Ai is no where near the level of leaving behind those that aren't using it. Especially not at the individual consumer level like this.

Anecdotally, as an educator, I am already seeing a digital divide occurring, with regard to accessing AI. This is not even at a premium/pro subscription level, but simply at a 'who has access to a device at home or work' level, and who is keeping up with the emerging tech.

I speak to kids that use LLMs all the time to assist them with their school work, and others who simply have no knowledge that this tech exists.

I work with UK learners by the way.

What are some productive ways students are using LLMs for aiding learning? Obviously there is the “write this paper for me” but that’s not productive. Are students genuinely doing stuff like “2 + x = 4, help me understand how to solve for x?”

My son doesn't use it but I use to help him with his homework. For example, I can take a photograph of his math homework and get the LLM to mark the work, tell me what he got wrong, and make suggestions on how to correct it.

Absolutely. My son got a 6th grade AI “ban” lifted by showing how they could use it productively.

Basically they had to adapt a novel to a comic book form — by using AI to generate pencil drawings, they achieved the goal of the assignment (demonstrating understanding of the story) without having the computer just do their homework.

Huh the first prompt could have been "how would you adapt this novel to comic book form? Give me the breakdown of what pencil drawings to generate and why"

At the time, the tool available was Google Duet AI, which didn’t expose that capability.

The point is, AI is here, and it can be a net positive if schools can use it like a calculator vs a black market. It’s a private school with access to some alumni money for development work - they used this to justify investing in designing assignments that make AI a complement to learning.

I challenge what I read in textbooks and hear from lecturers by asking for contrary takes.

For example, I read a philosopher saying "truth is a relation between thought and reality". Asking ChatGPT to knock it revealed that statement is an expression of the "correspondence theory" of truth, but that there is also the "coherence theory" of truth that is different, and that there is a laundry list of other takes too.

I recently saw someone revise for a test by asking chatgpt to create practice questions for them on the topics they were revising. I know other people who use it to practice chatting in a foreign language they are trying to learn.

Paste the lecture notes in and talk to it

It has been bad for not having access to a device for at least 20 years. I can’t imagine anyone doing well in their studies with a search engine.

The anology I would use is extended phenotype evolution in digital space as Richard Dawkins would say. Just as crabs in oceans use shells to protect themselves.

Even if its not making you smarter, AI is definitely making you more productive. That essentially means you get to outproduce poorer people, if not out-intellectualize them

Don't you worry; the "rich West" will have plenty of disenfranchised people out of work because of this sort of thing.

Now, whether the labor provided by the AI will be as high-quality as that provided by a human when placed in an actual business environment will be up in the air. Probably not, but adoption will be pushed by the sunk cost fallacy.

Productivity improvements (such as automation) increase employment.

The decreased employment case is when your competitors get the productivity and you don't, because you go out of business.

I think the tech-elite would espouse "raising the ceiling" vs "raising the floor" models to prioritize progress. Each has it's own problems. The reality is that the dienfranchised don't really have a voice. The impact of not involving them with access is not well understood as much as the impact of prioritizing access to those who can afford it is.

We don't have a post-cold war era response akin to the kind of US led investment in a global pact to provide protection, security, and access to innovation founded in the United States. We really need to prioritize a model akin to the Bretton Woods Accord

I’m watching some of this happening first and second hand, and have seen a lot of evidence of companies spending a ton of money on these, spinning up departments, buying companies, pivoting their entire company’s strategy to AI, et c, and zero of its meaningfully replacing employees. It takes very skilled people to use LLMs well, and the companies trying to turn 5 positions into 2 aren’t paying enough to reliably get and keep two people who are good at it.

I’ve seen it be a minor productivity boost, and not much more.

> and the companies trying to turn 5 positions into 2 aren’t paying enough to reliably get and keep two people who are good at it.

it's turning 5 positions into 7: 5 people to do what currently needs to get done, 2 to try to replace those 5 with AI and failing for several years.

I mean, yes, that is in practice what I’m seeing so far. A lot of spending, and if they’re lucky productivity doesn’t drop. Best case I’ve seen so far is that it’s a useful tool that gives a small boost, but even for that a lot of folks are so bad at using them that it’s not helping.

The situation now is kinda like back when it was possible to be “good at Google” and lots of people, including in tech, weren’t. It’s possible to be good at LLMs, and not a lot of people are.

Yes. The people who can use these tools to dramatically increase their capabilities and output without a significant drop in quality were already great engineers for which there was more demand than supply. That isn't going to change soon.

Ditto for other use cases, like writer and editor. There are a ton of people doing that work whom I don’t think are ever going to figure out how to use LLMs well. Like, 90% of them. And LLMs are nowhere near making the rest so much better that they can make up for that.

They’re ok for Tom the Section Manager to hack together a department newsletter nobody reads, though, even if Tom is bad at using LLMs. They’re decent at things that don’t need to be any good because they didn’t need to exist in the first place, lol.

I disagree. By far, most of the code is created by perpetually replaced fresh juniors churning out garbage. Similarly, most of the writing is low-quality marketing copy churned out by low-paid people who may or may not have "marketing" in their job title.

Nah, if the last 10-20 years demonstrated something, it's that nothing needs to be any good, because a shitty simulacrum achieves almost the same effect but costs much less time and money to produce.

(Ironically, SOTA LLMs are already way better at writing than typical person writing stuff for money.)

> (Ironically, SOTA LLMs are already way better at writing than typical person writing stuff for money.)

I’m aware of multiple companies that would love to know about these, because they’re currently flailing around trying to replace writers with editors + LLMs and it’s not going great. The closest to success are the ones that are only aiming to turn out stuff one step better than outright book-spam, and even they aren’t quite where they want to be, hardly a productivity bump at all from the LLM use and increased demand on their few talented humans.

Qwen has an open reasoning model. If they keep up, and don’t get banned in the west “because security”, it’ll be fun to watch the LLM wars.

> and don’t get banned in the west “because security”,

It's from Alibaba, which is Chinese, so it seems likely.

Yeah, but it’s a bit trickier with them, given how they still operate in US and listed in NYSE. Also if they keep releasing open source code, people will still just use it… basically the Meta way of adoption into their AI ecosystem.

If it's an open model, good luck preventing us from downloading and using it though.

tbh a lot of the rest of the world already has the ability to get tasks they don't want to do done for <$200 per month in the form of low wage humans. Some of their middle classes might be scratching their heads wondering why we've delegating creativity and communication to allow more time to do laundry rather than delegating laundry to allow more time for creativity and communication...

If $200 a month is the price, most of the West will be left behind also. If that happens we will have much bigger problems of a revolution sort on our hands.

That supposes gen AI meaningfully increases productivity. Perhaps this is one way we find out.

If the models are open, the rest of the world will run them locally.

If the models are closed, the West will become a digital serfdom to anointed AI corporations, which will be able to gouge prices, inject ads, and influence politics with ease.

Kai-Fu Lee's AI Superpowers is more relevant than ever.

The rich west will be in the lead for awhile and then get tiktok-ed.

The lead is just not really worth that much in the long run.

There is probably an advantage gained at some point in all this of being a developing country too that doesn't need to bother automating all these middle management and bullshit jobs they don't have.

No US company got TikTok’d, and China doesn’t even allow US social media companies in its country.

China is notoriously middle management heavy, by definition that’s what communism is.

Richer people always get products first, when they are still expensive and bad. Don't worry about too much.

I actually suspect the opposite. If you get access to or steal a large LLM you can potentially massively leverage the talent pool you have as a small country.

Has it really made that much of a difference in the first place? I have a feeling that we'll look back in 10 years and not even notice the "AI revolution" on any charts of productivity, creating a productivity paradox 3.0.

I can imagine the headlines now: "AI promised unlimited productivity, 10 years later, we're still waiting for the rapture"

No one is left behind, eventually. You think the ai companies don't want poor people's money?