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

US AI Action Plan

ai.gov

Post reply on HN
Print view View original

441–450 of 672 posts

Re: US AI Action Plan

#441

Post by fatata123 »

wredcoll wrote: post #78
AlanYx wrote: post #10

The most important thing here IMHO is the strong stance taken towards open source and open weight AI models. This stance puts the US government at odds with some other regulatory initiatives like the EU AI Act (which doesn't outlaw open weight models and does have some exemptions below 10²⁵ FLOPS, but still places a fairly daunting regulatory burden on decentralized open projects).

I don't know if this counts as amazing optimism or just straight up blinders if that's your takeaway compared to the emphasis placed on non-renewable energy and government enforced ideology.

[dead]

Re: US AI Action Plan

#442

Post by alganet »

jonplackett wrote: post #396
LorenDB wrote: post #60

> Encourage Open-Source and Open-Weight AI It's good to see this, especially since they acknowledge that open weights is not equal to open source.

How can this work with their main goal of assuring American superiority? If it’s open weights anyone else can use it too.

It doesn't say anything about open training corpus of data.

The USA supposedly have the most data in the world. Companies cannot (in theory) train on integrated sets of information. USA and China to some extent, can train on large amounts of information that is not public. USA in particular has been known for keeping a vast repository of metadata (data about data) about all sorts of things. This data is very refined and organized (PRISM, etc).

This allows training for purposes that might not be obvious when observing the open weights or the source of the inference engine.

It is a double-edged sword though. If anyone is able to identify such non-obvious training inserts and extract information about them or prove they were maliciously placed, it could backfire tremendously.

Re: US AI Action Plan

#443

Post by HPsquared »

rayval wrote: post #440
cardamomo wrote: post #173

Earlier quoted context omitted.

I wonder how this intersects with their interest in "unbiased" models. Scare quotes because their concept of unbiased is scary.

"unbiased", in the world of realpolitik, means "biased in a manner to further my agenda and not yours".

See also "fair".

Re: US AI Action Plan

#444

Post by davemp »

discordance wrote: post #239
davidmurdoch wrote: post #201

Earlier quoted context omitted.

How much land mass would need to be covered by solar panels to power this future AI infrastructure. Yes, I'm implying that solar would be impractical, but I'm also genuinely curious.

It's worth considering total lifecycle use of water (mining, production and operation) for nuclear and solar. Solar: ~300-800 L/MWh [0] Nuclear: ~3000 L/MWh [1] 0: https://iea-pvps.org/wp-content/uploads/2020/01/Water_Footpr... 1: https://www-pub.iaea.org/MTCD/Publications/PDF/P1569_web.pdf

That’s not really useful information. The nice thing about water is that it’s usually still water after it’s “used”.

The question how much is used for mining slurry or chemical baths.

Those 3000L/MWh might very well be more environmentally friendly than solar because most of it’s used for cooling.

Re: US AI Action Plan

#445

Post by HPsquared »

NewJazz wrote: post #299
bluefirebrand wrote: post #285

Earlier quoted context omitted.

Water we have plenty of. We can desalinate as much as we need to

Build subsurface wells and responsible brine dispersion infrastructure then come back and tell me we can desalinate as much as we need to.

If you put the brine back into the sea, and later put the waste water back into the same sea, doesn't it balance out? Also, the sea is pretty big.

Re: US AI Action Plan

#446

Post by HPsquared »

cbsmith wrote: post #433
bluefirebrand wrote: post #285

Earlier quoted context omitted.

Water we have plenty of. We can desalinate as much as we need to

Yeah, but you need energy to desalinate so...

How does it compare to ~3000 L/MWh? I assume it's a rounding error.

edit: Desalination uses 4 kWh per cubic metre of water. That is, it would yield 250,000 L/MWh.

Re: US AI Action Plan

#447

Post by preisschild »

nxm wrote: post #423
4gotunameagain wrote: post #421

Earlier quoted context omitted.

We're fine with not catching up to the likes of facebook, openai et al while maintaining very livable wages with actual work life balance & no need to toil ourselves away to have things as basic as health insurance.

Don't forget the sky-high taxes and energy costs as EU is quickly regulating itself into lower quality of life

Doesn't mean we should deregulate everything

Re: US AI Action Plan

#448

Post by davemp »

davidw wrote: post #75
Aaronstotle wrote: post #64

Earlier quoted context omitted.

The American Healthcare is so broken already that further breakage could be seen as an improvement.

A lot of people are about to find out with this admin, how things that were certainly imperfect can be so much worse. Building things is tough; tearing them down is relatively easy.

So much naivety. It’s like the new grad reading parts of a 500kloc project and proclaiming that it can only be saved by a full rewrite.

Re: US AI Action Plan

#449

Post by rs186 »

LorenDB wrote: post #60

> Encourage Open-Source and Open-Weight AI It's good to see this, especially since they acknowledge that open weights is not equal to open source.

Without providing actual support like money, the government saying they encourage open-* AI is no more meaningful than me saying the same thing.

In fact, if you open the PDF file and navigate to that section, the content is barely relevant at all.

Re: US AI Action Plan

#450

Post by giantg2 »

jordanb wrote: post #232
giantg2 wrote: post #82

Earlier quoted context omitted.

Aren't profit margins restricted by law? Ostensibly, any reduction in expenses should result in some reduction to patients.

The solution to this is to form a Health Group that combines insurance with a whole bunch of provider services. Then approve those claims that result in you paying yourself.

These are terrible. Then you have a very small in network option and get screwed when going anywhere else. This is basically what happens with CVS Caremark drug coverage - any recurring prescription must go through them or they won't cover it. They'll only cover one off prescriptions at competitors. It's really pretty terrible, especially if you need a compounded medication. I'm not sure how such an anticompetive racket is legally allowed to exist.
Post reply on HN