131–140 of 143 posts
Dario is an American patriot who wants the US to win. Don't see this happening.
What is the US "winning" by pissing off their allies?
Dario is an American patriot who wants the US to win. Don't see this happening.
Earlier quoted context omitted.
That’s assuming everything would be in time and no budget overflows which is never the case in public funded projects. I can give you one concrete example. Google for Stuttgart 21. It’s a project to build a main train station in Stuttgart, Germany. The project has been delaying for years. By the time it’s finished it would be 12.5 years overdue its original deadline and so far it has costed 14.5 Billion euros. One co…
train stations are boring and no military spends time thinking about train stations . ASI hacker nonsense, that gets their attn.
I think what's actually needed is two things: an EU training infrastructure that allows training of 10T+ models, and an EU inference infrastructure that is sufficient that it's possible to do RL on them. This effectively reduces the problem to a specialized supercomputing infrastructure problem which I think is relatively easy to solve. I think the chips are coming. I think Euclyd will be able to do the inference chi…
Curious how you can develop ai in Europe and still be compliant with eu ai regulations , dma, dsa and gdpr at the same time
|_: Company| false
Earlier quoted context omitted.
What is the US "winning" by pissing off their allies?
Ask yourself the same question, but replace it with France kicking out NATO or New Zealand kicking out the Americans. What did they gain by pissing off their allies, and what have been the honest long term effects for the country in question?
Have OpenAI or Anthropic ever had a model hacked/leaked? Is there any good reads on their cultures of preventing it from happening?
- confidential computing and trusted execution environments (TEEs) are the strongest primitive yet invented to prevent model weight exfiltration. there are other techniques. like bandwidth limiting of GPU workers and key sealing. but none as strong as those afforded by confidential computing and TEEs
- see https://confidential.ai/docs/confidential-computing-primer for a primer on confidential computing
- using confidential computing and TEEs to protect model weights in use is an area of active research
note: i work on confidential.ai
I think what's actually needed is two things: an EU training infrastructure that allows training of 10T+ models, and an EU inference infrastructure that is sufficient that it's possible to do RL on them. This effectively reduces the problem to a specialized supercomputing infrastructure problem which I think is relatively easy to solve. I think the chips are coming. I think Euclyd will be able to do the inference chi…
VSORA already finished their inference chip if you want to take a look at what they're doing. They got a handful produced at TSMC 5nm node. They are expected to be offered by Scaleway at some point in the near future (alongside with other French components from the AION consortium, like a SiPearl Rhea1 CPU and an Eviden Bull BXIv3 interconnect, all of them are also already produced, only integration is remaining). It…
Earlier quoted context omitted.
Deepmind I can undertand, as it was early days and no guarantee of success? and they needed massive investment. But the decision to allow Arm to be sold was surprising. But UK sells nearly everything - we need ongoing foreign investment.
> Deepmind I can undertand, as it was early days and no guarantee of success? Its companies with no guarantee of success that are the future innovators. Look at how important Deepmind's field has become. > But UK sells nearly everything - we need ongoing foreign investment. and then we complain about the resulting trade deficit! You cannot have both - it literally does not add up.
Most don't succeed, and DeepMind needed levels of money that would have taken the bulk of the Uk pot. Also no guarantees that a uk only based DeepMind would have succeeded in the same way as a Deepmind with multiple locations and able to use the Google Halo to attract staff.
Also the paper Attention is all you need was release by Google Brain researchers, not Deepmind.
UK backed Graphene financially stating it would take over the world, but it failed to succeed commercially.
Earlier quoted context omitted.
Actually I’m not so sure about that. The turning on and off quickly part. Because in my country (in Asia) there’s surplus energy from renewables (solar, wind) during the daytime, so much so that they’re having to shut things down because the grid can’t handle it (yet). However at night we need power from traditional sources (coal, basically. We do have hydroelectric and nuclear power also, but that’s not sufficient).…
Coal is different. Natural gas is commonly used for peaker plants that provide electricity only when needed. They’re a better fit for using along with other intermittent sources.