The promise of "free & infinite" intelligence.

What big piece of our daily activities could be 100X improved through a copilot?

During the last couple of weeks, the subs have been focused on the discussion about AI tools across different industries, DeFi, GameFI, and DePin. I decided to put on hold the discussion between decentralization and centralization. The purpose of this was to extend the conversation by giving examples of decentralized tools, and the benefits of setting up decentralized and open-source frameworks as a cornerstone of your development stack. 

But the core purpose of this substack has been to ask ourselves the question of how to democratize AI.

There is enough proof right now that closed-ended solutions won’t make mass adoption. A lot of them won’t even return the investment in training the models. Yes, the main reason is the cost of training the model, and the amount of data and computational resources required to get good performing results. This makes it exclusive for enterprise companies. But still, Large Language Models like the ones made by Open AI, Antropic, or Deep Mind are aiming to be turned into great generalists. In the words of Emad Mostaque (Stability AI former CEO):  

“When have you seen a bunch of great generalists outperform a specialist in her field?” 

So if you have the superpower to train your copilot and make it a specialist in whatever field you need to, what would you choose? What type of answers or task automation you will be expecting from it? 

I think this is the kind of question we should ask ourselves when it comes to AI adoption. What big piece of our daily activities could be 100X improved through a copilot? If we have that kind of leverage in our daily activities, What should we do with the saved time and energy?

One of the promises of AI is the abundance of knowledge and intelligence. This is also one of the most common fears, if intelligence cost turns close to zero what is the future of the graduate students? 

But let’s focus on the promise, What you will do with the abundance of knowledge, and On what you will focus your attention? 

Here is when the promise of having infinite resources to address cancer, famine, education, and many more should become a purpose and a roadmap rather than another propaganda fairy tail.  

“ We’re moving from a world of static files to one of dynamic information flows. The world we’re stepping into with AI is one where information is constantly updated, combined, and repurposed in real-time. This changes everyone’s relationship with information.” From SchellingAI Subsctack. 

In this kind of world, the capabilities you develop to shape and use the information will represent your competitive advantage as an individual and as an organization.

For me, AI should be decentralized and open source because this gives:

Governance, Collaboration, Trust, Boarder knowledge with a handful of perspectives, and personalization. 

Focused on the last one, what would happen if large enterprises become the sponsors of general infrastructure? And just drop general models as common goods?

This will enable individuals or smaller organizations to develop specialized problem-driven AI solutions. That might be the key to AI mass adoption. Sort of Mark Zuckerbergs' vision for his Llama AI.

 Now if we add to that vision the ability to create assets through our developments and value transition rails to build incentives around it. Transparency, trust, and data immutability guarantee the outcomes of what we have built. Some of the results can be awesome solutions to different problems. Something that resonates with me after hearing and reading Emad Mosquete is every region has its problems, and its priorities to solve them. Most of the time is not even every region, but every city, or even every neighborhood. I am thinking in a geographical way to make a better explanation, but this can be applied to any field and its subfields are just semantics. 

But the problems are as specific as they get, and we have as many problems as we have been able to develop or ignore in the current insane hypergrowth model. So we should be thinking about granular specific solutions to address those problems. Because large monopolies won’t they don’t have the incentives. 

Maybe the promise of abundant knowledge in AI can turn into an actual superpower if we develop the tools to use the machine in real-time, with specific problem-driven data. I think this will be possible in a decentralized world. Where the incentives are driven by the networks and communities, not by CEOs.