The wildcard: Google
If we're asking “Who has the best chance of eventually winning?”, rather than “who has the best model today?” I might actually choose Google DeepMind.
The above text is a response from ChatGPT when asked, “Who will build the best AI model?” And its reason for choosing Google is below:
Why? Because the endgame probably isn't just a better language model. It is a system combining:
reasoning + agents + multimodality + robotics + scientific discovery + enormous compute + persistent memory + real-world data.
Google has pieces of almost all of that.
Now, here's a part of its response to really think about:
The “winner” may not be the company with the single smartest model. The winner could be whoever figures out how to turn models into millions of reliable autonomous agents at the lowest cost.
I have a confession to make.
This post isn't about pointing out who I think will build the best model. I would say that it really doesn't matter more than what it means for our individual portfolios.
But this isn't an investment discussion.
What I do want to talk about is what being the best in the game could actually look like.
Expensive AI is a bottleneck that needs to be solved.
People think that trying to lower the cost of building AI and deploying it for use is about saving companies some money. But the reality is that it is a necessity. If it isn't achieved, it could cripple the industry at the consumer level.
AI needs to be cheap so that more people can use it.
If more people cannot use AI, not a lot will be produced with it, and the technology will effectively never produce enough economic value to justify its existence.
Think about it for a minute or two, the value of AI isn't just in how intelligent the model is. It's in what that intelligence allows people to do.
A model that is 10% smarter but costs 10 times more to run may not create more value than a cheaper model that is slightly less capable.
The real breakthrough might therefore not be another benchmark victory, although a lot of people in the industry tend to be obsessed with the data.
What would eventually count as a breakthrough worth reporting will be making intelligence so cheap and accessible that millions, or eventually billions, of people can afford to use it every day.
What this tells us is that the winners will be those that figure out a way to get the full benefits of the technology to the end user at the most affordable rates.
And this is where the idea of autonomous agents becomes particularly important.
If AI can move from answering questions to actually doing things like researching, coding, managing businesses, discovering drugs, operating machines, analysing markets, coordinating logistics, then the amount of economic activity it can generate becomes enormous.
But only if the cost of doing those things makes sense.
The lower the immediate cost and the higher the value of its output, the more successful the technology will be.
So I think that maybe people should not be asking the question of who will build the smartest AI and focus more on who will make intelligence the cheapest, most useful and most accessible resource in the world.
