Ask HN: What do you expect from the next 5 years of AI? | Hacker News Reader

Ask HN: What do you expect from the next 5 years of AI?

8 pointsmsuniverse202620 comments

One of my personal “pelican riding a bicycle” tests for an LLM was to prompt for the script of a Seinfeld episode with a contrived plot device. While hilarious, I doubt the rights holders to the Seinfeld franchise would approve of this use case.

In the next five years, I expect that LLMs will detect and prevent (or charge extra for) even “fair use” of registered copyright, the same way YouTube does today.

HNers are the last people I'd ask: they're the most cynical group about AI (and technology in general) there is.

You might as well ask horses in 1908 what they expect from the next 5 years of combustion engines

Are we reading the same site? Are you saying HN is not pro-AI enough for your tastes? Jesus.

Read the comments: it's people angry that AI is everywhere and pessimistic about its current and future use. Not in the "it'll become too powerful" sense but in the "it's now and always will be far less useful than people think" sense.

This is the least pro-AI tech community on the internet.

Some years ago when I was more willing to waste time, I used to argue with people on here who disabled javascript in their web browsers and complained that websites didn't work. That wasn't the dark ages... it was around covid era times. HNers are often proud luddites, or at least proud tech minimalists.

Read the rest of the comments. There are massive AI boosters on here - strong enough boosters that it seems legitimate to wonder whether they are shills. There are also massive skeptics.

It might be that OP is implying most of the boosters are bots themselves. I don't doubt it: AI equals money and ycombinator is a startup accelerator. I think this might just be a Brandolini's Law thing, as the "indifferent and/or against" people are just too goddamn tired of it all.

Speaking of horses... the internets tells me that the US horse population didn't peak until 1920 (at 25 million), so in 1908 they were still doing great, even though horses were beginning to be replaced by cars in the early 1900s. Millions were killed during WW1 though, which might have been avoided if ICE cars and trucks had been more advanced earlier.

The flood of LLM-generated content, code and communications might drown a lot of the things we love in noise. It will slip into every crevice of our existence, and people will adjust their behaviour accordingly.

Sure, it will make the bullshit part of our jobs more convenient, but for the most part it will just make the bullshitters a lot more prolific. I'm already feeling it.

I predict it will be regulated in the west, with approved models only to the public. Corporate and verified individuals only for the high-end level models. Most likely the end of open source competitive models. Not sure what the east will do.

Doubtful. When you can use an equivalent Chinese model, even a closed one, it’s hard to regulate local ones.

It’s basically the Napster era of AI at that point.

“You wouldn’t query a Chinese vector database, would you?”

Major steps forward come less often, as the low-hanging fruit gets picked and further improvement becomes harder.

Self-improvement turns out to be mostly a mirage, not a one-way ticket to the promised land.

Mind you, I'm only 60% convinced of these two predictions.

I think it will become deeply integrated into all spheres of life, and we won’t be able to imagine life without it (just as we can’t imagine life without the internet now). People get used to convenience.

I can't see governments not somehow getting more involved

I think large amounts of compute are big strategic asset for a country and governments will partner with the labs

It will tell lies more convincingly.

A big loss of money

More piracy by AI companies hiding behind the “training data” mantra.

Creators getting no compensation.

Bribing EU lawyers suffocate competitors.

Dumb Bots (“agents”) doing real world damage to companies and individuals

Division, violence, some possibly interesting theorems proved

Small, local models approach frontier-level prowess and reliability and so every person runs their own personal and private AI system. Frontier labs switch to serving large corporations and governments exclusively. An AI bifurcation, free AI for the masses and dangerously powerful AI for the psychopathic leaders of industry and global politics.

> An AI bifurcation, free AI for the masses and dangerously powerful AI for the psychopathic leaders of industry and global politics.

We got a small taste of that when Claude pulled Opus 4.6 from public use. You either paid a lot more to access better models or you were restricted to extremely limited access.

Robots / physical AI displacing workforce

Reply on news.ycombinator.com