Don't Be Fooled by this Summer of AI Hype | Hacker News Reader

Don't Be Fooled by this Summer of AI Hype

technologyreview.com

40 pointsjacquesm32 comments

If you're interested in an anti-AI article which isn't just the same trite takes you've already read a million times before, just move along. Go read a HN thread about any recent AI news and you'll get the same takes.

Lots of anti-AI takes seem to fall prey to the moralistic fallacy (ie because something shouldn’t be, thinking it isn’t). I’m very skeptical of the claims AI companies make, their motives, and whether their impact will be anything positive. But I am pretty tired of takes like this that are so reflexively skeptical of anything AI that they carelessly dismiss claims about the dangers and real impacts of AI that are happening.

(And to be clear: I’m not talking about being skeptical of fantastic claims about things that have not happened- ie an AI takeover, runaway super-intelligence, etc.)

What kind of real evidence about the dangers of AI that are happening do you have that didn't come from an AI company, wasn't about some internal test that conveniently can't be replicated or independently verified, where nobody disabled guardrails that would be in place for typical users, or where they didn't instruct the AI to roleplay as someone doing something evil? What danger are we seeing right now in the real world that people aren't taking seriously?

The article is very dismissive of recent mathematical proofs. It mixes together ethical problems (plagiarism, "strip mining" proofs) with capability problems (proofs not being novel, overstating capabilities). Again, it's fair to be skeptical of what the companies themselves are saying (and why). But from what I've read, many mathematicians are taking AI quite seriously. I've recently read several thoughtful essays from credible folks who are calling for dramatic changes to the field in response. For example: https://www.daniellitt.com/blog/2026/9/13/a-beginning-for-ma...

The article is also dismissive of the Hugging Face incident and other hacking incidents. This one is trickier, because we are more reliant on what the companies have released to piece together what happened. The METR report is probably the closest thing we have right now. Here again, I think the issue is that a lot of coverage blends the ethical issues (holding AI companies responsible for what their bots do) with capability issues (whether AI is or will be capable of developing novel and dangerous attacks).

That's not what happened, in any of the many incidents. But even if it were, if you can make AI do dangerous things by telling it to roleplay doing dangerous things, then that... Makes the AI dangerous.

> That's not what happened, in any of the many incidents.

That's exactly what happened when Google's AI "autonomously" hacked into companies coincidentally in the middle of tests of its cybersecurity abilities in a simulated highly controlled scenario on a totally isolated system that was just accidentally given unrestricted access to the open internet.

That's exactly what happened when AI was "threatening to blackmail" people. Internal tests, in roleplaying scenarios designed to push the model to do exactly wthat it did.

It's exactly what happened in the the Hugging Face incident. An internal test was conducted, AI was told to do tasks specifically to test AI's ability to take known vulnerabilities and turn them into working exploits, mostly using an internal research model which was not intended for production, but also another production model whose guardrails had been intentionally removed.

So yes, a company with billions and billions of dollars and years of research and investment into AI models likely will be able to use their AI in cyber attacks. I think pretty much everybody believed that was a risk.

What does that tell us about our risks in the real world right now? Well, it's extremely hard to say because we're still left trusting the AI company who is constantly lying about what their AI is to tell us what it was doing and how, and we're still unable to independently verify or replicate what happened, and there are no known examples of anyone who isn't an AI company doing anything like it. Certainty you or I or the vast majority of the population can't just browse to chatgpt.com and hack the planet.

Did AI do something that humans couldn't already do? No. Did they spontaneously "go rogue", or demonstrate "superintelligence" or "create AI civilizations"? No. Can we be sure they didn't just orchestrate the whole incident to push their "AI scary!" narrative? Nope. So we're pretty much left with what we already knew. All the largest risks are from AI companies themselves and governments building AI models. There isn't much any of us can do about that.

So, your claim is that OpenAI intentionally might have had its agents hack multiple companies, as well as multiple government agencies? And they're just pretending to not have known about it?

Okay.

The vast majority of the AI-doomerism (if one wants to call it like that) actually doesn't come from AI companies, rather, from a few (high-profile) individuals.

(I intentionally ignore the tsunami of trash journalism and the likes)

"Kill all humans" came from Anthropic.

AI companies were quick to sign the "Statement on AI Extinction Risk" https://aistatement.com/work/statement-on-ai-extinction-risk...

"White-collar work, where you’re sitting down at a computer, either being a lawyer or an accountant or a project manager or a marketing person, most of those tasks will be fully automated by an AI within the next 12 to 18 months" https://www.youtube.com/watch?v=YTrBz6Z5c0E

AI companies aren't leaving a lot of AI-doomerism ground uncovered.

Why should anyone listen to Timnit Gebru or Emily Blender after how consistently they've been wrong about everything? What insight or value do they bring to this?

You’re absolutely right! In 2019, Timnit and Emily published a paper criticizing LLMs because of water usage / environmental impact, deception, and propagation of bias. They were completely wrong, and thankfully none of those criticisms apply to LLMs today!

She actively lies on twitter

https://x.com/ByrneHobart/status/2105368471198380131

You're right to push back! Honestly, calling someone with EA-adjacent ideas an effective altruist is an active lie! On the other hand, when AI companies claim that they have solved Millennium Prize problems, when really they plagiarized other academics' work, that's just an honest mistake.

You have a keen and insightful understanding of honesty!

This is not a constructive way to engage with people. Please reply in good faith on HN - there are other platforms if you prefer otherwise.

Sigh. You're supposedly a PhD student. If you think LLMs are evil because of water usage, please, please do the least bit of due diligence and read a bit about it.

First, I never said LLMs were evil. Second, even though datacenters (and LLMs) consume relatively little water on an absolute scale, they can and often do overwhelm local water supplies, frequently in marginalized communities [1]. This is exactly what Timnit and Emily warned about [2]:

> When we perform risk/benefit analyses of language technology, we must keep in mind how the risks and benefits are distributed, because they do not accrue to the same people. On the one hand, it is well documented in the literature on environmental racism that the negative effects of climate change are reaching and impacting the world’s most marginalized communities first.

Please, please do the least bit of due diligence and understand the discourse before condescending to other people.

[1] https://cnr.ncsu.edu/news/2026/09/data-center-boom-water-res...

[2] https://dl.acm.org/doi/10.1145/3442188.3445922

Can you point to a single community or watershed where datacenter usage is even 10% of the agricultural usage in that same community? Or even just name a single community that's facing severe water shortages attributable to data centers?

What have they been wrong about? What is wrong in the article?

I believe the objection is because they are constantly negative, even before LLMs were a fraction of its usefulness today.

Every technology has its negative parts, but breathlessly pointing out the negatives gets really old.

Do you mean critical? They are expert in AI ethics, it’s pretty much their job to look for problems in the field. I’m not familiar with them but the posted article covers facts that are undeniable

This is such a bad article, which I suppose to be expected from Gebru/Bender. It is so strange to claim that AI capabilities are "hype" and useless, and simultaneously that we should censure these companies for the damage that rogue AIs have done.

We can't have both! The models are both insanely powerful, AND very dangerous! Anyone who's used a frontier model for an hour should understand this.

Things can absolutely be both useless and harmful. Case in point: Google "support."

If you want a glimpse of how techbros want to weaponize probabilistic automation against the population, you just need to look at how Google "support" operates. They'll fire human beings with actual responsibilities in favor of kafkaesque and unaccountable computer programs. Its sole purpose is making sure that complaints from pesky peasants go nowhere.

But maybe it isn't useless after all. It'll sure be "useful" for the oligarchs wanting to steal from everyone below them under the guise of "cutting costs."

Claude Anal is launching this fall, featuring an usage of quintillion liters of water per second and enough energy to pulverize the world, ten times over, kilowatts hour. It also managed to escape containment from our cutting edge UMBRELLA-like special facility by killing everyone and then impersonating one of our scientists to infiltrate human society. We've managed to apprehend this Claude Anal "bot" for now, but Claude Anal is so dangerous, we aren't sure what to do with it, if to release it at all. Anyways, it's coming out this fall for only $99/67 tokens. But it's totally dangerous so we might have to lobotomize it.

The ethical angles of AI are debatable, but how a company frames their own product as so superior that it went rogue and successfully attacked all the things should be critiqued and not believed until otherwise proven.

It's a race, and a speculative one, so flexing _constantly_ to stay seemingly ahead is expected.

> so superior that it went rogue and successfully attacked all the things should be critiqued and not believed until otherwise proven.

Your statement is very vague, but if you refer to the HuggingFace incident, a third party (METR) has been called to analyze it, and their report is public.

Regarding OpenAI and other hacks, I think the prompts and tools supplied to the agent should be made public or make it known to the investigators. It should be treated exactly the same way a normal hacker would be.

nice anti-ai article which made me disconnect with reailty but take the anti-ai pill for a trip

If you go there, you get a popup: "Click to claim AI offer".

Not me, but I'm using an adblocker and have JS disabled by default to avoid exactly that kind of thing. I probably miss out on a lot of ironic or inappropriate advertising like banner ads for guns on a news article about a school shooting, but that's a price I'm willing to pay.

One of the author has a pretty impressive Wikipedia page: https://en.wikipedia.org/wiki/Timnit_Gebru

“racing straight towards self-improving superintelligence and gambling with our lives.” sounds like a line from a bad TV show about AI written by i guess tech bros

Are they really paying people to be fired and deliver stupid lines to justify the data center overcommitment I want a new timeline the Mayan calendar was never meant to be stretched this long past $30/plate food trucks in Haight Ashbury

summer of ai hype, the fall of ai fear(reverse psycological antihype, hype) but the simple fact is that it is getting harder and harder to sell ANY of it, and the best thing people have found to do with all the technology, is to have sex in robotaxis, cheaper than a room,and it gets both of you together and then wherever you want to go after

Reply on news.ycombinator.com