The Real-Time Internet: Visions of AI-Generated Everything

Written by liamshotwell | Published 2024/01/25
Tech Story Tags: futurism | artificial-intelligence | augmented-reality | generative-ai | model-autophagy-disorder | real-time-internet | future-of-the-internet | quantum-like-websites

TLDRThe real-time Internet is a concept by Jared Ficklin where generative AI can create personalized and dynamic software on demand. Interfaces democratize technology. Generative AI will make augmented reality more accessible in the future with natural language prompts. MobiDev AI expert Liubov Zatolokina believes it's possible, but we're not there yet. Also, beware model collapse where AI trains on AI-generated content. via the TL;DR App

Example code generated by Microsoft Copilot. iPhone mockup by zlatko_plamenov on Freepik. Background by FWStudio on Pexels.com

The development of the Internet is a story of democratization. Computing power and information progressively become easier to access over time. Now, in the middle of generative AIā€™s hype cycle, visions of a farther future of the Internet are becoming clearer. In fact, the acceleration of AI technology makes me wonder just how far away we really are from revolution.

That revolution will overturn everything we know about apps, software, and programming. The ā€œreal-time Internetā€ may be on the horizon, and itā€™s wise that business leaders understand its implications now before itā€™s upon us.

An Internet Programmed in Real-Time

Generative AI has the capability to create not just blog posts, poetry, and images. With the right training sets, it can write entire pieces of software. As seen with tools like

Spline, it can work easily with 3D environments for augmented reality projects.

Argo Designā€™s Lead Creative Technologist Jared Ficklin predicts that the future of software and interfaces will become increasingly personalized and produced in real-time by AI.

In an episode of Freethinkā€™s Hard Reset, Ficklin goes into great detail about his vision of a world computed by generative AI in real-time. He describes it as a natural progression of the evolution of the Internet and of computing itself. He also shares insight into the new ways we will interact with computers in the future.

The Past and Future of Interfaces

Ficklin is fascinated by interfaces and how they democratize technology to the masses. Early computing was performed by punch cards, where programmers had to deeply understand how machines functioned. However, punch cards evolved into syntax and programming languages, and then to desktop environments. Now, anyone in the world can use a computer thanks to interfaces like Windows and MacOS.

As of now, creating augmented reality experiences is too complicated without the help of talented software developers. Similar to how Unix terminals and desktop environments made computing more accessible, generative AI may make augmented reality development, or rather, experiences more accessible than ever before.

ā€œQuantum-likeā€ Websites

One example Ficklin points to in the present day that relates to his ideas is ā€œquantum-likeā€ websites. These are web pages or sites that look completely different based on who is viewing them. Amazon is a great example. Every time you visit Amazonā€™s website, it looks different. This is especially true from person to person. Thanks to personalized shopping results, Amazon can programmatically suggest content to you based on your interests and history.

Just like how these websites are unique and personalized to their users, Ficklin imagines that the next step will be completely unique software for each person, all generated by AI on demand.

Creating New Experiences From Scratch: Realtime AR

Ficklinā€™s ā€˜real-time Internetā€™ is a radical concept that rewrites the rules of how software is developed. For augmented reality, this could mean that AR experiences could be written on demand by on-board or cloud-based generative AI. ā€œWeā€™re going to have a new [multimodal] vocabularyā€¦launch that, put it there, close that.ā€

Ficklin explains in a Freethink interview. ā€œā€¦using gestureā€¦voice, and certain control surfaces all together.ā€ In a nutshell, a future augmented reality experience using this ā€˜real-timeā€™ approach might look something like this:

  1. With a wearable mobile computer like an AR headset, dictate: ā€œShow me what it would look like if the wallpaper in this room was blue.ā€

  2. The generative AI model uses natural language processing and understanding to interpret the userā€™s prompt and begins writing an AR program from scratch, using the headsetā€™s sensors as environmental data. The AI creates the app using room-scanning training data and specially tailors the new app to identify the walls of the room.

  3. The program is ready for launch and shows the user what the room would look like with blue wallpaper. Once the user is finished with the experience, the generated software is deleted. Ficklinā€™s interview with Freethink fascinated me. I was curious just how close we are to the future he describes. So, I decided to dig deeper.

Reality Check With an AI Expert

Now, letā€™s be honest. Iā€™m not an AI expert. While I was coming up with this piece, I realized the best thing I could probably do was ask someone who knew what they were doing about their opinion.

Iā€™ve done a lot of writing on behalf of MoibDev in the past, so I couldnā€™t think of anyone better to ask than their AI experts. I asked MobiDev AI Team Leader Liubov Zatolokina for her opinion.

Sheā€™s been helping MobiDevā€™s clients around the world succeed for over four years, so I was really excited to get her opinion on the Real-Time Internet.

Is the Real-Time Internet Even Possible?

ā€œIn my opinion, we already partially have this experience,ā€ Liubov explains. ā€œwe can already create simple prompts like images, music, movies, poetry, and social media posts, and even basic programming code and prototypes.

It wonā€™t always look like a ready-to-use product, but itā€™s an excellent place to start. Very quickly weā€™ve seen an exponential increase in the quality of not only images, text, and software code, but also video creation."

Based on her analysis, it sounds to me like Ficklinā€™s future is possible if AI technology continues to advance at its current pace, but when?

In a very short period of time, weā€™ve seen an exponential increase in the quaity of not only images, text, and software code, but also video creation.

How Far Are We From That Future?

ā€œItā€™s hard to say exactly when the next step will be, but weā€™re already seeing amazing products from leading companies every month,ā€ Liubov continues.

ā€œThere are hundreds of AI-related articles and news from open-source projects every day. This only accelerates and fuels our curiosity. I expect weā€™ll see more improvements to the quality of generation and data processing this year, but it probably wonā€™t lead to AGI or the ā€˜Real-Time Internetā€™. Each innovation is iterative, and with some patience the future will be mind-blowing.ā€

Itā€™s hard to say exactly when the next step will be, but weā€™re already seeing amazing products from leading companies every month

How Would That Future Disrupt the Status Quo of Software Development?

ā€œEvery cutting-edge technology disrupts someoneā€™s status quo,ā€ Liubov asserts, ā€œand software developers are no exception. We are now seeing an increase in the amount of automation in development.

Code writing assistants and testing tools lead to faster development, less red tape, and more time spent on idea generation. I expect the iterative advances of AI to result in a ā€˜snowball effectā€™ in software development and technology sectors.ā€

Every cutting-edge technology disrupts someoneā€™s status quo, and software developers are no exception.

So, Do On-Demand AI-Generated Applications Pose a Threat to the Jobs of Programmers and Software Developers?

Iā€™m particularly interested in whether AI may start to replace existing jobs, and how Ficklinā€™s Real-Time Internet may impact programmers. ā€œOne way or another, progress affects each of us,ā€ Liubov explains.

ā€œBut I canā€™t call on-demand AI generation a threat to software developers. This is the same change for humanity as the invention of the wheel. Some professions may disappear, but many new ones will appear in their place.ā€

This is the same change for humanity as the invention of the wheel. Some professions may disappear, but a large number of new ones will appear in their place.

If More and More Applications Are Built Automatically With AI, How Will That Affect the Pool of Available Training Data?

ā€œSoon, I think weā€™ll see further growth and development in the data storage industry, especially with regards to processing and labeling. The incredible speed of generative AI contributes to the fact that thereā€™s more and more unlabeled, automatically generated data in the network.ā€ Liubov says.

"Although the content might be high quality, itā€™s not real-world data. Real-world examples in those datasets is crucial for proper and high-quality training of models. Re-use of that data leads to ā€˜model collapseā€™, a decrease in input data diversity, leading to a decrease in model quality.ā€ The phenomenon Liubov describes is also called Model Autophagy Disorder (MAD).

Image source.

Re-use of [AI-generated content in datasets] leads to ā€˜model collapseā€™, a decrease in input data diversity, leading to a decrease in model quality.

What Do You Believe Are the Primary Obstacles to Achieving On-Demand AI-Generated Apps, Such as AR Experiences? What Would It Take to Overcome Those Obstacles?

ā€œThe main obstacle is the intersection of available computational power and how much power the AI models need,ā€ Liubov says.

ā€œSo far, AI tasks are best suited for use by powerful cloud computing and parallel processing systems. However, the AI engineering community has already seen that scalability isnā€™t the cure-all in terms of the challenges AI faces. New data processing algorithms combined with new advanced hardware, both focused on energy and computational efficiency, could make things a lot easier.ā€

Based on Liubovā€™s analysis, I suspect that this may be why businesses are starting to pay closer attention to ā€˜on-deviceā€™ generative AI solutions.

Qualcomm and other companies have discussed on-device AI a number of times this year, and some companies like Samsung are already releasing products with these features.

New data processing algorithms combined with new advanced hardware, both focused on energy and computational efficiency, could make things a lot easier.

The Future of the Real-Time Internet

Jared Ficklinā€™s visions of the next generation of computing may not be here just yet. It remains to be seen just how far off the horizon AI-generated apps on demand are, but thereā€™s no denying that weā€™re seeing exponential growth in artificial intelligence. Each advancement iterates on the last, and thereā€™s no sign of stopping.

I still personally have a lot of questions, like how do we ensure that AI-generated apps are secure? How will AI safety play a role in Ficklinā€™s envisioned future? Thereā€™s a lot we donā€™t know, but thereā€™s also a lot to be excited about.

No matter what happens, AI is going to change the world, and we have to prepare for that future.

Thank you Liubov Zatolokina for your insights on this story! Also, thank you for reading! What do you think about the Real-Time Internet? How far do you think we are from that future? Let me know in the comments.


Written by liamshotwell | Freelance SaaS SEO and tech writer. I also play the trumpet.
Published by HackerNoon on 2024/01/25