The Human Playbook

An independent research project studying the transition into the AI era: part philosophy, part cultural critique from inside the tech industry.

[The Human Playbook @gil2almeida]

@nick_____t exactly the whole contradiction of what’s happening is that technology can succeed completely according to its metric while the human outcome fails.

[The Human Playbook @gil2almeida]

Cool. so here’s another possibility:

- individual mathematicians may become less productive, because generating proofs gets cheap while verifying, understanding and choosing among them becomes the bottleneck
- building new theories may become harder: AI can generate building blocks faster than humans can develop the intuition to know which ones matter
- math may become less democratized maybe producing mathematics gets easier; knowing whether it is meaningful or important may require more expertise than before
- theorem proving may become more relevant, because when proofs are abundant, trusted verification becomes infrastructure
- understanding a proof may become harder, because mathematicians will increasingly encounter proofs they did not construct and whose reasoning exceeds what they would naturally have produced
- mathematicians may have less time to socialize ideas, because AI dramatically increases the volume of ideas, conjectures and proofs competing for their attention

[The Human Playbook @gil2almeida]

Somewhere along the way, we confused the ability to build technology with the authority to define humanity’s future. Those are two separate things.

[Elon Musk @elonmusk]

This is the future we shall bring into being

[The Human Playbook @gil2almeida]

@ayojoestar Or, in a clever 4D chess game, that could actually benefit them a lot financially

[The Human Playbook @gil2almeida]

That’s interesting .., but it only makes sense if the objective function of civilization is computational efficiency. And why assume that? What if increasingly capable intelligence causes civilization to value something directly opposite of that … like physical experience, exploration and more activity/participation? What if intelligence becomes very cheap, and stops being the scarce thing we’re optimizing for. We would then we will focus on what the scarce thing is: matter, energy, attention, experience, human participation, physical presence, authenticity, relationships, time …

[The Human Playbook @gil2almeida]

@aiwithsally Ai promise is abundance for humanity but the business economics is saas enterprise … so do what you will with that conclusion

[The Human Playbook @gil2almeida]

Do you know what the AI business is actually built on? Mathematics. Linear algebra, probability, statistics and optimization are what make these systems work at all. And many of the biggest business challenges is all about how much compute you need, how efficiently models train and run, how reliably they reason, how you scale them … all of which are downstream of mathematical problems.

So dismissing what they say about the underlying technology because they aren’t businesspeople is a pretty strange way to understand an AI business.

[The Human Playbook @gil2almeida]

Im all for adaptation but this advice worked when technology automated tasks. Learn the next skill. But that’s not so simple with this technology. Companies pursue leverage precisely because leverage means more output per unit of labor. If AI becomes an increasingly powerful source of that leverage, then telling an individual worker to “become leverage” doesn’t explain why the company will continue to need their labor. Own the capital” isn’t enough either. If labor’s economic value declines relative to capital, how exactly are people supposed to acquire enough capital through their labor to participate in the upside?

This time around we might need to do some real thinking about the system we are creating if we want to have a saying into the future

[The Human Playbook @gil2almeida]

Unfortunately that will increase with AI productivity.

Less workers → much more output.

Companies are capturing the extra gains.

What’s absurd is that we don’t need zero jobs for something crazy to happen.

Meaning … companies can still employed a few people while workers becoming progressively less important as an input into production and therefore receive a declining share of what the economy produces.

Time for a great awakening. These tools are not here to liberate you unless we start designing the system more intentionally …

[unusual_whales @unusual_whales]

“Workers are earning a smaller and smaller share of the economy’s income,” per NYT

[The Human Playbook @gil2almeida]

Unfortunately that will increase with AI productivity.

Less workers → much more output.

Companies are capturing the extra gains.

What’s absurd is that we don’t need zero jobs for something crazy to happen.

Meaning … companies can still employed a few people while workers becoming progressively less important as an input into production and therefore receive a declining share of what the economy produces.

Time for a great awakening. These tools are not here to liberate you unless we start designing the system more intentionally …

[The Human Playbook @gil2almeida]

Same with people building slop products, they shouldn’t be building software. Or writing. Or … you name it here … ai is a super capability for those with the craft, taste, expertise and curiosity … if you want to simple live a life of grifting ai will be slop for you … maybe that’s the universe way of generating an equilibrium

[The Human Playbook @gil2almeida]

Maybe. But here’s the question I’m stuck on: if understanding is partly produced by the struggle required to arrive at an answer, what happens when we remove more and more of that struggle?

Tao may produce 10 lifetimes of mathematics. But will that mean he has 10 lifetimes of mathematical understanding or that the system does?

[The Human Playbook @gil2almeida]

Excellent qq and gets the heart of the issue. No, humans won’t stop understanding mathematics simply because AI becomes better at mathematics. But we may understand less mathematics if fewer circumstances require us to do the cognitive work from which mathematical understanding emerges.

E.g GPS - we didn’t lose our ability to navigate because GPS became better at navigation. We lost some navigational capability because the incentive to perform the cognitive work disappeared. Why build a mental map, remember places, tolerate getting lost, correct your model, and slowly learn a city when the technology removes the need?

The thing is. Ai is not a gps. It removes the cognitive capability. We happily took the trade with gps because we move our thinking elsewhere. The AI question becomes more consequential because it removes all layers at once.

So I think the debate is focused on the wrong question. It’s not really about humans still be capable of understanding mathematics … that’s trivial. Think about the incentives. What produces the incentive for us to develop understanding when competent cognition is available without it? It’s not just laziness … the annoying process of understanding was building the mathematician. Was building the person …

[Yann LeCun @ylecun]

I said "auto-regressive LLMs, in and of themselves, will not lead human-level AI"
That statement is still totally true.

First, the reasoning abilities of current AI systems are based non-auto-regressive search (which is what I have always advocated for). But AFAICT, they do it in token space, which is limited and inefficient. I have claimed that human-like reasoning must be a search in continuous representation space. It looks like the industry is moving towards that.

Second, the self-improvement methods, as currently practiced, only work for domains where the quality of outputs can be scored without human intervention, such as mathematics, code, and scenarios that can be simulated accurately. Not anything else. Humans and animals learn new skills way more efficiently than current RL methods.

Third, the multimodal capabilities of current AI assistants generally use separately-trained encoders (that are not LLMs). This is also what I've been advocating. Except that I think the best way to do this is with JEPA trained with self-supervised learning. The research community is clearly moving towards that (3000 papers on JEPA in just 4 years).

Fourth, if LLMs were a path to human-level AI, we would have domestic robots and Level-4 or Level-5 self-driving cars for consumers by now. And we don't. We certainly don't have cars that can learn to drive in 20 hours or practice like any teenager. We're still missing something pretty huge to claim human-level intelligence (let alone superhuman).

Sure, we now have computer systems that are impressive, very useful, and whose performance is superhuman in an increasing number of domains (coding being one of them).
But that's true of the entire history of progress in computer technology.

Lastly, there is a basic confusion about what intelligence actually is.
It is not the mere accumulation and regurgitation of existing declarative knowledge (which is essentially what LLMs do).
As Jean Piaget famously said, "intelligence is not what you know, it is what you do when you don't know."
It is your ability to solve new problem without any prior training, to act in previously-unknown scenarios, and to adapt very quickly to new situations with minimal training.

We're still far from that.

[The Human Playbook @gil2almeida]

@GregoryConti19 the whole contradiction of what’s happening is that technology can succeed completely according to its metric while the human outcome fails.

[The Human Playbook @gil2almeida]

@micsolana We would have a successful consumer ai product by now

[The Human Playbook @gil2almeida]

@wyakyro_ I think it’s so retarded that our idea of best bodies for robots to handle a multitude of tasks are human bodies. It shows our lack of creativity …

[The Human Playbook @gil2almeida]

To be honest, if the labs were straightforward about the path to IPO and the pressure of market expectations, I think it would have given them more leeway to discuss the possibilities.

But framing it as a human-apocalypse scenario where they get to emerge as the triumphant gods who save us just made everyone even more suspicious and more likely to reject a bailout outcome

[The Human Playbook @gil2almeida]

@Deniz_Genlik The commercial incentives around ai companies are not aligned with human / social outcomes.

[The Human Playbook @gil2almeida]

People are acting on a complete false premise that this is just another cool tool that we need to “adapt” and “learn” faster. This is deeper than that. It’s substituting “thinking” (the appearance of it) and we need to “think” how are we going to require these skills so we can understand the freaking world we’re building with ai … why is this so hard to understand?