Companies are collections of people, and these companies keep losing key developers to the others, I think this is why the clusters happen. OpenAI is now resorting to giving million dollar bonuses to every employee just to try to keep them long term.
If there was any indication of a hard takeoff being even slightly imminent, I really don't think key employees of the company where that was happening would be jumping ship. The amounts of money flying around are direct evidence of how desperate everybody involved is to be in the right place when (so they imagine) that takeoff happens.
I was a bit confused as to what "Cognition" was, but they're the makers of Devin (edit: that just got added to the title, for reference), so that makes sense. Just buying the competition, the only surprise is they had more money to spend than the big ones.
Well Google did also just pay $2.5B to license Windsurf in perpetuity. Cognition is probably spending a lot less than that for just whatever it left after that type of a deal. Remaining team members, etc.
Circumvents regulatory oversight and also shafts 99% of the employees. Seems to be a backdoor way to acquire the key founders/leaders and IP (via a perpetual license) while leaving behind a desiccated husk of rank and file employees, customers, and obligations.
No, as some portion of the $2.5b goes to Windsurf investors.
Basically, Google bought the top talent from the company. This cash was used (according to articles I read this morning) in part to pay directly out to shareholders, and in exchange Google got the top talent from the company and a license for the software (probably mostly so their new talent didn't have to worry about NDA, non-compete, and patent challenges).
Since this money went to shareholders, not to the company bank, and since top talent fleeing the company reduces the value of the company the overall value of Windsurf likely went down as part of the Google deal. This in turn likely made it cheap enough for the remainder to be purchased by Cognition.
They already do this, it's the moderation model.[1]
This name thing is an additional layer on top of that, maybe because training the model from zero per name (or fine tuning the system message to include an increasingly big list of names that it could leak) is not very practical.
Eric puts a ton of effort into AoC each year and has built a great reputation for it. They shouldn't have to worry about other events stealing/muddling that name.
Personally I didn't find the movie boring overall, but there were around five too many romance scenes between Driver and Emmanuel's characters that didn't seem to move anything forward, the kind of scenes that usually get cut for redundancy.
Old man wants to put his main actress in a sheer top, and have his director-analogue character fondle her. That's how I read those scenes. I felt like every single thing Nathalie Emmanuel did in that film was for the gratification of Coppola. He even put Nathalie in a lesbian relationship at the beginning!
I'm sorry if this is strawmanning you, but I feel you're basically saying it's in the public's interest to give more power to Intellectual Property law, which historically hasn't worked out so well for the public.
Copyright as far as I understand is focused on wholesale reproduction/distribution of works, rather than using material for generation of new works.
If something is available without contractual restriction it is available to all. Whether it's me reading a book, or a LLM reading a book, both could be considered the same.
Where the law might have something to say is around the output of said trained models, this might be interesting to see given the potential of small-scale outputs. i.e. If I output something to a small number of people, how does one detect/report that level of infringement. Does the `potential` of infringement start to matter.
Nah. What he is saying is that the existing law should be applied equally. As of now intellectual property as a right only works for you if you are a big corporation.
> you're basically saying it's in the public's interest to give more power to Intellectual Property law
Not necessarily. An alternative could be to say that all models trained on data which hasn't been explicitly licensed for AI-training should be made public.
I think the second alternative works too: either you sue these companies to the ground for copyright infringement at a scale never seen, OR you decriminalize copyright infringement.
The problem (as far as this specific discussion goes) is not that IP laws exist, but rather that they are only being applied in one direction.
HN generally hated (and rightly so, IMO) strict copyright IP protection laws. Then LLMs came along and broke everybody's brain and turned this place into hardline copyright extremists.
What do you mean by this? All I see in this thread is people who have absolutely no legal background who are 100% certain that copyright law works how they assume it does and are 100% wrong.
The difference is that before, intellectual property law was used by corporations to enrich themselves. Now intellectual property law could theoretically be used to combat an even bigger enemy: big tech stealing all possible jobs. It's just a matter of practicality, like all law is.
The main issue was that Meta quickly took down the first announcement, and the only remaining working submission was the information-sparse HuggingFace link. By the time the other links were back up, it was too late. Perfect opportunity for a rescue.
Hmm, it's not that simple, is it? Let's say the AI is trained on the tweet "Ben Adams drove to Mexico yesterday but I still haven't heard from him."
From this knowledge, you can ask the AI "Who has driven to Mexico" and it might know that Ben Adams did, and reply with that.
HOWEVER it's also baked into the model and can't be surgically removed after a complaint. That's the irreversibility part. You can't undo isolated training. You need to provide it a new data set and train it all over again. They won't do that because it's too costly.
The problem with the above example is of course that it can also contain sensitive or private user details.
I've easily extracted the complete song lyrics to the letter from GPT-4 even if OpenAI try to put up guardrails against it due to the copyright issues. AI is really still in the wild west phase...
The irreversibility is still important to highlight, as it is distinctively different from a similar consent issue with search: "Google indexed my website against my will, but I will just forbid them to include me in search results going forward".
It is irreversible similar to how a student reading a textbook from LibGen can remember and profit from that information forever. Kinda crazy how many in this community went from champions of freedom of knowledge to champions of megacorps owning and controlling of all of human creation in the span of like two years when it became clear other corporations could profit off that freedom too.
A bit of a shame about the exploit applying to THUG PRO. The mod is played to this day, since the more competitive side of the Tony Hawk franchise has been dead for almost twenty years (with the exception of the THPS1+2 remake, which was but a blip in the scene).
The mod itself is over 10 years old now, and I think the original developers are gone, explaining why no one was interested in fixing it when Ryan reported it. But this means that now the mod is unusable, no one is going to want to risk a full privilege exploit taking over their PC.
Hopefully this article reaches someone who's a bit more interested in patching the mod.
I wish I had the time, because it would be fun. Back when I DID have time, I actually got that thug1 source code almost playable on Windows. That source code was only for the console versions, and the code assumed if it was compiling for windows (and not Xbox windows..) it was only for tools, so a lot of pieces worked completely differently.
Two big issues with Archive.org are that 1. it's a single point of failure, they don't encourage mirror sites to emerge, and 2. they keep using the "brand" to fight unwinnable battles like hosting books they don't own online, risking the whole endeavor.
I still appreciate it, but just imagine if it goes down due to a lawsuit. Now that Google no longer shows cached results, an entire historical record would be gone.
Its surprising that archive.org is the only such outfit I have encountered. Just like we have had libraries since ancient times, why are there so few digital libraries? There must be others, but nowhere near the number (or awareness) that we should have.
Heck, existing paper-based libraries should probably each include a digital archiving department.
Maybe this is already happening or already exists, and is trivial to those studying library science or something. I can hope, anyway.
Local neighborhood libraries could have their own curated digital archive, as cache for fast local search, and archival backup for long-term resilience.
> I still appreciate it, but just imagine if it goes down due to a lawsuit. Now that Google no longer shows cached results, an entire historical record would be gone.
Or somebody accidentally `rm -rf`'s an empty variable. Or The Big One hits San Fran. Or somebody in crisis breaks in with a crowbar, matchbook, and jug of gasoline.
They're a rather old-school shop. Own their own servers, all in one location I think. Bare metal admin stuff, and data's only mirrored across two disks per file IIRC. Keeps costs down. It's what makes the whole operation possible. But I also wonder sometimes.
