I have, LLMs are less fragile, that’s why I like them. The ability to generalize isn’t just about being general purpose, it’s super robust, and so assuming the budget is there (I agree they are inefficient) end up performing better on many classical tasks that have ood inputs. Before LLMs / foundation models we all struggled with generalization and at least in the work I was doing people were independently converging to using bigger more general models for tasks anyway as compute got cheaper. LLMs are just the most popular version of this.
> The ability to generalize isn’t just about being general purpose, it’s super robust
I work with LLMs daily. 5 of my specialized tasks are outperformed by a custom model than a general purpose frontier model. The performance of my custom models not only beat them but are orders of magnitude low in costs and thus are able to be used by more customers.
I suspect you and GP are talking at different layers. I think you are using robust on specific tasks with measurable confusion matrix. I think GP is talking about robust in more complex and diverse workflows, with the ability to self correct over turns.
That's part of the irony here I guess. In specialized fields, think computer vision, there were lots of teams whose innovative state of the art model was essentially just a function of the limitless compute they could throw at the problem. Now there are just people with even bigger sticks.
There are lots of scenarios where specialized models still are the only option for real time, power efficiency, and so on. And transformers and other tech behind LLMs can equally produce better specialized models. But no sympathy for those who confused compute with innovation.
Yeah, I don’t understand why people have a problem with this. I think it should basically always be true between peer nations that there is free movement and trade.
Irony aside (now there's more people from the middle-east migrating to the UK- which if you're racist you tend to hate more than Polish/French/German/Swedes/Estonians)...
I think a lot of people who argued for brexit did so mostly out of ignorance.
"There are regulations on bananas, this makes things inefficient!" (forgetting that the regulations on bananas was about categorising them, not forbidding sale).
They felt out of touch and disconnected from EU parliament, and it seemed complicated and scary. Ironically the reason it's so complicated is because it's a lot more democratic than our own systems- but instead of attempting to understand it they threw their hands up and thought "since I didn't vote for every person who works there, it must be undemocratic!"
It's very true that there was an anti-migration sentiment inside brexit, but I wouldn't go nearly so far as to say that this was the only feeling people had.
I'd say it's mostly feeling a loss of control, which they couldn't place and our politicians were not open enough about explaining that this loss of control of the population was caused by.... the politicians themselves.
> I'd say it's mostly feeling a loss of control, which they couldn't place and our politicians were not open enough about explaining that this loss of control of the population was caused by.... the politicians themselves.
Yeah, the EU is essentially run by the national governments.
However, they typically (and the UK were particularly bad here) use the EU as a scape-goat for unpopular decisions even when they'd agreed to them in Brussels.
This is what I meant by weaponisation of outrage. The media got people so riled up about things that didn't really affect them so that they eventually felt like the EU didn't represent them.
Fishing being the obvious example. I'd never seen many people care about fishing rights until it was suddenly a hot topic leading up to the vote, but suddenly people really cared about what was happening around the scottish coastline
There are a lot of jingoists around who want to use that character trait to amass power/wealth, or to find an excuse for things in their life they don't like. Sometimes, their grievances might even be legitimate.
I just read the first part and if I understand he thinks we shouldn’t be allowed to train LLMs to act like they are conscious because then people will think they are and give them rights? Seems more an education problem than a problem needing rules about what persona you can fine tune in. People who want to will find ridiculous misinterpretations no matter what you do.
Openrouter lets you pick your provider and see their policy re retention and training. You can pick US providers e.g. digitalocean that don’t retain or train on your data. Openrouter is also now owned by stripe, so I don’t see any reason why using them with a trusted provider is riskier than using the big two (especially in light of some of the whole Navier Stokes thing)
If it’s really sensitive then don’t use a cloud provider.
I’m not super familiar with the options, but isn’t there O-1 for anyone who’s world class commanding a million dollar salary? The impression I have (mostly from online discussion) is that H-1B is basically only for (comparatively) low skill workers to come work for lower salaries. Are there cases where globally competitive people would come on a H-1B?
I came in on an H1B as an IC3 at Meta. Now, with the skills I gained during my career, I maintain core infrastructure that has a massive benefit to the industry and is critical to tech companies of all sizes (I learned last week that a major auto manufacturer uses my infrastructure as well). Immigration's benefit to society is tail-distributed, which is why it's important to have broad-based immigration (you can't predict who will be at the tails).
The public discourse was very different then. Misinformation, bias, etc. Those in power were much more concerned about an LLM saying something unapproved.
Wild that we accept that OpenAI didn't want to deal with the social/political fallout of people realizing that their tool was responsible for flooding the internet with misinformation and bigotry?
Sure, now that all seems quaint, but I can definitely see how a fledgling AI company would care about their reputation.
Look up how the wireless radio helped spread European fascism.
It's not about "the people can think this but can't think that", it's "how does this new communication platform change the existing political culture and how will we adapt?"
So I first read this comment and thought is looked like AI and thought I’d give the benefit of the doubt. Then I saw this where there is no doubt: https://news.ycombinator.com/item?id=49687774
I thought, I wonder if it’s the same guy? Sure enough.
If I take your two sentence comment and remove one sentence it also is non-sensical. Stop removing context.
>Your mileage went from VW->telematics vendor->broker->carfax, thats the chain it usually follows. The fuse stops it but the real...
Prior commentary advised removing the fuse to stop the telemetry. This comment points out that the chain of information flow precludes the owner of the car.
>the real issue is the owner is the only one without the record
I don't understand what "the record" is referring to here. The mileage is presumably still on the odometer of the car. So what is "the record" that the owner does not have and how is that causing the owner an issue?
You have to read it in context with what it is referring to, and perhaps to know a little bit about the issue. In that respect, it's like a lot of conversations around here.
I read this as:
"Removing the fuse that provides power to the telematics module can stop data collection from a vehicle, but the real issue is that the putative owner is the only one who isn't instantly privy to all of the collected data."
In any case, the complaint above was obviously about AI usage, and there are a lot of good reasons to use AI for editing your words, starting with (for example) if you're not a native speaker.
How much of the “danger” is from better models vs the harness?
Isn’t the current risk due to how AI is configured, like giving it a full set of tools and internet access and a goal to hack stuff?
If we think we need laws or gate keeping, why isn’t it at this level? I already can’t ddos someone or fuzz their server or whatever right, I imagine if I threw equivalent compute at old school hacking I’d just get arrested.
The quality of the “frontier” model doesn’t really matter, they just generate transcripts, they can take no action.
If this was real they’d be calling on people to stop hooking them in to “dangerous” harnesses as opposed to pausing research. But it’s not.
What in this proposal gives you the idea that they're only talking about pausing research itself? It doesn't say that at all.
But yeah, directionally I think you're right. It's insane we ever let these things connect to the Internet or to write, never mind execute code.
But for the economic counterargument, the distinction doesn't matter much. Continuing research but constraining harnesses is just accepting nearly all of the costs and foregoing nearly all of the benefits.
reply