hgoel

To clarify a little, I don't doubt that a decent portion of this story is embellished to make it sound more impressive/shocking than it was.

Yet even if we dismiss the drama as marketing (say, the sandbox intentionally left holes, the zero days weren't actually zero days, even that huggingface was in on it and the model was instructed to break in to a system), we're left with a model that seemingly broke into another company's servers.

I wonder how many more high profile incidents some of you need before you stop insisting that this is all just marketing.

Is it going to take Chinese companies also talking about contributing to long standing math problems and accidental sandbox escapes? Or is that also going to be interpreted as some conspiracy?

They aren't talking about vision LLMs though. SAM and DINO are vision models, no LLM involved.

I think a big thing that the enterprise comparison misses is that this is petabytes worth of dense data that has to be put through fairly heavy processing (some of which is currently custom for the specific experiment) and studied by a human.

It isn't just a giant database of small files and metadata blindly feeding a recommender system.

Several petabytes is definitely a staggering amount of data in that context of being analyzed by human eyes to extract some scientific value.

The models in question have source and weights available. AFAIK not fully FOSS because the weights require registration to access.

This is a very tired insult/joke outsiders throw around about vtubing (though I don't get the impression that you had that intention). It's somewhat ridiculous, vtubers stream very often and voice changers tend to be pretty obvious. There are some men who use female avatars, but it isn't common.

It's also kinda misogynistic, plenty of women enjoy anime culture. Many prominent artists for vtuber models are also women (and occasionally vtubers themselves). To the level that calling the model artist "mama" and the model rigger "papa" is a common trope.

It isn't sustainable in the context of their desire to pivot to being a streaming service, along with their desire to blow resources on repeatedly redesigning the apps to make the UI slower and buggier.

Some VTubers treat it as a gateway into Japanese idol culture. So being part of an agency helps them with access to resources and opportunities for that sort of thing, e.g. voice acting gigs, appearances at events, planning concerts, recording studios, 3d motion tracking studios and brand sponsorships.

You might prefer things to be that way, but it's important to remember that's an absurd expectation. Especially in the west with how frequently vtubers graduate and show up under a different name. Even hololive JP, who push the entire persona thing very hard, have to periodically break immersion and remind fans that there are real people behind the avatar.

The "different persona" stuff is responsible for some of the most serious abuses the industry has surfaced, and the increasing treatment of VTubing as an avatar is a direct consequence of that.

Yes, of course, that would be interesting info to have. I just mean that from what we have already we can reasonably infer that the LLM played a role in the result being obtained.

If I am not mistaken, all of the flurry of novel results has come from existing mathematicians. This makes me suspect that the models aren't at the level where just any layman can get results. They require a skilled human in the loop to keep them on the rails and to properly explore the solution space.

Aren't deep learning models themselves a case where we have hints of some deeper underlying logic to why some things are more effective than others, but we lack the mathematical tools to properly work it out for anything of practical size?

All we're able to do is apply flawed analogies, generic information theoretical models, trial and error, post-hoc rationalizations and benchmarks without really understanding why.

Engineering and biomedicine, probably in the long term (if at all). But accelerated development of new mathematical methods has a possibility of proving to be relevant for fundamental physics research.

Occasionally large improvements in our models of the universe have been associated with the development of mathematical tools that allow those models to be expressed and/or tested.

The labs do tend to kind of anthropomorphise their models, as do us users, but admittedly part of the issue is that, for example, "thinking" is a decent layman description of the process even though it implies more human-like intelligence than there really is.

While I agree that we need the inputs to properly evaluate what this means for LLM capabilities, I don't really believe that the amount of knowledge input matters much for the overall significance of the result.

These kinds of results are interesting for LLMs because mathematicians have been working on them for decades. If the result doesn't already exist, there's no way it's in the training data, and if mathematicians have been unsuccessfully tackling the problem for decades, it is believable that the use of a new tool made the result possible, even if guided by a great mathematician.

I have played with CAD Sketcher, it's usable, but yes, not really good enough for serious use. Doesn't quite match up to FreeCAD after the significant improvements that has had over the past couple of years.

I really wish they'd go back to the blobs and stick to them.

I explained what I was thinking about when asking that here: https://news.ycombinator.com/item?id=48940887

I think what they're saying is that any morality related arguments during the free phase are irrelevant because their actions are not based in morality. Put differently, they aren't giving you something for free out of the goodness of their hearts.

Don't these estimates assume launching from the surface, fully via rocket? On Earth, having air breathing stages to gradually build up speed, or using other launch mechanisms, isn't worthwhile because rockets are more cost effective here, but those tradeoffs change if you're on a planet with higher gravity and a denser atmosphere.

And as a separate matter, any tool for evaluating students should be applied fairly, safely, and with adequate human review and due process.

Agreed, that's a fair and reasonable stance.

The reason I asked is that I have a hard time understanding the point of these tools. When it comes to education, it can be a matter of learning objectives. But outside that, what's the point?

The prediction from the tool is pointless for deciding on copyright or contract issues, and other text should be judged on its correctness or applicability to the task.

If all the tool is good for is "maybe this student cheated, but only an in-depth investigation would maybe prove it", it isn't a very useful tool, because it's more straightforward to just mandate that evidence is submitted regardless of what the tool says. On top of that, even the lack of evidence of manual work isn't good proof of using LLMs.

So, if the decision from Pangram determined, on every assignment, if you would be expelled from university for plagiarism, would that be acceptable to you regardless of how you actually did the work?

If you would not be okay with that, what level of consequence would be acceptable for the output from this tool?

I don't know if the Chinese text implies something different, but I think even in English it's pretty normal for people to claim they 'faked' their way through something without referring to fraud.

E.g. "I faked my way through the interview!" = "I did my best to respond to questions I did not feel fully prepared for, and managed to get through the interview"

Agreed. I think it's wild that such a position seems relatively common in the US (though at a surface level I understand that it's just a cultural difference).

The only situation in which I would consider charging my kids 'rent' is, if as adults, they were being irresponsible with their life, e.g. being a NEET and not helping out around the house. Even then I would hold the money in a separate account to gift back to them later.

I think it's parents' role to always be a source of almost unconditional comfort and security for their kids. Though I also think that this is unsustainable unless kids also do their best to maintain and respect that relationship.

Tools are neutral so we shouldn't do anything to reduce the possibility of someone consuming alcohol while or right before driving. Tools are neutral, so we shouldn't do anything to mitigate blatantly obvious risks, in fact we should actively engage in the risky behavior, just to show how neutral the tool is!

Grok was replying to public posts on X with the compromising deepfakes. Musk was actively joking about it right up until many countries blocked it, and several European countries, India, South Korea, Australia, Canada and Brazil all started investigations against X for violating local laws against producing intimate imagery without consent. Internet companies often enjoy a lot of leeway for cases where their safety measures are bypassed and they take reasonable actions to mitigate or respond to bypasses, that evaporates when they openly support the abuse.

That attitude is exactly how to prevent things from ever getting better.

Selectivity is just a numbers game, doesn't translate well in the way you're trying to present.

I'm surprised that UAF as an acronym is apparently unusual even among people familiar with use-after-free as a concept. I thought that was a pretty typical acronym in the context of software.

I like to use subagents a lot, but I find them to be most useful when explicitly specified. E.g. "assign these tasks to 2 Sonnet, 2 Opus and 1 Fable subagent". Helps keep allocation consumption under control.

Why would they interpret a system being used for the purpose it was designed for as terrorism? Using it for small purchases is one of the intended use cases, e.g. allowing street vendors to accept digital payments, incentivising them to have a bank account and reducing the risk of theft.

In America there's a very sharp geographical distinction between which people oppose the melting pot and which see it as a core part of the American experience.

People from the big immigrant cities like NYC, SF, LA are more likely to hold the latter position.