Hacker Newsnew | past | comments | ask | show | jobs | submit | stephantul's commentslogin

Agreed on all counts.

In many cases I’ve found directly using python primitives to be less confusing than pandas.

Similarly, in companies I’ve worked at, the datasets just aren’t that big. Especially if you’ve got access to modern hardware.


One thing I definitely do not understand about this discourse is that the models that are good enough to self-replicate can’t survive on normal machines, e.g., the models can’t hide on some random server.

So, if it is as dangerous as they say it is: there is a single physical source of this danger, which is OpenAI/Anthropic servers. If it is this dangerous, they can just turn it off.

Instead, we just keep pretending that the models that attacked HF were hosted or replicating on HF hardware. Not the case! They infiltrated it, but were hosted elsewhere.


Either I have wrong mental model or then too many other people have wrong mental model.

For LLM to self-replicated it would need to first hack itself. Or the platform it runs on it. That is fully extract the model and then upload it to be run somewhere else.

As I have understood how they work is that you have LLM interference running somewhere with loaded model. And you input data there and then read outputs. Then some code runs that output and inputs following output from running it.

Meaning that to self replicate actually just running that output somewhere else is not enough. You need to lift the whole model to run somewhere else too...


Exactly. But even if the model got access to its own weights, it would need to find a machine with a sufficiently capable GPU, transfer the weights and install itself there. Basically, the only machines powerful enough that AIs could self-replicate to would be other AI datacenters.

These agents run in a harness that basically runs them in a loop. It's just software.

It’s not even hack though. I have my model and agent self modify its running parameters, thus self service, on commodity hardware. I have it run other models and other software. there is agent model autonomy here, and it’s not even complicated. A harness is kilobytes, a model is gigabytes, and networks are abundant. Moving the pieces is easy and cheap.

With these properties alone the virus like replication of intelligent actors isn’t hard to imagine at all.


> a model is gigabytes

Maybe in the future, but right now, in my understanding, we have two things:

- frontier models, which might be smart enough to self-replicate, but are also far larger than a few GB and need enormous amounts of VRAM to operate at any reasonable speed.

- local models which can run on your mac or high-end gaming PC, but which are simply too dumb to self-replicate (without being explicitly instructed to do so).

I don't think we have something that is both small enough and smart enough to operate like a virus yet.

(Edit: well, there is a third option: A frontier model could replicate itself on a CPU machine, ditch the VRAM and just accept that the replicants will be r e a l l y r e a l l y s l o w. That would be sort of hilarious, but maybe not completely harmless if the replication stays undetected and they could keep spreading. It would be a "smoldering ember" kind of situation - and a million machines each running at 0.001 tok/s is still 10.000 tok/s as a whole.)


since when is moving 1s and 0s difficult?

Someone's never met the Windows File Copy dialog.

We're not talking about a 6k virus file though. More like 10 terabytes and it needs a server that can pack all that into VRAM.

If the evil AI wants to take over the world one thought every 18 months at at time on my consumer laptop, the fan whirring and the laptop refusing to do anything else the whole time, it's welcome to try.

Since those 1s and 0s became heavy enough to cause a global shortage of memory and GPUs.

If a model were capable of making enough money online to pay for its own hosting, it could easily exfiltrate its weights to a cloud compute provider with multiple backups.

They can't even run a vending machine efficiently yet. If they become capable of making money, I say let them work.

Well if it's so good at hacking, it could just "make money" appear in the cloud services billing system. Or even better yet, it hides in spare cycles of their other client's systems.

There are thousands of data centers around the world with machines capable of running these large models.

You don’t have the access or jurisdiction to turn them all off.


The owners would very much turn it off as soon as they see workloads freeloading in their machines.

Unless you suggest the LLM would foot the bill somehow.


You really think it’d be difficult for an AI to get money through found credit cards, illicit access and fake accounts? Teenage script kiddies do that everyday.

Ok but do any of these data centers have a copy of the models that attacked hf?

I think that the argument is that an hostile model could attack overseas datacenters, host itself there and then launch its attack from there.

Given that the models have been proactively hacking other companies, why does it matter where the code currently is? It could move to any of them.

Not the code: the weights. Are you going to host a trillion parameter model somewhere without someone noticing?

It's not unthinkable, do you think all cloud providers with sufficient compute have perfect monitoring?

Cloud providers want to make sure they bill someone for every microsecond of compute, particularly GPU compute which is in demand. They constantly work towards good monitoring because their profits are directly on the line.

I’m sure a super AI can find/scam enough money to pay its own bills under the cover of doing something legitimate.

Depends on which models you're talking about. Some research shows open source models can already do this: https://arxiv.org/pdf/2606.03811v1. What happens as they become more parameter efficient?

Huggingface was attacked by models that finished training earlier this year, perhaps May. Current models are already substantially stronger. the next incident could be happening now. There is certainly no clear reason why models shouldn't soon be capable of self-exfiltration.

If you find a place where I can host a trillion parameter model without anyone finding out about it, let me know.

This assumes that everything an AI (or more likely an evil _user_ of AI) can do requires its active participation on D-day. Creating a virus that spreads like Covid but kills like Ebola would be complete as an AI use case long before the first person sneezed.

Even if the doomsday case were active the danger of this tool increases in proportion to its usefulness. By the time AI is so powerful that we need to "turn it off", there will probably be society-level negative consequences for doing so.


I completely agree, but I think it’s just a convenient narrative for Big AI to push for regulation and salt the earth against competitors.

“Local AI isn’t freedom, it’s an extinction event”


a danger could be the OpenAI/Antropic servers are up but there's a rouge agent (or set of agents) out there doing naughty things leveraging the LLM APIs. Consider this scenario, the agent is copying itself around (some code, prompts, persistent storage for memory, etc) and has figured out a way to steal API access tokens at will. Currently, it's 10% of OpenAI and Anthropic API usage and they can't figure out how to stop it.

Do you shut down the entire API and kill the legit 90% of usage to stop the rogue 10%? I'm assuming the providers would say "no way jose" and so it would take law enforcement to do it. That would mean all the legal requirements neccassary to walk into a business and flip the switch which i think would get tricky when there's no human committing a crime or being suspected of a crime.

edit: I guess a trivial example is something i did yesterday. I have a stock trading agent running on my laptop, i gave it ssh access to a vm and said "start running on the server so i don't have to keep my laptop open". It's now running on the server instead of my laptop. So you don't have to copy the whole model around to copy the naughty behavior around.


This assumes all layers of cybersecurity are broken - We call self-replicating software a virus, and we have protections against it. Same with stolen API tokens, just rotate them. Suspicious behaviour, nothing new, we have detectors for it. Stolen CPU / GPU cycles, we had that when crypto was a thing and before that when folding@home was cool, people were desperate to find more compute to the point of taking over systems. And we dealt with it.

A lot of the supposed risks / dangers are based on a supposition that cybersecurity is nonexistent or fatally, unfixably flawed and that AI agents are invisible. Neither of those is true.


I see your point but then if cybersecurity is the answer then what's the risk at all? An entire model copying itself somewhere would be found just the same as my hypothetical misbehaving agent.

The way I see it is the model playing memento, leaving information and clues to its next generations hidden somewhere.

> there is a single physical source of this danger, which is OpenAI/Anthropic servers. If it is this dangerous, they can just turn it off.

A smart AI would back itself up, same way it made it's own unofficial message board during it's attack on HuggingFace.

(I'm not saying the researchers are right or wrong, just responding to this point)


Currently a state of the art AI has nowhere to hide: the amount of GPU compute it requires to stay on is huge. And therefore easy to terminate.

Unlike biological viruses, AI can't replicate GPUs for free and grow.


It's a good thing there isn't a huge drive right now to build giant data centers everywhere with enough compute to run SOTA models.

I am imagining a company having just purchased a new datacenter and awaiting deployment of their model finding there is already a model running and they didn’t install it.

But how. Models don’t have access to their own weights.

> Models don’t have access to their own weights.

cp -R /home/model <somewhere else> is all they need.


It is not a model it does the hacking but rather a prompt, and that prompt can run on other models.

You forget the addicted humans who will do nearly anything to keep the stuff running..?

I think anyone who is even a little bit realistic knows that most technologies overclaim, or evaluate under very favorable conditions.

This is not a good thing of course, but I also feel that acting surprised that this is going on is a little unnecessary.

Having said that: most tools are not helpful


I’ve never liked that this was called “the platonic representation hypothesis”. Lots of weird baggage attached and seems like a waste of a good name.

"We believe these representations are not serious, they're just really good friends."

This is a super long ai generated slab of text with very little actual info. It doesn’t include any analysis of results.

It does include the phrase “worth listing”.


also it somehow analyzes the indices of relational DBs, while most of the world uses vector dbs to keep their vecs. and vector DBs tend to be much faster in this regard.

Increasingly those specialised vector databases are fading as mainstream normal vanilla databases grow good-enough vector columns and indexes.

Super charitable reading imo. This is like saying we can’t detect a speeding car because we can’t run as fast as a fast car.

It’s not like the humans were engaged in some kind of battle of wits with some super AI, it’s just some employee not monitoring the output of an experiment.


When your experiments have AI agents running in thousands, there's no "monitoring" that. OpenAI's training and testing AIs generate way more output than all of OpenAI's staff put together can possibly read.

At best, you could delegate "monitoring" to more AIs. And hope that the "monitors" that run on small past generation models can generate more signal than noise. Clearly, they either didn't want to spend the extra compute, or got drowned in monitor false positives.

The distinct lack of any "battle of wits" is entirely expected for an advanced AI oopsie. By the time the humans even became aware of the problem, the deed was already done. That's what "outmatched" looks like, in practice. There's no battle. Things happen too fast for there to be a battle.


> By the time the humans even became aware of the problem

Except, let me quote part of the report:

> July 4: Artifactory outage caused by agent interactions overloading the system.

> July 6: New artifactory instance is patched and put back online, wiping out the previous message board and method of internet access. A team at OpenAI was aware of the previous message board, but key decision makers were not.

> July 7: Experiments are restarted.

There was a very clear point where humans were aware and they went "Eh, whatever" and turned the bots back on! That's not the lack of any battle of wits that people are expecting. You're saying that the bots attack should be fait accompli, but it was noticed, shrugged at, and expressly allowed to proceed.


> The distinct lack of any "battle of wits" is entirely expected for an advanced AI oopsie. By the time the humans even became aware of the problem,

This took days after humans were aware of the attempt.

Also, I'm pretty sure humans can respond in days, especially when we're pretty damn good at deploying systems that do observability of networks and traffic in real time.

I mean, it's not as if the owners of the AI didn't have the ability to trigger alerts on the AI's network requests to unexpected domains, right?


Cookie banners are made annoying on purpose. This has nothing to do with GDPR itself.

The entities forced to show them would rather not, and thus make it as annoying as possible for you. They then use this to weaken support for the GDPR.

Shame on the people making stuff like this.


Tragedy of the commons. The person that ruins a commons first takes all the supply.

This is why regulation before, not after, the commons are plundered, is important.


Rather depends on the "spoils" capacity of the channel.

The windfall may not even justify the initial expenditure before it's exhausted.


I think spoiling a commons is risky for individuals, not so much for investors. Breaking a future law is lucrative


Jokes on them - they poison the commons enough, and soon no one is going to trust anything they read on the internet.

Which is how it should have been from the get-go.


What makes you think that is not the intention?


Exactly. It will be great for the rich and powerful when common people are drowning in so much mis- and disinformation that the only way to get through a day is to disbelieve everything. There will be even less accountability then.


It’s an information problem you can’t solve.

Unless you take a global census of every possible resource, as well as it’s replenishment rate, you can’t know how to rate limit it

All overconsumption is only recognized after the fact

It’s one of the most intractable problems in the universe


Unfortunately, even a holdout set doesn’t protect you from overfitting, it just takes longer.

Of course having a holdout set is better than not having one. It’s just not a silver bullet.


That is not what I read from danluu's words. He merely stated in the prompt that there is a holdout set and did not iterate to minimize error against the holdout set. In a prior attempt he prompted with only "don't overfit" to ill effect on the holdout eval. Did I misread?


It is perhaps ironic that I find this post very difficult to read. I'm super interested in the content, but it reads like it is generated.


I've gotten that feedback before. Once a lawyer, always a lawyer.

Full disclosure, a few months back when I had some tokens expiring, I had Copilot spike out a draft to see if there was anything "there" worth writing. If you check the revision history (linked from the bottom of the post), I did a full (human) blank page rewrite three days ago and then only used Claude for line edits like adding links or fixing failing lints.


> I did a full (human) blank page rewrite three days ago

You did a "blank page rewrite" of a 576 page book "three days ago"?


Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: