Right, it seems like with celebrities no press is bad press (with a couple exceptions) and Twitter feuds and drama bring more interest. But doing it on behalf of a company can only look bad.
The obvious example is Musk who regularly makes himself look like an idiot, though I think the kind of whiny defensive posts mentioned in this article look way worse.
Somewhat surprisingly, I think Yan LeCunn does a decent job of posting fairly aggressive tweets but not looking like a whiner, crazy or someone shouting pointlessly into the void.
Ideally I agree. The issue seems to be that barriers to entry are too high. There is so much regulatory capture in the automotive industry, everything is arrayed in favor of incumbents.
I rented a skoda this year and I own a VW. They share the same "infotainment" system and it's shit. I otherwise like volkswagen but I can't recommend them based on their awful computer. I joke that it must have been designed by someone who had never been in a car and didn't know what it was for, and by that standard I guess it's decent.
Maybe not head to head, but you can definitely compete. Facebook has tried at least twice to materialize communities out of nothing (threads and metaverse), and they fail because it's not organic, they inherit all the bad stuff that comes with scale and none of the good stuff they organically leads to scale. We were talking about someone being imaginary groped in the metaverse and what kind of censorship it should have before anyone actually used it.
Twitter is the same, Musk bought it and drove it over a cliff.
Create a cool community and you can easily compete with those two.
It doesn't say what books. Shouldn't matter to some extent, but odds are it's a deliberate provocation as opposed to a book that the library is "organically" including in it's collection. "If your enemy has a choleric temper, antagonize him" works very well. But don't pretend it's not a tactic. I don't think this is the same as the "sensitivity readers" censoring Dr. Seuss or Ian Fleming or Agatha Christie or stuff that's been on long standing publication. Provoke, and then talk up your opponents bad reaction.
I would expect these are more likely to be used either for testing or for very niche applications, and would never be expected to be competitive for general purpose workloads. So I'm not too surprised by the cost.
I can imagine that the spec would to some extent reflect an attempt to capture the behavior of the jvm they actually wrote, as opposed to being created first and then having Sun implement afterwards? It sounds like an interesting exercise to recreate it based on the spec and see what got missed.
What is the upside for google? Their business is advertising and they have a massive platform for this already. What does a chatbot add? Showing that they are keeping up with evolving technology and that they also can train a competitive offering? From a pride perspective I see why they'd want to compete with OpenAI but from a business perspective? GPT+Bing has come and gone... I'm sure google worries about their business model being disrupted by LLMs but it's clear the mere existence of a chatbot isn't enough. So why rush to a competitive commerical offering?
Like it or not, they've become more efficient at maximizing shareholder returns. 1978 to present included all sorts of financial shenanigans designed to turn a company from the old school "return on assets" model to a financial vehicle that profits on the spread between it's revenue stream and financing costs. I think it's shit too, but the job turned from running a company go making money appear, and the pay went with it.
Don't hate the player, hate the game (seriously, the system is screwed up horribly)
This basically could have been written at any time.
Perhaps the most concerning response to the survey for anyone running a consultancy was that one in three consultants said they hoped to be doing something else in five years’ time.
Several regarded consulting as a stepping stone to a role in industries they advised. But one 25-year-old vice-president said it was “not a rewarding long-term career”. A 34-year-old engagement manager was more blunt: “This is a job for partner[s and] a bunch of juniors . . . [it] makes no sense to work here unless you are on a partner track.”
It's that sort of stuff that's a mainstay of the industry. Some of the reasons differ, economic cycles come and go, but this stuff is always the same.
Skimming it, I think they're not talking about academic deep learning research but about all the posturing around "AI" that various charlatans are up to.
Actual AI research often isn't very rigorous and is more "try a bunch of stuff and see what works" but it's not magic, it's just experimental as we don't really understand all the factors that affect model performance.
I tried that on slack for a while, it made no difference. I don't think most people read the status message. The medium lends itself to the "Hi" type messages unfortunately, there's not really a way go constrain human nature, other than to not use instant messaging at all (I also tried changing my status to a note telling people to phone me, that didn't work either)
I'd say we're currently choking to death on the self actualization that came from living in a safe society where all our needs are met. That remains more pressing than any externalities, which tend to get addressed as they become dangerous.
Not sure I understand all the subtlety but as a Canadian (where we have about 3 big cities and the rest has no voice) I'd love to be able to divert my tax away from the city I live in and to my home town. If course we don't really pay income tax directly to the city, so it would be complicated and probably just make more bureaucracy.
Training LLMs on data that you don't own the rights to is copyright infringement.
No it isn't. There is nothing indicating this is true and lots of evidence that it's either fair use or more likely not a use at all. Just pretending or asserting that it's a violation doesn't change it's legal status.
Before chatgpt people were putting stuff online and were fine with their lot. OpenAI found something else to do with that data, and it feels like people are upset they aren't getting a cut, even if their economics haven't changed.
I can see that potentially having info available on chat gpt would mean less page views for content that exists to show ads, but I'm not sympathetic to that business model.
People who have models other than trying to get views for ads are just as incentivized to create new content as they were before.
The flagged post didn't put it eloquently but the sentiment is right, and complaining about naming violates HN rules about not complaining about tangential issues. As it is they've at least temporarily ruined the discussion by having the top post be feigned concern about something that's not at all material to the content.
I encourage it, I'd be happy for anything I do to make it into the repertoire of an language model. What is the downside, I'm sharing it on the internet anyway. It's people trying to protect dying business models (ads, thankfully) or who are upset someone else found a use for their data and want to retrospectively rent seek that get worried about crawling.
It's talking about 70B and 160B models. Even heavily quantized can ggml run those that fast? (I'm guessing possibly). So maybe this is for people that dont have a high end computer? I have a decent linux laptop a couple years old and there's no way I could run those models that fast. I get a few tokens per second on a quantized 7B model.
What problem does this solve? It appears to add precision where it's mostly already clear - perhaps it can enforce some kind of rigor... but then like the example given uses "fair market value" as a term which I'd expect to be the kind of thing that's in contention, rather than any of the actual "logic", and it doesn't help with that.
The reason we have courts and lawyers is because of the need for interpretation beyond just writing good logic, so I don't see how this can really do anything. Or is it for something else?
Maybe I'm misunderstanding your question. Off the top of my head, there is codellama and starcoder, I'm sure I've seen various other ones. These are downloadable coding models that you can run on your compute. Did you mean something else?
That wouldn't let people make it about them instead of the project. Fundamentally that's what this stuff is, "my issue is most important" and trying to use popular language and concerns about things being "toxic" to pay attention to them. If they just did a fork, they'd actually have to develop software (or more likely split into groups that disagreed about some other political thing and fight about that)
The only time I've ever seen CoC invoked is as a kind of weapon to settle petty disagreements. Something someone disagrees with becomes "making me feel unsafe" or such nonsense. They're one of those ideas that seems harmless but in practice is only something agitators use to make the project about them (usually some woke ideology thing) instead of the project. I would never get involved in project that had a CoC.