andy99

I’m on the fence reading the other comments. On one hand I hate stock AI art, it looks like crap and signals that someone didn’t care at all and has zero taste. On the other hand, most peoples Halloween decorations are dollar-store made in China garbage, it’s not like we’re retreating from a world of lovingly hand-crafted decorations towards AI hell. I think a lot of it is really the anger it inspires due to the connotations (for me primarily of low effort and ignorance) rather than anything objective. I still hate it.

I don’t really support pushing out people who don’t vibe code, but I do think buy-outs are a good way to get rid of people who don’t really want to be at an org but hang around for the wrong reasons. I’ve seen this personally at much smaller companies where for whatever reason you accumulate a bunch of checked out people - giving them something fair but not so generous that actually engaged people take it makes it better for everyone, especially if the alternative is eventually doing layoffs.

Would love to see public evidence of this. I believe it’s possible but I’ve never heard this before

There’s been a tendency since well before LLMs for AI demos to focus on “here’s a new problem we made up that it solves that you could imagine will soon transfer to an existing real problem” as opposed to “AI solves X”. This has been an ongoing challenge, it does cool stuff and looks so promising but has been very hard to harness for useful work. It’s still going to take time before the hype dies down and we see what’s left.

I don’t think it’s going away, LLMs and other ML stuff will no doubt remain part of people’s workflow, but I also see no evidence for superabundance or anything close to it.

The people that are doing this are self absorbed assholes and confronting them will almost always lead to an escalation that makes it worse for you. Management is unfortunately the better approach here.

In the US anyway it feels like the battle for public spaces has been lost. Everywhere I go people are listening full blast to their phones with no headphones - I find this about 100x more obnoxious than filming. I think both should be banned with Singapore-style punishments but I expect even if companies posture about this nothing will happen. United famously prohibited using your audio without headphones but based on my trip with them last week that has done nothing. At this point it’s a societal problem that needs a tougher approach.

There’s probably some double fallacy here, like “x is valuable in this specific situation so it must be valuable in all situations” - I think that’s actually a named fallacy but anyway.

I’m sure it’s similarly true for travel agents, real estate agents, probably lawyers, etc. Just because there are circumstances when someone values them doesn’t automatically justify them everywhere or refute that many of them basically don’t do anything.

Stupid question probably because I’ve gotten out of sync with the new releases but is Flash better or worse than vanilla 5.3? I’ve been using 5.3 extensively (and maybe ironically have used flash only for viewing images which I have as a skill in Pi because 5.3 vanilla seemingly doesn’t support vision). I find it generally competent but still very far behind opus 5.5. I’ve been trying to avoid Anthropic but the lift of opus vs GLM is enough that I end up coming back to it. I feel like currently there’s no comparison.

anyway, tldr I would not be happy if stuck with vanilla 5.3 so is flash better and should I move to that?

I get this is an unpopular opinion, but it’s even more important given the nature of the allegations. They “look bad” but it’s all accounting stuff in Cyprus, none of us know what’s actually legal here and what isn’t, it’s the BBC making insinuation about stuff to promote an investigative reports show or something they run.

I don’t want to defend anyone involved in the allegations, and don’t personally think they’re above shady business dealings, I just don’t like usual sensationalist “reporting” - essentially advertising like this and how people eat it up.

I feel like this isn’t the best way to draw attention to the problem. Presumably having the police visit is what the YouTuber wanted? I’m not really sure of the lesson here but drama’s presumably the point as opposed to a side effect so it’s hard to take seriously.

I hate that too, but I think it’s more the ux than the concept of a default. There are definitely situations (“here’s the default install path, press enter to confirm”) where defaults are a good experience, and LLM generated ones can fall in this category.

What I hate about sentence completion or suggestion is usually that it happens right when I’m trying to think and so destroys my focus, it’s actually way worse than just a passive option, it’s actively harmful to the task I want. The worst is google docs “help me write” - that may be gone now, I’ve blocked it with ublock origin, that waits until you’re thinking amount what you’d write and then hits you with a distracting pop up. It’s obviously PMs that don’t care about their users and want to maximize some AI use metric.

Anyway rant aside, it’s the interface more than the concept that’s the big problem.

Softmax over logits doesn’t give calibrated probabilities either as a rule. I don’t want to comment specifically on Jev but as a rule it’s very hard to get good calibration because it’s somewhat in tension with minimizing training loss for neural networks, e.g. Guo et al (2017) https://arxiv.org/abs/1706.04599

While I know there are ways to improve calibration, I’d personally want to see a lot of evidence the probabilities were actually more meaningful before trusting them. I agree of course that asking an LLM to provide a confidence estimate is meaningless.

So nothing to do with AI, their anonymization scheme seems to be flawed and somebody figured that out. But AI is a better scapegoat than whoever designed the system

It’s an incentive problem. If “sovereign” becomes your claimed value proposition, you can claim success even if the models not competitive, so nobody is pushed sufficiently hard to actually make it good.

Sovereign works when talking about building a commodity supply or something, not in literally the world’s most competitive and fast moving field.

Those seeking sovereign capability would be better off aiming to be best at something, even something much narrower than an all round LLM. Or just fast following and making something that matches leading performance, which is close to what the Chinese labs do currently.

Not really. The upside of competitive newer models is all in the proprietary data they are trained on. This is why data labeling, RL environments et al have been such a big industry, OpenAI and Antrhopic are paying literally billions to get the data they need. Do people think the ability to do research level math or advanced cybersec comes from just training on more public data?

A real sovereign effort could invest heavily in this, whatever people accuse China of “stealing” I’m sure they are also generating tons of their own data and are probably the primary sovereign doing so outside the US labs.

Yeah I think a model has to be actually good to claim sovereignty, as in competitive enough that people want to use it. Chinese and US LLMs are the only ones in these categories right now. Mistral and Cohere have the same problem, yes they are made in different countries but they are not competitive. They (France and Canada in this case) would be better off just downloading Chinese LLMs, even if they get cut off they still have the weights.

I’m most familiar with Canada, where sovereign is usually just an excuse to overpay someone connected for an inferior product with no strategic value.

> they are literally trying to build a superpower that exceeds the impact of the atomic bomb

I think people (the government, powers that be) have given away what they really think by their actions. There are few that take this seriously. The “exceeds the impact of the internet” view has much more support based on investment, but the “danger” aspect does not based on actions. If this was a company making novel prions or whatever that were outside regulation but obviously dangerous, government would be all over them. What regulation we do see is much more power grab than mitigating real danger

I’d go as far to say that no good writer uses AI for editing. It’s a tool for people that don’t know how to write, don’t know what good looks like, and are not confident enough to choose raw over soulless.

Canada has a charter of rights and freedoms, but there’s a section that allows governments to override it at will so it’s effectively worthless. And nobody knows that better than Quebec.

This I believe is part of why forgery is dealt with so harshly, copying a few thousand bucks isn’t a big deal but destroying the integrity of hard currency is.

Agree unfortunately. I hate apple, for the same reasons as everyone else, but I have both a MacBook and iPhone and can’t do much about it. It’s like politics a little, you can only choose the least bad, there is no good.

I have a Linux server, and use cloud Linux machines extensively, but at least the last time I checked, Linux desktop sucks. There are always memory management issues, doing a video call makes my fan sound like a jet taking off, not sleeping, put my laptop in my bag and it gets to hot to touch - that’s just off the top of my head, I’ve tried Linux laptops on multiple occasions and they suck.

It’s often struck me how many “leaders” basically jump to step 6, and come up with an urgent set of superficial actions without even clarifying what they’re trying to do. There’s probably some value in “bias for action” as you might call it. But it’s been one of the difficult adjustments for me, how little people want to think.

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.

> If it wasn't a problem, the government wouldn't have to do it.

This is certainly very different from my experience with what governments think they have to do.

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.

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?

The public discourse was very different then. Misinformation, bias, etc. Those in power were much more concerned about an LLM saying something unapproved.

See also the 2023 call to “ pause for at least 6 months the training of AI systems more powerful than GPT-4.” https://futureoflife.org/open-letter/pause-giant-ai-experime...

Signed by Benjio, Musk, Woz, etc