it's akin to finding out which parts of the resource processing can be automated to free up time and boost efficiency. These tasks could range from ad-hoc analyses required for primary product metrics reporting, studies essential for A/B tests (such as power analysis for the new test and post-test analysis of the A/B test metrics), or even routine data gathering, joining, and transformation to answer recurring business questions. When setting up your initial data infrastructure, you'll often encounter a piece of code that no one dares to touch. It's like the support beam in a crumbling building - we don't know why it's holding everything up, but we're not about to pull it out to find out. As you identify these tasks, you adopt some of three primary strategies to streamline your workflow: : In "Factorio," a factory is an assembly line of automated machinery that takes in raw materials and churns out finished goods. In the world of data analytics, a 'factory' is a suite of tools and systems designed to automate your workflow. These may include workflow schedulers like Airflow, which automate regular tasks, dashboard systems such as Redash or Grafana, which automate reporting and foster a self-service environment for stakeholders; and ETL/ELT pipelines which automate data transportation and transformation. Building a factory : This is similar to inviting friends for a cooperative game in Factorio. Expanding your team involves hiring more data specialists. They bring diverse skill sets and perspectives, which accelerates progress and helps build robust analytics capabilities. Expanding your team : In Factorio, a technology tree represents all the possible upgrades and technological advancements. Similarly, in a startup, it refers to planning and implementing the necessary technology stack that supports your data needs, ensuring the scalable, long-term growth of your analytics operations. Developing a technology tree However, it's essential to maintain a balance during this phase. While it may be tempting to automate every process and hire extensively, ensuring these decisions align with the startup's financial resources and strategic goals is critical. Additionally, remember to design flexible systems that can adapt to evolving business needs – avoid creating a complicated, high-maintenance infrastructure that could become a burden in the future. The transition from stage one to stage two reminds me of evolving from an explorer to an architect in Factorio. It marks a significant shift in how you interact with data and sets the stage for your analytics team's scalable, efficient growth. The journey is far from over, and there are still stages to navigate, but the groundwork you've laid here will be instrumental in the coming stages. Practical actions you can take at this stage: This could include data cleaning tools, reporting software, dashboarding tools, etc. Look for tools that can help you automate your routine tasks. It can save you considerable time and increase efficiency. Identify tasks you perform frequently and look for ways to automate them. It involves setting up data pipelines, implementing a data warehouse, and setting up ETL processes. Start developing your data infrastructure. Look for individuals with skills that complement your own and can help you scale your operations. As you move to this stage, you may need to expand your team. Phase 3. Scaling In the previous stages, you've laid the foundation and initiated automation for your analytics operations, much like building a rudimentary factory on an alien planet in "Factorio." Now comes the time to scale, expand your operations, and prepare for growing demands. This phase is similar to increasing your factory, optimizing production lines, and preparing for larger, more complex challenges in the game. At this juncture, your analytics system is mainly automated, doing much of the routine work without your constant involvement. The system has morphed from a basic rocket ship factory into a more sophisticated operation. Yet, this is not the moment to rest but to gear up for more significant challenges. It's similar to the game scenario where resources become scarcer and demand multiplies - it calls for a strategic expansion. The scaling phase has three crucial components: : In "Factorio," as you exhaust local resources, you must build train networks to import materials from distant mines. Similarly, as your internal data resources need to be improved to drive deeper insights in a startup scenario, you might need to integrate third-party data or expertise. This could mean adopting external data APIs, forging partnerships for data sharing, or hiring external consultants or agencies with niche skills. External resource integration : Much like setting up dashboards to monitor your factory's efficiency in Factorio, it is crucial to establish robust performance tracking in a data analytics setup. Regularly monitoring your analytics system's performance, peak capacity, and identification of bottlenecks ensure its smooth operation. It also provides insights for future improvements and helps in proactive troubleshooting. Monitoring : In "Factorio," you often have to restructure your factory to accommodate new technologies or to make the production line more efficient. Similarly, in a data environment, refactoring or re-engineering your data pipelines, models, or reports might be necessary to improve efficiency, reduce computational costs, or incorporate changes in the business context. Refactoring Throughout this stage, the critical emphasis is on efficiency and optimization. While aiming for massive growth is tempting, the scale phase is more about 'smart' growth. It's about ensuring that your operations are extendable, maintainable, and flexible enough to adapt to the rapid changes characteristic of startup environments. Scaling up your processes is a bit like building your IKEA furniture - you start off with manual labor, realize you need some automation (or at least a power drill), and eventually, you’re standing back admiring your work, just to realize you’ve got some leftover parts. Did we really need that data pipeline, or was it just an optional extra? Scaling also includes planning for future growth. While you want to ensure that your current setup is efficient, you also want to anticipate the needs of your growing organization. This might mean investing in scalable technologies, planning for increased data storage and processing capabilities, or even initiating discussions about building a data warehouse. Much like the game "Factorio," the challenges are becoming more complex and demanding, but the opportunities for creativity and impact are also growing. As you navigate this stage, your role evolves from an architect to a city planner, meticulously planning and overseeing the growth and development of your analytics operations. And as you do so, you lay the groundwork for the final stage - the flight. Practical actions you can take at this stage Implement tools and processes to monitor the performance of your data infrastructure. Identify bottlenecks and areas for improvement. Continually look for ways to optimize your processes and data infrastructure. As your operations grow, you may find that specific processes or systems no longer meet your needs. Be prepared to refactor or redesign these as necessary. You may need to start integrating external data sources or systems. Establish new data pipelines or expand your data infrastructure. Phase 4. Flight Much like in "Factorio," where the ultimate goal is to build a rocket and escape the alien planet, in the context of a startup's analytics team, the 'flight' phase represents reaching a stage where your analytics operations are fully-functional, robust, flexible, and are driving strategic decisions and value creation. At this juncture, you're not just managing a data analytics operation; you're overseeing a well-oiled machine contributing significantly to the startup's trajectory. You have built a system that functions almost independently, only requiring your attention for high-level strategic decisions and occasional troubleshooting. However, just as in the game, reaching the final stage doesn't mean you're free from challenges. Even at this level, there are hurdles to overcome: : In "Factorio," the aliens evolve and become more of a threat as you progress. Similarly, competitors will continue to grow and innovate in a startup environment. You must ensure that your analytical capabilities continue to develop, staying ahead of the curve and providing the strategic edge your startup needs to thrive. Evolving competition : Time is a critical resource, especially for startups. Despite having a successful operation, you may need more help. It's essential to ensure that your analytics operation runs efficiently and that resources are allocated strategically to support continual growth and innovation. Resource constraints : Over time, some of your data processes and models may lose effectiveness or become outdated. Just as you would upgrade your factories and technologies in "Factorio," it's essential to regularly revisit and improve your analytics operations. Keeping your analytics tools and techniques up-to-date ensures you continue generating valuable insights. Decreasing marginal utility Therefore, the 'flight' phase is not about coasting but continual improvement, adaptation, and innovation. It's about maintaining the momentum built up and pushing for greater heights. Practical actions you can take at this stage Continually look for ways to improve your operations, whether through new tools, processes, or data analysis approaches. Don't rest on your laurels. This might involve learning new tools or techniques, attending conferences, or reading relevant publications. Keep up with developments in the field of data science. Use your data insights to inform these decisions and drive growth. You should contribute to your startup's strategic decisions at this stage. . Ensure that you are allocating your time, your team's time, and your budget to maximize return on investment. Be mindful of your resources Conclusion Establishing and growing an analytics team in a 0 → 1 startup environment can be an exciting, challenging, and rewarding journey - much like progressing through the stages of the game. Starting from basic manual work, you gradually build your operations, initiate automation, scale your capabilities, and finally reach the flight stage, where your analytics operations become vital to your startup's journey. However, each step requires different skills, strategies, and perspectives. Like in the game, you need to continually adapt, innovate, and strategize to ensure the success of your analytics operations and, consequently, your startup. By understanding and implementing this four-stage approach, you can survive and thrive in a 0 → 1 analytics team.

From Crashing to Lift-Off: How to Thrive as the First Data Scientist in a Startup

ক্র্যাশিং থেকে লিফ্ট-অফ পর্যন্ত: একটি স্টার্টআপে প্রথম ডেটা সায়েন্টিস্ট হিসাবে কীভাবে উন্নতি করবেন

Vom Absturz zum Start: Wie Sie als erster Datenwissenschaftler in einem Startup erfolgreich sind

충돌에서 이륙까지: 스타트업에서 최초의 데이터 과학자로 성공하는 방법

Çökmeden Kalkışa: Bir Startup'taki İlk Veri Bilimcisi Olarak Nasıl Başarılı Olunur?

You’re a founder. You know you need to have a “data play” (or worse, “AI play”). Investors and clients are asking about machine (or worse, deep learning). The question is no longer why, but when. startup learning So, you hire your first data scientist. A fresh graduate (those with industry experience are too expensive and hard to recruit). They have a PhD (so you can nonchalantly drop ‘I have my best PhDs looking into this’ in the conversation with potential clients and VCs.) They‘re excited about doing machine learning — and so are you. This is going to rock! Your data scientist starts, ready to take your data and build models from it, just like you agreed. Back in academia, they were working with datasets that several generations of grad students have been using. They’re not naive, though — they know this time around they’ll have to create their own. No problem! Your startup has a year worth of logs in S3, ready to be mined and used as training data. How not to think about the data science Venn diagram There’s no infrastructure for data analysis, so everything has to be done from scratch. Your data scientist is trying to install various tools incompatible with your stack or architecture and, for now, is doing one-off log parsing to answer every question. The engineering team doesn’t feel it’s safe to give them prod access so they provide an offline copy of your database. you say to the data scientist. There! You have access to all the data you need! Except… data in that offline database is, by design, not structured in a way that’s easy to analyze or combine with the data in the logs. Values don’t make sense or are missing for weeks at a time, the only events instrumented are those relevant to the ops team, and even the simplest query takes forever to run. Everybody is frustrated . The data scientist wanted to work on machine learning — those elegant algorithms they spent their career thinking and publishing about. Yes, they expected to put some effort into gathering and cleaning data — but they didn’t expect it to be so complex and messy, with so much missing or hard to access. They didn’t expect to spend so much time in meetings or Slack with questions about how the data was gathered and what, exactly, is in that json blob. They didn’t expect the rest of the company to care so little about how each redesign messes up month to month comparisons. They didn’t realize nobody logged all the options a user saw before they clicked, nor the version of the UI at the time, so that one year of training data is not exactly usable. You hired a data scientist to work on machine learning. They’ve been warned about spending 80% of their time cleaning data. That sounds like a distant dream now. Instead, they spend 80% of their time begging for data to be created, accessed, moved, or explained. They’re spending the other 20% lobbying for data science-friendly tools, security policies, and infrastructure, or looking for a new job. The engineering team is frustrated, too. They have to take time away from their own work to do a ton of thankless grunt work for the scientist who sits around, does nothing useful and still manages to constantly complain about the data not being good enough. You, the founder, are even more frustrated. It’s been 2 months, and the data scientist didn’t even produce a decent dashboard, much less the magic machine learning everybody keeps raving about. Plus, they seem like a bad culture fit, slowing everybody else down for no reason. Deep down, you start suspecting machine learning is snake oil and . AI is just hype You keep this to yourself, though, and tell your investors and clients about your AI play and your crazy smart PhD data scientist who’s super excited to work on machine learning any day now. If you’re wondering how to avoid this painful but too frequent scenario, here’s the Quora answer that inspired this post : “ What are the challenges of building a data team at a startup ?”. Think other people would benefit from this article? Click the heart to help them find it ❤

How not to hire your first data scientist

With data breaches and cyberattacks on the rise, you can never be too careful. Backup testing prevents loss and minimizes cyberthreats’ impact, making it fundamental to cybersecurity. However, like many in your position, you may not consider it a priority. Are you opening yourself up to unnecessary risks by not making it part of your security strategy? Why Make Backup Testing Part of Your Security Strategy? Backups are copies of files, databases, applications or systems. Testing involves checking them to verify they are accurate, intact, and up to date. Its purpose is to evaluate the effectiveness of your strategy and schedule, letting you make actionable changes. It also helps you recognize and address issues proactively — meaning before a threat actor makes recovery impossible. Data loss is more common than you might think. In 2023, 46% of organizations worldwide lost some amount of sensitive information. Since these incidents can be caused by anything from human error to cyberattacks, preventing them can be challenging. This is why backups are so helpful — redundancy is crucial to recovery. Intact backups make recovery possible. Without them, the process would be expensive. Organizations spent about $4.35 million per breach on average in 2022. In extreme cases, recovery may be impossible — you could lose weeks, months, or years of valuable files. Testing these copies is critical to your security strategy. If you’re like many cybersecurity professionals, you don’t consider redundancy a priority. Even though 72% of organizations had to use a backup to recover at least once in 2021, 41% admitted they rarely — if ever — back up their data. Crucially, even if you create and update copies, you may still be at risk of loss. Unfortunately, backups aren’t indestructible. Human error, bugs, hardware damage, incompatibilities, and mismatched encryption keys can cause data loss, rendering them useless. Alternatively, a tech-savvy attacker could use ransomware to encrypt all your files simultaneously, preventing you from using your copies to recover. The Fundamentals of a Strategic Backup Testing Policy You must first understand the fundamentals of backup testing to incorporate a high-quality policy into your security strategy. What to Test Several test scenarios exist, with different kinds for each data loss cause. Corruption, accidental deletion, and cyberattacks are some of the most common, but it may pay to prepare for unlikely incidents. Your backup’s storage medium, location, and size will also affect your approach. Be mindful that you’re trying to improve your security strategy. When to Test Prioritize periodic testing instead of kicking the can down the road for the sake of convenience. Work files may need to be backed up weekly, depending on how often information changes. Also, frequency varies depending on the type of data you’re backing up. While archives can go months without an update, you should back applications up more often. Where to Test If you can't prioritize everything, which backups should you test? Consider creating a risk-based schedule. You typically want to test on-site data first since it enables rapid recovery. Generally, off-site copies should only take priority if a catastrophic event threatens the integrity of all local files. How to Incorporate Backup Testing Into Your Strategy Testing involves selecting a scenario and simulating the recovery process. Once you create an isolated test environment to prevent unintended changes, you reproduce real-world events like accidental deletion or cyberattacks. Then, you go through the step-by-step process to restore your systems, applications, files, or databases via your backup. Upon completion, you should have a fully restored environment. If you notice discrepancies when comparing the original to the restored version, your backups may not be fully intact. If that happens, search for anomalies like corruption, duplicate entries, or mistyped fields. Even if you don’t find anything wrong, testing helps you identify potential process improvements. Documentation is key. You should record how long restoration took, what issues emerged, which steps you took, and whether you succeeded on the first try. This way, you can accurately estimate how long recovery will take. It also lets you know what to expect in the future. If someone else is in charge of the next round of testing, they’ll appreciate those details. Incorporating testing into your strategy means making room in your busy schedule. Consider automation to free up time. Tools like robot process automation and artificial intelligence can automate backup creation, restoration simulation, and discrepancy detection. Since 72% of companies have adopted AI as of 2024, your employer may already have such tools. If you automate testing, consider whether you’ll leverage periodic or trigger-based tests — or some combination of both. While evaluating various systems, applications, and files weekly or monthly is fine, it may lead to minor data loss. Setting your tools to activate when you find evidence of a breach or detect unusual network activity mitigates this issue. Considerations for Uncommon Backup Testing Scenarios Sometimes, assigning roles, creating schedules, and refining strategies isn’t enough. You should consider edge cases — tests for issues that only occur under unusual circumstances — for a comprehensive backup testing policy. Imagine you only prioritize accidental deletion and breach resolution. What happens when an insider threat intentionally wipes a system? Preparing your organization for potential data loss incidents by incorporating unlikely test scenarios into your security strategy improves its security posture. Since cyberthreats and the digital landscape are ever-changing, being proactive can help. Setting yourself up for success can save time and preserve your employer’s reputation. With data breaches and cyberattacks on the rise, you can never be too careful. Backup testing prevents loss and minimizes cyberthreats’ impact, making it fundamental to cybersecurity. However, like many in your position, you may not consider it a priority. Are you opening yourself up to unnecessary risks by not making it part of your security strategy? With data breaches and cyberattacks on the rise, you can never be too careful. Backup testing prevents loss and minimizes cyberthreats’ impact, making it fundamental to cybersecurity. However, like many in your position, you may not consider it a priority. Are you opening yourself up to unnecessary risks by not making it part of your security strategy? Why Make Backup Testing Part of Your Security Strategy? Why Make Backup Testing Part of Your Security Strategy? Backups are copies of files, databases, applications or systems. Testing involves checking them to verify they are accurate, intact, and up to date. Its purpose is to evaluate the effectiveness of your strategy and schedule, letting you make actionable changes. It also helps you recognize and address issues proactively — meaning before a threat actor makes recovery impossible. Data loss is more common than you might think. In 2023, 46% of organizations worldwide lost some amount of sensitive information. Since these incidents can be caused by anything from human error to cyberattacks, preventing them can be challenging. This is why backups are so helpful — redundancy is crucial to recovery. 46% of organizations worldwide 46% of organizations worldwide Intact backups make recovery possible. Without them, the process would be expensive. Organizations spent about $4.35 million per breach on average in 2022. In extreme cases, recovery may be impossible — you could lose weeks, months, or years of valuable files. Testing these copies is critical to your security strategy. spent about $4.35 million spent about $4.35 million If you’re like many cybersecurity professionals, you don’t consider redundancy a priority. Even though 72% of organizations had to use a backup to recover at least once in 2021, 41% admitted they rarely — if ever — back up their data. Crucially, even if you create and update copies, you may still be at risk of loss. 72% of organizations had to use 72% of organizations had to use Unfortunately, backups aren’t indestructible . Human error, bugs, hardware damage, incompatibilities, and mismatched encryption keys can cause data loss, rendering them useless. Alternatively, a tech-savvy attacker could use ransomware to encrypt all your files simultaneously, preventing you from using your copies to recover. backups aren’t indestructible backups aren’t indestructible The Fundamentals of a Strategic Backup Testing Policy You must first understand the fundamentals of backup testing to incorporate a high-quality policy into your security strategy. What to Test Several test scenarios exist, with different kinds for each data loss cause. Corruption, accidental deletion, and cyberattacks are some of the most common, but it may pay to prepare for unlikely incidents. Your backup’s storage medium, location, and size will also affect your approach. Be mindful that you’re trying to improve your security strategy. When to Test Prioritize periodic testing instead of kicking the can down the road for the sake of convenience. Work files may need to be backed up weekly , depending on how often information changes. Also, frequency varies depending on the type of data you’re backing up. While archives can go months without an update, you should back applications up more often. may need to be backed up weekly may need to be backed up weekly Where to Test If you can't prioritize everything, which backups should you test? Consider creating a risk-based schedule. You typically want to test on-site data first since it enables rapid recovery. Generally, off-site copies should only take priority if a catastrophic event threatens the integrity of all local files. How to Incorporate Backup Testing Into Your Strategy Testing involves selecting a scenario and simulating the recovery process. Once you create an isolated test environment to prevent unintended changes, you reproduce real-world events like accidental deletion or cyberattacks. Then, you go through the step-by-step process to restore your systems, applications, files, or databases via your backup. Upon completion, you should have a fully restored environment. If you notice discrepancies when comparing the original to the restored version, your backups may not be fully intact. If that happens, search for anomalies like corruption, duplicate entries, or mistyped fields. Even if you don’t find anything wrong, testing helps you identify potential process improvements. Documentation is key. You should record how long restoration took, what issues emerged, which steps you took, and whether you succeeded on the first try. This way, you can accurately estimate how long recovery will take. It also lets you know what to expect in the future. If someone else is in charge of the next round of testing, they’ll appreciate those details. Incorporating testing into your strategy means making room in your busy schedule. Consider automation to free up time. Tools like robot process automation and artificial intelligence can automate backup creation, restoration simulation, and discrepancy detection. Since 72% of companies have adopted AI as of 2024, your employer may already have such tools. 72% of companies have adopted AI 72% of companies have adopted AI If you automate testing, consider whether you’ll leverage periodic or trigger-based tests — or some combination of both. While evaluating various systems, applications, and files weekly or monthly is fine, it may lead to minor data loss. Setting your tools to activate when you find evidence of a breach or detect unusual network activity mitigates this issue. Considerations for Uncommon Backup Testing Scenarios Sometimes, assigning roles, creating schedules, and refining strategies isn’t enough. You should consider edge cases — tests for issues that only occur under unusual circumstances — for a comprehensive backup testing policy. Imagine you only prioritize accidental deletion and breach resolution. What happens when an insider threat intentionally wipes a system? Preparing your organization for potential data loss incidents by incorporating unlikely test scenarios into your security strategy improves its security posture. Since cyberthreats and the digital landscape are ever-changing, being proactive can help. Setting yourself up for success can save time and preserve your employer’s reputation.

Is Backup Testing Part of Your Security Strategy?

Je testování zálohování součástí vaší bezpečnostní strategie?

Er sikkerhedskopieringstest en del af din sikkerhedsstrategi?

¿Las pruebas de respaldo son parte de su estrategia de seguridad?

Babeskopia-probak zure segurtasun-estrategiaren parte al dira?

バックアップ テストはセキュリティ戦略の一部ですか?

არის სარეზერვო ტესტირება თქვენი უსაფრთხოების სტრატეგიის ნაწილი?

ການທົດສອບການສໍາຮອງຂໍ້ມູນເປັນສ່ວນຫນຶ່ງຂອງຍຸດທະສາດຄວາມປອດໄພຂອງທ່ານບໍ?

Adakah Ujian Sandaran Sebahagian daripada Strategi Keselamatan Anda?

Maakt back-uptesten deel uit van uw beveiligingsstrategie?

උපස්ථ පරීක්ෂාව ඔබගේ ආරක්ෂක උපාය මාර්ගයේ කොටසක්ද?

A është testimi rezervë pjesë e strategjisë suaj të sigurisë?

Чи є резервне тестування частиною вашої стратегії безпеки?

One data scientist recently argued with me that we should stop using the term “hallucination” and just call these outputs what they really are: errors. The premise is simple—calling them errors would set more reasonable expectations for AI’s behavior. While this argument might make sense for someone coming from a statistics or programming background, it fundamentally misunderstands how AI works and why the term “hallucination” is actually more appropriate, albeit imperfect. The Misunderstanding of AI’s Nature From a statistical perspective, an error is a deviation from a known value or standard. It’s easy to quantify, easy to detect, and there’s a clear sense of right and wrong. If a model predicts a 20% probability when the true probability is 30%, we call it an error. We can measure this deviation, adjust the model, and move forward. For a data scientist accustomed to working within this framework, it’s natural to see LLM outputs that are factually incorrect as just another type of error. If an AI model says, “A circle has three sides,” then it’s clearly made a mistake—like a regression model producing a bizarre outlier. The problem with this comparison is that it applies a narrow definition of error to a system that generates language, not discrete values. AI models like GPT-4 do not “make errors” in the traditional sense because they lack a clear objective standard like those found in statistics. Why We Call It “Hallucination” AI language models generate sequences of tokens—words and phrases—based on probabilistic patterns. These patterns are learned from vast amounts of text data, and the models produce the most probable sequence of words given the context. Because of this structure, a language model might generate a phrase that is syntactically correct but semantically flawed. Imagine these two statements: “A circle has three sides.” “The professor said a circle has three sides.” If you evaluate these on a token-by-token basis, both might be highly probable outputs based on their preceding context. The first statement is objectively false, while the second statement could be contextually true if a professor actually made that erroneous claim. The model doesn’t have an internal understanding of geometry or truth—it’s simply generating words that fit together well. That’s why when the output is nonsense, we call it a “hallucination” instead of an error. The model didn’t make a mistake according to its own mechanics; it just produced an improbable yet plausible-sounding string of text. The Inadequacy of Calling It “Error” Labeling such outputs as errors creates the false impression that these models should know better. But knowing requires understanding, and that’s not how AI models are built. They don’t “know” in the way humans do, nor do they validate statements against a factual baseline. There is no inherent “truth” within the architecture of an LLM. The model operates based on patterns and associations, not logical consistency or factual accuracy. When I hear the argument that AI hallucinations are “logical errors” or “input errors,” it reminds me of the Garbage In, Garbage Out (GIGO) adage in traditional programming. If your inputs are flawed or your logical framework is off, the system produces wrong outputs. But that analogy only goes so far with AI because language and context are far more nuanced than a spreadsheet of values or a database query. An LLM doesn’t make logical errors in the programming sense. Instead, it lacks an internal verification process to ensure that what it’s saying is true. Imagine trying to apply the same logic to artistic styles or creativity. If an AI-generated image combines elements of a cat and a cloud, calling it an “error” is inappropriate. The model didn’t err—it produced a plausible but nonsensical creation based on its input data and generative process. Hallucinations and Human Perception This distinction is crucial because calling these outputs “errors” could mislead people into thinking that LLMs are making simple mistakes that can be fixed with better logic or more data. But hallucinations in LLMs are not bugs that can be patched. They are a byproduct of how these models work—of generating text without a built-in sense of truth. The onus is still on the human user to guide the AI’s output by crafting better prompts, setting clearer parameters, and using post-processing tools to verify facts. This is why the term “hallucination,” while imperfect, is closer to the mark. It conveys that the model is not just producing an error but fabricating new content—fabrications that are a natural consequence of its design, not the result of poor logic or faulty inputs. It serves as a reminder that these models are generative, not analytical. They don’t have a grasp of reality and can produce completely fabricated content even when all the inputs seem perfect. Setting Expectations for AI’s Outputs Ultimately, the data scientist’s argument comes from a place of wanting to simplify AI’s behaviors for the end-user. If we call these outputs “errors,” maybe people will understand AI’s limitations better. But in reality, this simplification does a disservice to the complexity of these models. We need to help people understand that AI models do not think, reason, or understand—they predict and generate. Their hallucinations are not “errors” in a mechanical or statistical sense but are intrinsic artifacts of how they operate. Until we develop AI models with built-in mechanisms for validating truth, we will continue to see outputs that are not just factually wrong but sometimes fantastical. And as we refine our language and expectations around AI, we need to keep terms like “hallucination” in place—if only to remind ourselves that these models are not yet capable of true understanding. So, while I appreciate the sentiment behind labeling these outputs as errors, it inadvertently strips away the nuances that make AI hallucinations fundamentally different from the statistical errors many data scientists are familiar with. As we continue to refine our models and our language, it’s worth keeping these distinctions in mind to set clearer expectations for what AI can and cannot do. About Me: 25+ year IT veteran combining data, AI, risk management, strategy, and education. 4x hackathon winner and social impact from data advocate. Currently working to jumpstart the AI workforce in the Philippines. Learn more about me here: https://docligot.com One data scientist recently argued with me that we should stop using the term “hallucination” and just call these outputs what they really are: errors. The premise is simple—calling them errors would set more reasonable expectations for AI’s behavior. While this argument might make sense for someone coming from a statistics or programming background, it fundamentally misunderstands how AI works and why the term “hallucination” is actually more appropriate, albeit imperfect. The Misunderstanding of AI’s Nature From a statistical perspective, an error is a deviation from a known value or standard. It’s easy to quantify, easy to detect, and there’s a clear sense of right and wrong. If a model predicts a 20% probability when the true probability is 30%, we call it an error. We can measure this deviation, adjust the model, and move forward. For a data scientist accustomed to working within this framework, it’s natural to see LLM outputs that are factually incorrect as just another type of error. If an AI model says, “A circle has three sides,” then it’s clearly made a mistake—like a regression model producing a bizarre outlier. The problem with this comparison is that it applies a narrow definition of error to a system that generates language, not discrete values. AI models like GPT-4 do not “make errors” in the traditional sense because they lack a clear objective standard like those found in statistics. Why We Call It “Hallucination” AI language models generate sequences of tokens—words and phrases—based on probabilistic patterns. These patterns are learned from vast amounts of text data, and the models produce the most probable sequence of words given the context. Because of this structure, a language model might generate a phrase that is syntactically correct but semantically flawed. Imagine these two statements: “A circle has three sides.” “The professor said a circle has three sides.” If you evaluate these on a token-by-token basis, both might be highly probable outputs based on their preceding context. The first statement is objectively false, while the second statement could be contextually true if a professor actually made that erroneous claim. The model doesn’t have an internal understanding of geometry or truth—it’s simply generating words that fit together well. That’s why when the output is nonsense, we call it a “hallucination” instead of an error. The model didn’t make a mistake according to its own mechanics; it just produced an improbable yet plausible-sounding string of text. The Inadequacy of Calling It “Error” Labeling such outputs as errors creates the false impression that these models should know better. But knowing requires understanding, and that’s not how AI models are built. They don’t “know” in the way humans do, nor do they validate statements against a factual baseline. There is no inherent “truth” within the architecture of an LLM. The model operates based on patterns and associations, not logical consistency or factual accuracy. When I hear the argument that AI hallucinations are “logical errors” or “input errors,” it reminds me of the Garbage In, Garbage Out (GIGO) adage in traditional programming. If your inputs are flawed or your logical framework is off, the system produces wrong outputs. But that analogy only goes so far with AI because language and context are far more nuanced than a spreadsheet of values or a database query. An LLM doesn’t make logical errors in the programming sense. Instead, it lacks an internal verification process to ensure that what it’s saying is true. Imagine trying to apply the same logic to artistic styles or creativity. If an AI-generated image combines elements of a cat and a cloud, calling it an “error” is inappropriate. The model didn’t err—it produced a plausible but nonsensical creation based on its input data and generative process. Hallucinations and Human Perception This distinction is crucial because calling these outputs “errors” could mislead people into thinking that LLMs are making simple mistakes that can be fixed with better logic or more data. But hallucinations in LLMs are not bugs that can be patched. They are a byproduct of how these models work—of generating text without a built-in sense of truth. The onus is still on the human user to guide the AI’s output by crafting better prompts, setting clearer parameters, and using post-processing tools to verify facts. This is why the term “hallucination,” while imperfect, is closer to the mark. It conveys that the model is not just producing an error but fabricating new content—fabrications that are a natural consequence of its design, not the result of poor logic or faulty inputs. It serves as a reminder that these models are generative, not analytical. They don’t have a grasp of reality and can produce completely fabricated content even when all the inputs seem perfect. Setting Expectations for AI’s Outputs Ultimately, the data scientist’s argument comes from a place of wanting to simplify AI’s behaviors for the end-user. If we call these outputs “errors,” maybe people will understand AI’s limitations better. But in reality, this simplification does a disservice to the complexity of these models. We need to help people understand that AI models do not think, reason, or understand—they predict and generate. Their hallucinations are not “errors” in a mechanical or statistical sense but are intrinsic artifacts of how they operate. Until we develop AI models with built-in mechanisms for validating truth, we will continue to see outputs that are not just factually wrong but sometimes fantastical. And as we refine our language and expectations around AI, we need to keep terms like “hallucination” in place—if only to remind ourselves that these models are not yet capable of true understanding. So, while I appreciate the sentiment behind labeling these outputs as errors, it inadvertently strips away the nuances that make AI hallucinations fundamentally different from the statistical errors many data scientists are familiar with. As we continue to refine our models and our language, it’s worth keeping these distinctions in mind to set clearer expectations for what AI can and cannot do. About Me: 25+ year IT veteran combining data, AI, risk management, strategy, and education. 4x hackathon winner and social impact from data advocate. Currently working to jumpstart the AI workforce in the Philippines. Learn more about me here: https://docligot.com About Me: 25+ year IT veteran combining data, AI, risk management, strategy, and education. 4x hackathon winner and social impact from data advocate. Currently working to jumpstart the AI workforce in the Philippines. Learn more about me here: https://docligot.com

Why Your Data Scientists Will Struggle With AI Hallucinations

Achieving Optimal Service Reliability: Insights Into Service Level Objectives (SLOs)

Every time I share the latest PR insights, I end up hearing the same feedback over and over again: “Well, now it’s clear. All you need is money.” I don’t know how those keen minds came to such a twisted conclusion, but to me, it’s obvious that there’s always an option to start small. In this post, I’ll share some takeaways for how to gain media coverage and leverage your PR endeavors with no news peg and zero budget. Option 1: Media Coverage Through Op-Ed, Guest Posting, or Columns Media are very fond of experts who tackle industry issues through the lens of their experience and provide editors with fresh and insightful material that would catch a reader’s eye and drive traffic to their website. So, gaining media coverage through op-eds and guest posts is the easiest way to start. Why? Because through the personal expertise of your C-level representative, you will manage to tackle industry issues that somehow relate to your product. How? Here’s a step-by-step guide: Step 1: Decide on a spokesperson and map the topics for them. In one circle, the list of the topics that would resonate with the industry that you are working in. In the second circle, note down the vectors in which your speaker has expertise. Then check where those circles overlap, and choose 3-4 major areas to cover. Those ideas should both demonstrate your speaker’s valuable experience and tackle the industry issues for which you already have a solution at hand. Generate a bunch of topics, headlines, and detailed outlines for each. Step 2: Create your media list with relevant contacts. Check which media are covering this or that topic, make a list of such media, find opinions editor’s email, and send them a proposal for approval(consisting of the headline and a detailed outline). Some media like Daily Hodl or HackerNoon have a submission form, but in this case, you have to submit the whole article for review, not just the outline. I would also recommend doing small research to see whether this particular angle hasn’t been covered yet. Step 3: Wait for a response. Great things take time. Don’t share the same content with all the outlets simultaneously. You may end up gaining coverage across multiple outlets and thereby ruin your relations with journalists. Instead, wait for approval from the editorial team (they indicate in the requirements how much time they usually need to review your submission). If there’s no answer within a specified timeframe, then share your idea with others. Pros of leveraging your media relations through guest posts and op-eds: Your brand name and maybe the link to the website will appear in the author’s bio. You will be able to insert some information about your product. If it’s relevant and doesn’t look like a promotion, you can use your product and company’s data to back your stance. You establish thought leadership and build a strong personal brand for your speaker. If you are a no-name startup with a no-name CEO, building a personal brand is crucial to make people trust you and in you. No wonder smaller companies are ready to pay a fortune to hunt C-level executives with sound names and spotless reputations.
Because by accepting a job offer, this person confirms their trust in the company, making it more credible in the eyes of the industry by default. Option 2: Media Coverage Through Commentaries Nobody accepts commentaries from people with no prior trace left in the media. So, you better start with the opinion part and gradually add commenting atop this. Crypto is a very wavering environment with a lot of stuff going on. Keep track of everything happening in the industry. Start your morning by checking Feedly and Google News to never miss a beat. There are several ways how you can ensure media coverage through expert commentaries: As soon as you have secured 5-6 pieces of coverage for your speaker, you can create a press portrait, indicating the speaker’s expertise, topics of interest, background information, and the links to op-eds you’ve already published in media. Once something happens in the industry that is relevant to your company and/or speaker, write a comment and pitch it to a journalist whose beat you know for sure will most likely imply working on this particular story. You can also check #journorequest on Twitter, Quoted, or HARO, as well as check the journalists’ personal Twitter account for similar shoutouts. If you have a ready-to-use press list and a well-established connection with journalists, you can send them a press portrait and introduce your speaker without waiting for the right time. In this case, a journalist may occasionally contact you once they need expert commentary. Option 3: Media Coverage Through Newswire I used to be quite skeptical about the Newswire until one day I shot it and checked the results: 406 pieces of coverage secured across top-tier media outlets, including Business Insider, Seeking Alpha, Benzinga, and others. I made it SEO-wise and inserted the links to the website along with a bunch of anchors and keywords that reinforced our SEO endeavours. Moreover, the journalist who spotted our press release reached out for additional information and wrote his own story based on what we had shared with him. Press release services like PR newswire or crypto-specific Chainwire offer a great deal of features to align your PR efforts with SEO requirements, thereby killing two birds with one shot. Don’t expect a newswire service to bring you a great deal of awareness right away, but rather think out of the box, and see how else can you benefit from distribution and what exactly can you share that would pay off in a long run. A small tip to keep in mind: Don’t rush to the newswire as soon as you have a news peg. Try to pitch it to journalists first; either as an exclusive or under embargo, and only then upload your press release into the newswire system to ensure maximum coverage. Final Words Each option I’ve mentioned above leaves a trace in the media, ensuring greater visibility for your company and products. Forbes isn’t going to cover your news unless you do have the news that is worth being covered on Forbes, and no PR wizard can fix it. Strive for the best you can, but don’t underestimate small achievements and the power of smaller steps that you are making on the way to your goal. Every time I share the latest PR insights, I end up hearing the same feedback over and over again: “Well, now it’s clear. All you need is money.” I don’t know how those keen minds came to such a twisted conclusion, but to me, it’s obvious that there’s always an option to start small. In this post, I’ll share some takeaways for how to gain media coverage and leverage your PR endeavors with no news peg and zero budget. Option 1: Media Coverage Through Op-Ed, Guest Posting, or Columns Media are very fond of experts who tackle industry issues through the lens of their experience and provide editors with fresh and insightful material that would catch a reader’s eye and drive traffic to their website. So, gaining media coverage through op-eds and guest posts is the easiest way to start. Why? Because through the personal expertise of your C-level representative, you will manage to tackle industry issues that somehow relate to your product. How? Here’s a step-by-step guide: How? Here’s a step-by-step guide: Step 1: Decide on a spokesperson and map the topics for them. Step 1: Decide on a spokesperson and map the topics for them. In one circle, the list of the topics that would resonate with the industry that you are working in. In the second circle, note down the vectors in which your speaker has expertise. Then check where those circles overlap, and choose 3-4 major areas to cover. Those ideas should both demonstrate your speaker’s valuable experience and tackle the industry issues for which you already have a solution at hand. Generate a bunch of topics, headlines, and detailed outlines for each. Step 2: Create your media list with relevant contacts. Step 2: Create your media list with relevant contacts. Check which media are covering this or that topic, make a list of such media, find opinions editor’s email, and send them a proposal for approval(consisting of the headline and a detailed outline). Some media like Daily Hodl or HackerNoon have a submission form, but in this case, you have to submit the whole article for review, not just the outline. I would also recommend doing small research to see whether this particular angle hasn’t been covered yet. Step 3: Wait for a response. Great things take time. Step 3: Wait for a response. Great things take time. Don’t share the same content with all the outlets simultaneously. You may end up gaining coverage across multiple outlets and thereby ruin your relations with journalists. Instead, wait for approval from the editorial team (they indicate in the requirements how much time they usually need to review your submission). If there’s no answer within a specified timeframe, then share your idea with others. Pros of leveraging your media relations through guest posts and op-eds: Pros of leveraging your media relations through guest posts and op-eds: Your brand name and maybe the link to the website will appear in the author’s bio. Your brand name and maybe the link to the website will appear in the author’s bio. You will be able to insert some information about your product. If it’s relevant and doesn’t look like a promotion, you can use your product and company’s data to back your stance. You will be able to insert some information about your product. If it’s relevant and doesn’t look like a promotion, you can use your product and company’s data to back your stance. You establish thought leadership and build a strong personal brand for your speaker. If you are a no-name startup with a no-name CEO, building a personal brand is crucial to make people trust you and in you. No wonder smaller companies are ready to pay a fortune to hunt C-level executives with sound names and spotless reputations.
Because by accepting a job offer, this person confirms their trust in the company, making it more credible in the eyes of the industry by default. You establish thought leadership and build a strong personal brand for your speaker. If you are a no-name startup with a no-name CEO, building a personal brand is crucial to make people trust you and in you. No wonder smaller companies are ready to pay a fortune to hunt C-level executives with sound names and spotless reputations. Because by accepting a job offer, this person confirms their trust in the company, making it more credible in the eyes of the industry by default. You establish thought leadership and build a strong personal brand for your speaker. If you are a no-name startup with a no-name CEO, building a personal brand is crucial to make people trust you and in you. No wonder smaller companies are ready to pay a fortune to hunt C-level executives with sound names and spotless reputations. Because by accepting a job offer, this person confirms their trust in the company, making it more credible in the eyes of the industry by default. Option 2: Media Coverage Through Commentaries Nobody accepts commentaries from people with no prior trace left in the media. So, you better start with the opinion part and gradually add commenting atop this. Crypto is a very wavering environment with a lot of stuff going on. Keep track of everything happening in the industry. Start your morning by checking Feedly and Google News to never miss a beat. There are several ways how you can ensure media coverage through expert commentaries: As soon as you have secured 5-6 pieces of coverage for your speaker, you can create a press portrait, indicating the speaker’s expertise, topics of interest, background information, and the links to op-eds you’ve already published in media. Once something happens in the industry that is relevant to your company and/or speaker, write a comment and pitch it to a journalist whose beat you know for sure will most likely imply working on this particular story. You can also check #journorequest on Twitter, Quoted, or HARO, as well as check the journalists’ personal Twitter account for similar shoutouts. If you have a ready-to-use press list and a well-established connection with journalists, you can send them a press portrait and introduce your speaker without waiting for the right time. In this case, a journalist may occasionally contact you once they need expert commentary. Option 3: Media Coverage Through Newswire I used to be quite skeptical about the Newswire until one day I shot it and checked the results: 406 pieces of coverage secured across top-tier media outlets, including Business Insider, Seeking Alpha, Benzinga, and others. I made it SEO-wise and inserted the links to the website along with a bunch of anchors and keywords that reinforced our SEO endeavours. Moreover, the journalist who spotted our press release reached out for additional information and wrote his own story based on what we had shared with him. 406 pieces of coverage Press release services like PR newswire or crypto-specific Chainwire offer a great deal of features to align your PR efforts with SEO requirements, thereby killing two birds with one shot. Don’t expect a newswire service to bring you a great deal of awareness right away, but rather think out of the box, and see how else can you benefit from distribution and what exactly can you share that would pay off in a long run. A small tip to keep in mind: Don’t rush to the newswire as soon as you have a news peg. Try to pitch it to journalists first; either as an exclusive or under embargo, and only then upload your press release into the newswire system to ensure maximum coverage. Final Words Each option I’ve mentioned above leaves a trace in the media, ensuring greater visibility for your company and products. Forbes isn’t going to cover your news unless you do have the news that is worth being covered on Forbes, and no PR wizard can fix it. Strive for the best you can, but don’t underestimate small achievements and the power of smaller steps that you are making on the way to your goal.

Media Coverage Options for a Startup With No News Pegs and Zero Budget 

Table of Links Abstract and I. Introduction II. Preliminaries III. Proposed Design: IBis IV. Detailed Construction V. Implementation on DAML VI. Evaluation VII. Conclusion and References II. PRELIMINARIES A. Background AI model training. In general, the training process for AI models is continuous and iterative, containing training, retraining, and fine-tuning [29]. As seen in Fig.1, training begins with data collection, where initial training datasets are gathered through data scraping. These datasets are then fed into the model training step, where a preliminary model is trained. To ensure the model remains effective and up-to-date, it undergoes periodic retraining with newly collected data, allowing it to adapt to new information. Additionally, a model may undergo a fine-tuning phase, where it is slightly retrained to meet specific domain requirements, enhancing its accuracy and relevance for targeted applications. Copyrights. Copyright grants creators exclusive rights to their original expressions such as literary, artistic, and musical works. This legal framework safeguards creators’ rights, allowing them to control how their work is used, reproduced, and distributed. Copyright protection is automatic upon creation, but it is implicit, requiring additional steps for proper protection. First, registering the work with the copyright office offers authoritative legal evidence of ownership and eligibility for statutory damages in case of infringement. Second, adding a copyright notice (©) with the creator’s name and the year of creation informs others of the copyright claim [30], akin to a signature on a picture. Additionally, NFTs [31] offer a novel method for embedding ownership of digital art via blockchain technology, with ownership automatically claimed upon minting. Licensing is the primary method for granting or transferring the rights to a work. Creators can control the scope of rights by specifying terms and conditions within the license agreement. Licenses may vary widely, from granting broad permissions to restricting usage to specific purposes or timeframes B. Related Work Protecting copyright/data in AI. Data and copyright protection in AI services is a long-standing topic. Existing methods can classified in several aspects [32]: Data-modifying approaches involve modifying or sanitizing user data to unlink them from specific individuals (e.g., k-anonymity [33], differential privacy [34], and watermarking [35]). This minimizes the risk of reidentification by removing or concealing Personally Identifiable Information (PII). Data-encrypting approaches encrypt user data to ensure integrity and confidentiality during data sharing, leveraging techniques such as homomorphic encryption [36] and secure Multi-Party Computation (MPC) [37], [38]. Data-minimizing approaches aim to boost efficiency by reducing the volume of personal data needed [39], often observed in general model training where PII data are not required during training and minimally during inference. Dataconfining approaches involve AI methods that operate without sharing PII data beyond user boundaries [40], ensuring data integrity and confidentiality while enabling effective personalization through local access to personal data. Blockchain-empowered copyright management. Liang et al. [41] employed smart contracts to establish a homomorphic encryption mechanism aimed at safeguarding circuit copyrights. Liu et al. [42] employed a blockchain-based fraudproof protocol to secure ownership rights over AIGC (artificial intelligence-generated content). Numerous similar solutions are outlined in studies such as [43]–[45]. It is worth noting that most existing blockchain-based studies treat each copyright merely as a form of non-fungible online property, akin to an NFT. However, this approach restricts its practical utility in real-world scenarios that require varied operations like registration, renewal, and termination – features that our framework offers in contrast. Leveraging blockchain in AI. Recent studies made efforts to empower AI and foundational models with blockchain technology, aiming to build a more robust and trustworthy AI in distributed environments. IronForge [46] proposes a decentralized federated learning framework that integrates a distributed ledger and a Directed Acyclic Graph (DAG)-based data structure to asynchronously distribute training resources. Petals [47] is a distributed deep learning system that can effectively operate and refine complex models. It utilizes volunteer computing, outperforming traditional RAM offloading, particularly in autoregressive inference tasks. BlockFUL [48] introduces a decentralized federated unlearning framework that utilizes a redesigned blockchain structure leveraging Chameleon Hash. It decreases the computational and consensus costs associated with unlearning tasks. GradientCoin [49] introduces a theoretical concept for a decentralized LLM that functions akin to a Bitcoin-like system. Authors:
(1) Yilin Sai, CSIRO Data61 and The University of New South Wales, Sydney, Australia;
(2) Qin Wang, CSIRO Data61 and The University of New South Wales, Sydney, Australia;
(3) Guangsheng Yu, CSIRO Data61;
(4) H.M.N. Dilum Bandara, CSIRO Data61 and The University of New South Wales, Sydney, Australia;
(5) Shiping Chen, CSIRO Data61 and The University of New South Wales, Sydney, Australia. This paper is available on arxiv under CC BY 4.0 DEED license. Table of Links Abstract and I. Introduction Abstract and I. Introduction II. Preliminaries II. Preliminaries III. Proposed Design: IBis III. Proposed Design: IBis IV. Detailed Construction IV. Detailed Construction V. Implementation on DAML V. Implementation on DAML VI. Evaluation VI. Evaluation VII. Conclusion and References VII. Conclusion and References II. PRELIMINARIES A. Background AI model training. In general, the training process for AI models is continuous and iterative, containing training, retraining, and fine-tuning [29]. As seen in Fig.1, training begins with data collection, where initial training datasets are gathered through data scraping. These datasets are then fed into the model training step, where a preliminary model is trained. To ensure the model remains effective and up-to-date, it undergoes periodic retraining with newly collected data, allowing it to adapt to new information. Additionally, a model may undergo a fine-tuning phase, where it is slightly retrained to meet specific domain requirements, enhancing its accuracy and relevance for targeted applications. AI model training. Copyrights . Copyright grants creators exclusive rights to their original expressions such as literary, artistic, and musical works. This legal framework safeguards creators’ rights, allowing them to control how their work is used, reproduced, and distributed. Copyrights Copyright protection is automatic upon creation, but it is implicit, requiring additional steps for proper protection. First, registering the work with the copyright office offers authoritative legal evidence of ownership and eligibility for statutory damages in case of infringement. Second, adding a copyright notice (©) with the creator’s name and the year of creation informs others of the copyright claim [30], akin to a signature on a picture. Additionally, NFTs [31] offer a novel method for embedding ownership of digital art via blockchain technology, with ownership automatically claimed upon minting. Copyright protection Licensing is the primary method for granting or transferring the rights to a work. Creators can control the scope of rights by specifying terms and conditions within the license agreement. Licenses may vary widely, from granting broad permissions to restricting usage to specific purposes or timeframes Licensing B. Related Work Protecting copyright/data in AI. Data and copyright protection in AI services is a long-standing topic. Existing methods can classified in several aspects [32]: Data-modifying approaches involve modifying or sanitizing user data to unlink them from specific individuals (e.g., k-anonymity [33], differential privacy [34], and watermarking [35]). This minimizes the risk of reidentification by removing or concealing Personally Identifiable Information (PII). Data-encrypting approaches encrypt user data to ensure integrity and confidentiality during data sharing, leveraging techniques such as homomorphic encryption [36] and secure Multi-Party Computation (MPC) [37], [38]. Data-minimizing approaches aim to boost efficiency by reducing the volume of personal data needed [39], often observed in general model training where PII data are not required during training and minimally during inference. Dataconfining approaches involve AI methods that operate without sharing PII data beyond user boundaries [40], ensuring data integrity and confidentiality while enabling effective personalization through local access to personal data. Protecting copyright/data in AI. Blockchain-empowered copyright management. Liang et al. [41] employed smart contracts to establish a homomorphic encryption mechanism aimed at safeguarding circuit copyrights. Liu et al. [42] employed a blockchain-based fraudproof protocol to secure ownership rights over AIGC (artificial intelligence-generated content). Numerous similar solutions are outlined in studies such as [43]–[45]. It is worth noting that most existing blockchain-based studies treat each copyright merely as a form of non-fungible online property, akin to an NFT. However, this approach restricts its practical utility in real-world scenarios that require varied operations like registration, renewal, and termination – features that our framework offers in contrast. Blockchain-empowered copyright management. Leveraging blockchain in AI. Recent studies made efforts to empower AI and foundational models with blockchain technology, aiming to build a more robust and trustworthy AI in distributed environments. IronForge [46] proposes a decentralized federated learning framework that integrates a distributed ledger and a Directed Acyclic Graph (DAG)-based data structure to asynchronously distribute training resources. Petals [47] is a distributed deep learning system that can effectively operate and refine complex models. It utilizes volunteer computing, outperforming traditional RAM offloading, particularly in autoregressive inference tasks. BlockFUL [48] introduces a decentralized federated unlearning framework that utilizes a redesigned blockchain structure leveraging Chameleon Hash. It decreases the computational and consensus costs associated with unlearning tasks. GradientCoin [49] introduces a theoretical concept for a decentralized LLM that functions akin to a Bitcoin-like system. Leveraging blockchain in AI. Authors: (1) Yilin Sai, CSIRO Data61 and The University of New South Wales, Sydney, Australia; (2) Qin Wang, CSIRO Data61 and The University of New South Wales, Sydney, Australia; (3) Guangsheng Yu, CSIRO Data61; (4) H.M.N. Dilum Bandara, CSIRO Data61 and The University of New South Wales, Sydney, Australia; (5) Shiping Chen, CSIRO Data61 and The University of New South Wales, Sydney, Australia. Authors: Authors: (1) Yilin Sai, CSIRO Data61 and The University of New South Wales, Sydney, Australia; (2) Qin Wang, CSIRO Data61 and The University of New South Wales, Sydney, Australia; (3) Guangsheng Yu, CSIRO Data61; (4) H.M.N. Dilum Bandara, CSIRO Data61 and The University of New South Wales, Sydney, Australia; (5) Shiping Chen, CSIRO Data61 and The University of New South Wales, Sydney, Australia. 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 available on arxiv

How Blockchain Enhances Copyright Protection in AI Model Workflows

Do you read confidentiality agreements? Few people waste time. They all look alike, are boring, and probably don't lie. Companies, especially big ones, try to be honest about their responsibilities regarding user data if only to avoid hefty fines. And yet, something is wrong with the world's privacy protections. What if the concept of privacy itself is the problem? Decentralized projects don't usually say “privacy”; rather, they say data is “owned by the user” (data ownership). What's the difference? I came up with a fun analogy and want to share it. Let's take a more understandable subject - land, for example. If you own a plot of land, you have the right to get there whenever you want, build a house on it, make money with it, invite your friends to a party, and forbid anyone to enter. Ownership is the right to make decisions about some object. From a legal point of view, it consists of three elements: Possession (i.e., the possibility of unrestricted access)


Use (i.e., benefiting from the property)


Disposition (the right to sell, gift, or destroy the object) An owner can transfer some of his or her rights to another. For example, if you lease land, you will allow the tenant access and use, but not disposal. This is the classic, universal understanding of ownership. There are good reasons to extend it to include information. First, data plays a role in the modern economy; in the 21st century, it has become a valuable asset with measurable value. Look at the largest companies in the world: 4 out of 10 derive a significant part of their revenues from user data. Second, thanks to decentralized architecture and asymmetric cryptography, this has become feasible in principle: data ownership can be enforced. “Privacy” means that the data is kept secret and protected by the server. The server takes over the authentication and authorization functions to make the protection work. At its discretion, it can allow or deny someone access and use. In other words, the server owns the decisions - and so, it turns out, does the data. Can the server block a user? It can. Can it grant access to someone else for some reason - like a hacker? It can. Can someone else use your data for profit? Yes, that's the whole point of free digital services. The provider can destroy your account at will, as Google does, or take away your name, as Twitter does. And users are used to this state of affairs. As the former owner of the Music account on the service formerly known as Twitter put it, “Ultimately, they own the account.” As strange as it may sound, the idea of privacy on the Internet is reminiscent of medieval feudalism. In the Middle Ages, land was the main economic resource. Peasants used the land, but it belonged to the aristocracy. In return, the feudal lord was obligated to protect that land. Sound familiar? By allowing internet services to be in charge of privacy, we are adopting a feudal model: when our digital identity is protected, it is at the mercy of someone else. That's why tech companies cling so tightly to privacy. And why privacy laws being passed around the world exacerbate digital inequality. Internet giants limit our access to our own data, calling it an advantage. The very concept of privacy masks the fact that data does not belong to us. Let's be clear: this is not a global conspiracy theory but rather an attempt to find the root of the problem. Here's another classic legal principle: Cuius est solum, eius est usque ad coelum et ad inferos - To whomsoever it belongs, it is his all the way to the heavens (and all the way to hell) (Latin) I like the sound of that. This is exactly the kind of complete and unconditional ownership of one's own information made possible by decentralized, trustless networks with strong cryptography. That's why I believe Web3 is more than cryptocurrency speculation. It is the beginning of a new era of people's relationship with technology. Do you think the analogy with the Middle Ages is clickbait? Or does it seem to be true? Do you read confidentiality agreements? Few people waste time. They all look alike, are boring, and probably don't lie. Companies, especially big ones, try to be honest about their responsibilities regarding user data if only to avoid hefty fines. And yet, something is wrong with the world's privacy protections. What if the concept of privacy itself is the problem? Decentralized projects don't usually say “privacy”; rather, they say data is “owned by the user” (data ownership). What's the difference? I came up with a fun analogy and want to share it. Let's take a more understandable subject - land, for example. If you own a plot of land, you have the right to get there whenever you want, build a house on it, make money with it, invite your friends to a party, and forbid anyone to enter. Ownership is the right to make decisions about some object. From a legal point of view , it consists of three elements: From a legal point of view Possession (i.e., the possibility of unrestricted access) Use (i.e., benefiting from the property) Disposition (the right to sell, gift, or destroy the object) Possession (i.e., the possibility of unrestricted access) Possession (i.e., the possibility of unrestricted access) Use (i.e., benefiting from the property) Use (i.e., benefiting from the property) Disposition (the right to sell, gift, or destroy the object) Disposition (the right to sell, gift, or destroy the object) An owner can transfer some of his or her rights to another. For example, if you lease land, you will allow the tenant access and use, but not disposal. This is the classic, universal understanding of ownership. There are good reasons to extend it to include information. First, data plays a role in the modern economy; in the 21st century, it has become a valuable asset with measurable value. Look at the largest companies in the world : 4 out of 10 derive a significant part of their revenues from user data. Second, thanks to decentralized architecture and asymmetric cryptography, this has become feasible in principle: data ownership can be enforced. largest companies in the world “Privacy” means that the data is kept secret and protected by the server. The server takes over the authentication and authorization functions to make the protection work. At its discretion, it can allow or deny someone access and use. In other words, the server owns the decisions - and so, it turns out, does the data. Can the server block a user? It can. Can the server block a user? It can. Can it grant access to someone else for some reason - like a hacker? It can. Can it grant access to someone else for some reason - like a hacker? It can. Can someone else use your data for profit? Yes, that's the whole point of free digital services. Can someone else use your data for profit? Yes, that's the whole point of free digital services . whole point of free digital services The provider can destroy your account at will, as Google does , or take away your name, as Twitter does . Google does Twitter does And users are used to this state of affairs. As the former owner of the Music account on the service formerly known as Twitter put it, “Ultimately, they own the account.” As strange as it may sound, the idea of privacy on the Internet is reminiscent of medieval feudalism. In the Middle Ages, land was the main economic resource. Peasants used the land, but it belonged to the aristocracy. In return, the feudal lord was obligated to protect that land. Sound familiar? By allowing internet services to be in charge of privacy, we are adopting a feudal model: when our digital identity is protected, it is at the mercy of someone else. That's why tech companies cling so tightly to privacy. And why privacy laws being passed around the world exacerbate digital inequality. Internet giants limit our access to our own data, calling it an advantage. The very concept of privacy masks the fact that data does not belong to us. Let's be clear: this is not a global conspiracy theory but rather an attempt to find the root of the problem. Here's another classic legal principle: Cuius est solum, eius est usque ad coelum et ad inferos - To whomsoever it belongs, it is his all the way to the heavens (and all the way to hell) (Latin) Cuius est solum, eius est usque ad coelum et ad inferos - To whomsoever it belongs, it is his all the way to the heavens (and all the way to hell) (Latin) I like the sound of that. This is exactly the kind of complete and unconditional ownership of one's own information made possible by decentralized, trustless networks with strong cryptography. That's why I believe Web3 is more than cryptocurrency speculation. It is the beginning of a new era of people's relationship with technology. Do you think the analogy with the Middle Ages is clickbait? Or does it seem to be true?

Stories, News & Mentions about FIRSTDATA

Firstdata’s Stories on HackerNoon

Firstdata's News & Mentions