Web3 AI: a change that matters

Web3 AI can be seen as an umbrella term for projects that contribute to the Decentralized AI space, it could be as big as distributed Enterprise Grade GPUs what we do at Aethir , all the way to the amazing Web3 AI influencers that you will soon see from SUBBD and AI robots powered by decentralized Spacial AI powered by Auki Labs .

In between, we see the emergence of (trusted) data-layers needed to train AI, including Barcelona based Context Protocol and Dutch Nuklai . And there are even projects binding it all together with Decentralized Operating Systems for AI, such as 0G.ai and Autonomys .

We even see giant mergers happening in the Web3 AI space, similar to those we have seen in the Web2 industries as well. The latest, and now most famous one being the merger between SingularityNET, Fetch and Ocean protocol into the Artificial Superintelligence Alliance

You have been using AI for quite some time already, it is in the camera of your phone and in the Smart TV you use to watch Netflix on every evening. Whereas Netflix uses tech to estimate what they think you will want to see, based on your 'thumbs up' and viewing history, your camera can optimize images that are taken in the dark of from very far away.

So what?

You have probably already been playing around with generative video creation, a GPT to chat with (voice or text), AI voice overs or AI generated podcasts, and you have likely seen fake news generated by AI already as well. In fact, we see loads of AI generated stuff each day.

In these days it is almost unimaginable that someone (you?) still have not trained your own AI assistant to help you with your planning, bounce ideas off, get feedback on conversations from etcetera!

Some say that the web as we know it is a thing of the past, and that all the information we consume will soon come to us through AI.

This is a likely development, and one that puts a lot of power in the hands of a very select group of companies, that have one thing in common. They need to make a profit (from you).

During a small gathering with Binance 's CZ, Arthur Hayes and Nils Pihl I asked Nils the question "What keeps him up at night". His answer is one that touches us all. We use AR glasses, Metaverse tools, Roomba Robots and other smart devices and all these devices generate data. They literally have the power to map your home and in some cases give a literal view inside your home.

You don't invite strangers into your home and allow them to freely browse around, right?

Now that is exactly what is happening in Spatial AI. Large corporations do everything they can to get as much data from you as possible, including what is in your home, how you wake up, what time you wake up, what you eat etc. All of that with one objective, to influence you!

These developments are, scary to say the least. At least for those who love Web3, and AI at the same time (crazies like me, and perhaps like you). In Web3, we design alternatives. We won't replace these centralized AI's and systems, but we can offer the choice.

At Aethir we acknowledged that one part of the problem is that smaller companies hardly have any access to hardware needed to train their AI. And, as compute is often underutilized in data-centers, we saw that as an opportunity to create an environment where everyone benefits. We connect underused hardware onto our marketplace and we offer that to projects that need it. These projects are often smaller scale Web3 AI projects that, in turn, create community owned AI's (often Small Language Models).

Decentralization means, in this case, that data centers benefit as their GPU is being repurposed, Aethir community benefits as they secure the network and get rewarded for it in $ATH tokens, and the AI partners benefit as they get access to affordable GPUs that power the next generation of Web3 AI. We all win, a true ecosystem play!

Web3 AI is

Web3 AI is all about ownership and privacy, where does the data come from that is used to train an AI? If you create your own Small Language Model (SLM), who then actually owns/controls it and is there a party that can take it away from you? Can you store your own trained AI on your own device, and get it to learn from other small language models?

Can you train your own AI models, on data you own and that is personal to you, in a way that the AI becomes a part of your life? Like a coach that helps you through different phases in your life? Can you even store it on a device you own?

There are many Web3 AI companies that are trying to answer the aforementioned questions, and I freaking love that!

An example of an SLM

I have offered the following idea to my colleagues at the University (this really is a step too far for many, but it will come). What if we help students create their own AI-powered assistant that is a specialist on our programme and that can help a student grow throughout her 4-year studies with us? We train the model on everything the student needs to know, such as industry resources, course guides, programme guides, information on institutional FAQs etcetera. And from there onwards the student continues to train her assistant every day.

What were you good at? What topics did you love? What topics did you hate? What were your exam scores and what was the feedback given? When do you learn the best and when is the best time to relax? What competencies are you working on, and how do you do that?

You know, stuff every coach asks.

This is an absolute game changer, and lets agree that it is something super special that we can do today. The same idea can be applied in many other scenarios of course.

The problem though is, who owns this AI? Can OpenAI just take the information the student used to train the AI-assistant? And what if OpenAI has to file for bankruptcy, will someone else be able to buy your data and your assistant? If you use data to train an AI, then who owns that data? What about privacy concerns?

Even in that case, a Web3 approach to AI very much makes sense.

A new AI-driven Web that you (co-)own?

These ownership-, control- and privacy topics are challenges that Web3 people love to crack. And they take it further, what if you can monetize the AI you created? What do new AI-economies look like (Read my earlier article about that)? Can we have machine-AI to Machine-AI economies?

In Web3 we all know that one can not, and should not build all of that alone, the more we do it as an ecosystem the more likely it is that we will succeed. Together are building alternatives to the centralized (and in our view dangerous) AI-systems.

One more vision of the future

I envision a world where people own hardware in their homes, like Aethir Edge devices. They will be able to store their personal AI on that device, and that AI can live alongside of us for as long as we live. They can be trained on day one on our personal data (like medical records etc.) and they can learn what is important to us. They evolve with us, as we grow older and they get smarter. We ask it questions, use it to get help, and perhaps one day even upload it into our own robot assistants, powered by our own AI.

Each of these AI's will be unique, unique to us, our lives and the world we live in. It will be able to support us, from the day we are born until the day we .....