This is regulatory capture in action. This will make it hard/impossible for new vendors to come into the market and only established companies will get to play, and charge, for LLMs. What does this mean for open source? Will it become illegal to download weights? What about train your own? Are we heading to a world where GPU use is regulated to ensure that illegal LLMs aren't being processed on your machine? More bro…
> I can train insainly powerful models on my laptop.
This is such hyperbole. You might be able to train a model that's merely useful in a single domain, but to say you can train an 'insanely powerful' model on consumer hardware is laughable.
As far as Europe is concerned they have recently signed up to the 'Pax Silica'[1] and willingly givrn the LLM space over to the US incumbents with buildtin legislation banning Chinese models and coperation with them. So EU will be a renter of the LLMs that the US allows them to use. In the long run OpenSource will dominate as it did in the DB(MySQL/Postgres)/ServerOS(Linux/BSDs) versus Proprietery rent seeking alts l…
I hope that open models will dominate. The difficult part to reconcile for me is the amount of compute that's required to create and run such models. Small models are fine, I run local llms 27b param on a gpu, but it's not even close to frontier in capability. Who wants to drop $40k+ on hardware to run these things. Companies, maybe/perhapts. On the other hand, to run a DB I can get a server for $3k and handle tons o…
Im not worried about this at all. The OpenAI, Anthropic and the US government can play this game all they want... They're just accelerating the development of open source models; and helping destroy the lead the US has built in AI, and their profit margins along with it. This is like the battle between PostgreSQL and Oracle all over. Move up market, isolate yourself to enterprises, and watch while everyone else build…
The government will just claim that unsanctioned models have the potential to deliberately introduce security vulnerabilities when working on IT projects (e.g. be trained to strongly yet covertly favoring introducing compromised dependencies when you are not looking). Then laws will be made to forbid organizations who use models other than those from the sanctioned labs to participate in critical projects on national…
And companies from outside the US will outcompete those from the US forcing more protectionism, higher prices in US etc. I think there have been several cycles of several industries that have gone through this cycle (cars, shipping?), and mostly forced to roll back.
Well I was expecting it but if it's not going to become available in this subscription cycle for me I will cancel oai subs as I did with claude ones months ago... moving to open weight models is trivial now, with optimizations and stuff glm 5.2 is roughly the same price as the best models around from multiple vendors. unless I could atleast try and see Sol perform like 10x better I don't really have a reason to switc…
> I used Fable for like what 2-3 days at most and didn't really feel it was so much better It was a lot better. I can't believe people say this.
It felt a lot better, but it was just a feeling. None of the stuff I had Fable do actually worked, but it looked great.
This is regulatory capture in action. This will make it hard/impossible for new vendors to come into the market and only established companies will get to play, and charge, for LLMs. What does this mean for open source? Will it become illegal to download weights? What about train your own? Are we heading to a world where GPU use is regulated to ensure that illegal LLMs aren't being processed on your machine? More bro…
Nick Bostrom wondered aloud in Superintelligence (2014) why states would allow individuals and private organizations to develop AGI. If one takes the possibility seriously, AGI would a source of immense power and any state would to take that opportunity for itself.
Edit: Not saying whether AGI is right around the corner, that's a different discussion. I'm just saying that a serious possibility of AGI and an understanding of possible consequences will make a state act.
As far as Europe is concerned they have recently signed up to the 'Pax Silica'[1] and willingly givrn the LLM space over to the US incumbents with buildtin legislation banning Chinese models and coperation with them. So EU will be a renter of the LLMs that the US allows them to use. In the long run OpenSource will dominate as it did in the DB(MySQL/Postgres)/ServerOS(Linux/BSDs) versus Proprietery rent seeking alts l…
I hope that open models will dominate. The difficult part to reconcile for me is the amount of compute that's required to create and run such models. Small models are fine, I run local llms 27b param on a gpu, but it's not even close to frontier in capability. Who wants to drop $40k+ on hardware to run these things. Companies, maybe/perhapts. On the other hand, to run a DB I can get a server for $3k and handle tons o…
I believe until the hardware designs catch up to be more commodized ala cryto mining evolution from GPUs to ASICS for specfic algos.
Designs (like Google TPUs equivalent) would also need to evolve to be more memory dense to be able to handle them.
Untill then it seems will be system time shares for the larger models , probably with a bring your own model and pay as you go.
I hope this doesn't become the new norm where government becomes the bottleneck for innovation in the AI space. It's worrying that with no formal and transparent policy framework that the government will be picking winners and losers and stifling innovation. There's been no public policy, executive order, legislation, or otherwise on this, I wonder if anyone has filed FOIA requests for these decisions or the conversa…
Well, when the leaders of this movement go around doom-trolling for years on end this is what happens. It turns out you need to be careful what you say if you're a highly visible public figure. Amazing! Honestly, with open source models I don't think this regulation means anything because there's no way they can really regulate what's coming out of china. I don't think this affects innovation in AI much at all (unles…
I hope that open models will dominate. The difficult part to reconcile for me is the amount of compute that's required to create and run such models. Small models are fine, I run local llms 27b param on a gpu, but it's not even close to frontier in capability. Who wants to drop $40k+ on hardware to run these things. Companies, maybe/perhapts. On the other hand, to run a DB I can get a server for $3k and handle tons o…
Has anyone tried to run a data center as a Co-Op?
Would actually be a good biz model for the Colo facilities that keep shutting down as everyone moves to the big cloud providers.Now if they can get their hands on enough GPUs and RAM.
This is regulatory capture in action. This will make it hard/impossible for new vendors to come into the market and only established companies will get to play, and charge, for LLMs. What does this mean for open source? Will it become illegal to download weights? What about train your own? Are we heading to a world where GPU use is regulated to ensure that illegal LLMs aren't being processed on your machine? More bro…
Nick Bostrom wondered aloud in Superintelligence (2014) why states would allow individuals and private organizations to develop AGI. If one takes the possibility seriously, AGI would a source of immense power and any state would to take that opportunity for itself. Edit: Not saying whether AGI is right around the corner, that's a different discussion. I'm just saying that a serious possibility of AGI and an understan…
Well states are made of individuals, and at least in the US we should be able to self determine via elections and public discourse..
If that's impossible in any meaningful way, then yes, doesn't matter which color jersey the government is wearing, it's authoritarian.
Ah, so the specter of Biden doing it was bad, but this administration putting into practice is great.
Is anyone in favor of this?
Amazon, NSA, apparently a few other 3 letter agencies around the world, but the lattermost probably did not expect access to their agencies would be limited but appreciate that they aren't going to be exposed as security frauds by just anyone