Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

Nvidia, Microsoft, Meta warn against overregulating open-weight models

cnbc.com

Post reply on HN
Print view View original

321–330 of 337 posts

Re: Nvidia, Microsoft, Meta warn against overregulating open-weight models

#321

Post by bigyabai »

Danox wrote: post #314
bigyabai wrote: post #86

Earlier quoted context omitted.

Apple is allergic to industry-spanning consortiums.

Apples allergy goes back to Motorola saying no, IBM saying no, Intel saying no, which led to buying Pa Semi, Intrinsity, and Anobit which in turn led to the birth of Apple Silicon with in house Soc’s. Waiting for the crowd means no vertical computing, no gorilla glass, multiple, ecosystems, Tandem OLED or thunderbolt five releases to a wide audience before most of the competition.

Apple is a trillion-dollar business, you can't blame their participation on Motorola. They can perfectly well hold down a consortium alone, we've seen it happen before.

Features like Corning Gorilla Glass, tandem OLED and "vertical computing" should have been cut in favor of a GPGPU redesign. Apple Silicon is now dead-last in the GPGPU architecture race, in the middle of a $4 trillion CUDA market. Nothing that you listed is a $4 trillion feature, Apple absolutely should have put aside their pride and helped Khronos kill CUDA.

Re: Nvidia, Microsoft, Meta warn against overregulating open-weight models

#322

Post by bigyabai »

trollbridge wrote: post #320
dartharva wrote: post #259

Earlier quoted context omitted.

Apple is not an enterprise tech company. It makes and sells luxury consumer devices.

Then why does enterprises have fleets of MBPs?

Because enterprise positions are white-collar luxury jobs.

Re: Nvidia, Microsoft, Meta warn against overregulating open-weight models

#323

Post by DrewADesign »

hangrybear666 wrote: post #238
DrewADesign wrote: post #179

Earlier quoted context omitted.

There are a lot of unknowns. It will probably be harder to get an acceptable quality outcome in most fields than in the text-based, limited vocabulary, comparatively highly structured, granularly documented, relatively unambiguous world of code. For the ones where it’s even possible with our current approach, the economics might not even be feasible to provide the service. Based on how much money these companies need…

Also most people building these things are engineers and coders, what they know best is code. So naturally this aspect develops at an elevated speed. Getting non-tech experts to build a cutting edge tech product does not sound intuitively feasible.

Maybe to some extent, but with the way LLMs work, code generation is about as ideal of a use case as you get. Especially given the available training data.

Re: Nvidia, Microsoft, Meta warn against overregulating open-weight models

#324

Post by DrewADesign »

monster_truck wrote: post #184
DrewADesign wrote: post #179

Earlier quoted context omitted.

There are a lot of unknowns. It will probably be harder to get an acceptable quality outcome in most fields than in the text-based, limited vocabulary, comparatively highly structured, granularly documented, relatively unambiguous world of code. For the ones where it’s even possible with our current approach, the economics might not even be feasible to provide the service. Based on how much money these companies need…

??? These are not unknowns. It is pretty well documented how much more an employee costs over their outright comp

You seemingly misunderstood what I deemed unknown. It definitely wasn’t that.

Re: Nvidia, Microsoft, Meta warn against overregulating open-weight models

#325

Post by novaleaf »

trollbridge wrote: post #317
novaleaf wrote: post #141

Some irony: I subscribe to Claude and Codex (20x plans), and now Kimi. Why Kimi? because K3 is the only frontier model I can have a serious conversation with about my product's security. (I did apply for OpenAi's Cyber Pilot but got no response)

I’m in the exact same boat, although Codex isn’t quite as bad. Fable smacked me for asking it to design a secure app without obvious security flaws and to double check it wasn’t using libraries with known security problems.

Agree, out of the two, I can get Codex to design and implement security systems (Fable just refuses to discuss). I've only used K3 a little bit, but I found that being able to discuss attack vectors and their mechanics gives me details to paste back into Codex for it to ingest. I could never have gotten there through Codex alone.

Re: Nvidia, Microsoft, Meta warn against overregulating open-weight models

#326

Post by applicative »

trollbridge wrote: post #318
applicative wrote: post #267

Everyone keeps assuming that China will forever be publishing these models, but it is clear as a bell that that day is over. In the end every model will be buried under state control for the simple reason that they will soon be really dangerous, as China understands better than anyone. They will give Kimi K3 to the Sinaloa cartel and the US military, but that's about the end of it. Perfecting these autonomous agentia…

I don’t think the Sinoloa cartel cares at all about frontier models, and the military already runs their own locally deployed editions of Mythos.

Of course the cartel crazies use frontier models! https://www.wired.com/story/how-mexicos-cjng-drug-cartel-emb...

Re: Nvidia, Microsoft, Meta warn against overregulating open-weight models

#327

Post by satvikpendem »

2cynykyl wrote: post #286
satvikpendem wrote: post #36

Earlier quoted context omitted.

Banned in the US, obviously, not worldwide. The problem is if the US wanted to, they actually could enforce such a ban nearly worldwide, by putting Chinese AI companies on a sanctions list and disallowing any US company (or anyone that does business with a US company, which is much of the world) from touching them.

This is the most insightful comment on this subject I've read yet. I hadn't put those pieces together.

It's been talked about before, it's what essentially happened to Huawei.

Re: Nvidia, Microsoft, Meta warn against overregulating open-weight models

#328

Post by trollbridge »

bigyabai wrote: post #322
trollbridge wrote: post #320

Earlier quoted context omitted.

Then why does enterprises have fleets of MBPs?

Because enterprise positions are white-collar luxury jobs.

MacBooks tend to be a better value than any other option. Except at the ultra low end.

Not really “luxury”.

Re: Nvidia, Microsoft, Meta warn against overregulating open-weight models

#329

Post by bigyabai »

trollbridge wrote: post #328
bigyabai wrote: post #322

Earlier quoted context omitted.

Because enterprise positions are white-collar luxury jobs.

MacBooks tend to be a better value than any other option. Except at the ultra low end. Not really “luxury”.

> Except at the ultra low end.

You just conceded the point. Only some businesses buy them, because they want to lure and retain expensive, white-collar talent.

No enterprise servers deploy to macOS. Nobody's edge compute or inference cluster is a Mac. Apple does not make hardware for those people, they are a luxury bauble company.

Re: Nvidia, Microsoft, Meta warn against overregulating open-weight models

#330

Post by sigmoid10 »

novaleaf wrote: post #302
sigmoid10 wrote: post #301

Earlier quoted context omitted.

I found GPT 5.3 was the last model that was sufficiently competent and still open to discuss security on my own github repos. 5.4 started refusing to even look at potential problems, albeit not consistently. I'm considering a Kimi subscription, but I know many employers will simply not be on board with this and I don't know if I get enough personal use out of this for the few things I run on servers. When those compa…

keep in mind that Kimi trains on your data, so if you care about that, sandbox what it has access to.

They only train on the subscription memberships. Like everyone else. They also have a metered API that is not used for training (like most others too).
Post reply on HN