Decentralized AI, The Perfect Synergy

🏝 TheTechOasis 🏝

part of the:

Breaking down the most advanced AI systems in the world to prepare you for the future.

10-minute weekly reads.


  • Leaders: Decentralized AI, the Perfect Symbiosis Between AI & Blockchain

  • Did you know about… Elon Musk’s Grok 1.5 Vision New Model & The New Unexpected King of AI Music?

🥇 This week on Leaders… 🥇

While there’s a lot to celebrate with the amazing progress being made in the AI space, one trend keeps getting worse.

AI is gradually becoming less open and more suited for a couple of incumbents to accrue trillions of dollars in value leaving only but the crumbs for the rest.

Hence, Decentralized AI is one of the last bastions where the open community may be capable of holding its ground. Thus, even though you may be skeptical right now, I can guarantee you that blockchains hold immense, non-speculative value for AI, a value you simply cannot disregard.

I am not a financial advisor, hence I have purposely avoided mentioning any single cryptocurrency in this post, as I want to convey the importance of the underlying technology, not any single crypto in particular.

The Increasing Dangers of Centralized AI

When even King Charles III of England describes AI as “one of the greatest leaps in human history”, you know things are for real.

Heck, he even compared it to the “harnessing of fire”, which is a hell of a statement if you ask me.

Hype aside, the potential for AI to disrupt everything is unlimited.

And no, I am not referring to AI as a productivity booster, which is undoubtedly the first hill it will conquer, or as the brain of the humanoids that will one day inhabit our world, but also about AI’s potential as a ‘discovery tool’.

From Halicin and CRISPR to Maths

For starters, AI is already playing an increasingly vital role in healthcare. For instance, back in 2019, it helped humanity discover a new antibiotic, halicin.

But this is just the tip of the iceberg. As covered by GEN, AI’s role in drug discovery is already invaluable in detecting key patterns that are otherwise oblivious to us.

Also, as recently covered in a previous Leaders segment, EVO, a biological foundational model for DNA, is showing great promise in uncovering the secrets of the human genome, and has even shown potential in supporting the creation of novel gene-editing CRISPR systems that can help engineer future human generations to being more resistant to illness, or even to climate change.

Through the use of semantic spaces, AI is also helping us discover new emotions, breaking from the theories that have shaped our understanding for decades.

What’s more, semantic space theory is also helping us create odor maps to create new smells that help us protect our crops from insects, among other exciting new possibilities acknowledged by the Melinda & Gates Foundation.

Still, as I said earlier, AI is ubiquitous, and it’s also helping us reach new heights in maths, as Google Deepmind’s AlphaGeometry is discovering new simpler and more elegant ways to prove geometry theorems, or FunSearch, another system that helps us discover new solutions to maths problems with Large Language Models (LLMs).

But as I mentioned earlier, besides being a discovery tool, AI is also a massive productivity booster. And that, well, has some unintended consequences.

The Great Displacement

Today, you won’t find a single study on the future of work where AI isn’t given a protagonist role as a net job destructor.

Even the most skeptical acknowledge its job destruction capabilities, and many prominent organizations like the IMF have been highly outspoken on the risks.

Probably the first real example, besides the tech industry (where we could make the case that the layoffs were more about being completely overweight in staff terms) is Klarna.

Acknowledged by the company, their AI customer agent already handles two-thirds of support chats and performs the equivalent throughput to 700 human agents.

But, truth be told, in the grand scheme of things, the poop is nowhere close to hitting the fan. As predicted by OpenAI, McKinsey (America-only study), or PwC (UK-only study), the number of AI-exposed jobs will reach insane numbers by 2030.

Yet, job destruction isn’t the only issue that AI generates, as labs are indulging in some dubious acts too.

Suits Abound

A recent New York Times article we mentioned on Thursday made headlines recently. In it, the NYT described how most big tech companies committed copyright fraud to train their models by using unlicensed data.

Besides the ongoing litigation between NYT and OpenAI, OpenAI is also being hit from other fronts.

However, as Google has omitted any lawsuits against OpenAI despite knowing the latter used YouTube videos to train Sora (which basically means they are into it too) chances are that the endgame is that all data will become licensed-due very soon.

This scenario, where the most valuable asset in AI, data, becomes scarce and expensive, is the worst possible thing that could happen to the open-source community, and the best thing possible for the super-rich incumbents as only they will be able to access it.

And then there’s the threat of “God AGI”, the worst possible outcome for AI.

Do We Really Want God AGI?

Whatever Artificial General Intelligence (AGI) actually is, the trend today is that, when achieved, it will happen behind closed doors and in a centralized manner.

As Emad Mostaque described it four days ago, it’s like a “God AGI”. Assuming this god-like ‘thing’ can be contained, the incumbents leveraging it could hold the biggest power and control asymmetry the world has ever seen.

This threat, although may seem distant, is very, very real, and players like Microsoft might be making moves already, with examples like Project Stargate.

So, the question now is, how can blockchains help in all this mess?

A Just and Decentralized World

Even though the crypto industry is 99% a useless pile of speculation-fueled crap, the underlying technology (blockchain) has real value and powerful synergies with AI.

But first, what is a blockchain?

Decentralized Ledgers

A blockchain is a decentralized ledger. In layman’s terms, it’s a registry of transactions that is extremely hard to tamper with (immutable), where every transaction can—and is—verified (trustless), where everyone can participate (uncensorable and permisionless), and geographically unbound (globally distributed).

Truth be told, most dedicated compute today is heavily concentrated in some areas, so the last point is yet to be fully achieved.

But importantly, this ledger is distributed. What this means is that the validation of the transactions occurring in it is performed by a decentralized network of ‘nodes’, instead of one central validator, with a consensus mechanism that allows them to ‘vote’ and decide, for every single transaction, if it’s valid or not.

Importantly, not only does it imply that it’s extremely hard to tamper with the network (for example the Bitcoin network has even been hacked), but this verification is performed through public-key cryptography, meaning that decisions are made purely based on maths.

And where do cryptocurrencies fit in this picture?

Simply put, they are the economic incentivize for nodes to participate in securing the network. “Show me the incentive, and I’ll show you the outcome” as the late Charlie Munger once said.

It’s fairly simple, if I provide my resources to verify the network (which usually means providing compute power or staking some of my crypto) I need a way to get rewarded for it.

Consequently, the value of these cryptos in open markets are a statement of the value of the underlying network. However, it must be said that most of the value of cryptocurrencies today is purely speculative, period.

But why is having such a distributed registry so valuable?

Toward a trustless society

Today, our entire economy is based on trust. The biggest proof of this is fiduciary currencies, like the US dollar or the Euro, whose value is backed by the issuing government, which is a euphemism for ‘trust me bro, my currency is valuable’.

Promulgated by the ‘Nixon shock’, in 1971 the US dollar stopped being backed by gold. This meant that the US dollar’s value, from then on, would be valued based on ‘vibes’.

Jokes aside, what this meant was that the US dollar stopped having an absolute value in gold, but would be based on relative value to other fiduciary coins, aka the famous exchange rates.

The Euro was born directly as a fiduciary currency in 1999.

It’s funny if you think about it. Today, the vast majority of citizens hold the vast majority of their wealth on an asset that has no intrinsic value (that’s literally the definition of fiduciary). Yet, the overall trust of society over these currencies gives it its value.

It’s a beautiful thing when it works, but when extremely expansive monetary policies like the ones we saw in 2021 are executed, where US dollars and Euros flood the market, they cause an effect known as inflation.

And in some extreme cases, like the Argentinian Peso or post-World War I Germany’s German mark, when trust in the value of the currency is lost hyperinflation begins, which is a race to the bottom in terms of value.

Consequently, with fiduciary currencies, banks take a prominent role in society as trust enablers, as they guarantee that people’s commitments to pay are based on actual money, and banks guarantee the liquidity and solvency of their clients.

Despite this, in a trust-based society corruption is a common thing, as there are ways to circumvent these guardrails to your benefit.

But what if we could create a system where trust is not required and where people’s claims about how much money they have can be mathematically verified?

And more importantly, how does that affect AI in all this?

Well, in fact, blockchains are the only way we can guarantee that the value generated by AI is accrued by society and not by a selected few.

Distributing AI’s Value to the People

By now you probably have an idea of what Decentralized AI is.

In simple terms, these are networks where the data gathering and ingestion, the training and fine-tuning of AI models, and the inference are performed in a distributed manner.

This may sound very similar to ‘distributed computing’ where current huge LLMs can be trained throughout a network of GPU data centers. However, the key difference is that these datacenters are all controlled by the same entity.

But the key role of blockchains in all of this is traceability. Today, we are seeing multiple accusations against AI, not only to OpenAI but also to other companies like Suno or Udio in the case of music, where artists are ready to go to war.

But I can guarantee you would not want to be in the position of that judge if a lawsuit came her/his way.

A Deceiving Art

Copyright suits against AI are extremely complex to manage. How do you prove you were indeed scammed?

Sure you can prove that an AI system ingested your music, or your art, but how is that different from a human that learns music by reviewing Mozart’s pieces?

Part of the creative process of art is to be inspired by the works of others, and in that humans are no different from AI, so the latter shouldn’t be treated unfairly.

The only way a copyright lawsuit earns a victory for the plaintiff is if the output generated by the AI is too similar to the works of others (just like copyright infringement between humans works today, as being inspired by someone else’s music as part of your creative process and plagiarizing them is two completely different things).

But here’s the thing. While copyright infringement lawsuits won’t probably hold in court, the artists whose data was used to train the models have every right to be compensated for their work.

And here is where blockchains come in as a way to create new economies at scale around the biggest goldmine in the 21st century: data.

But how?

Subscribe to Leaders to read the rest.

Become a paying subscriber of Leaders to get access to this post and other subscriber-only content.

Already a paying subscriber? Sign In

A subscription gets you:
High-signal deep-dives into the most advanced AI in the world in a easy-to-understand language
Additional insights to other cutting-edge research you should be paying attention to
Curiosity-inducing facts and reflections to make you the most interesting person in the room