Should the US Fund a 'National Cloud' for AI Research to Compete With China? (nbcnews.com) 48
Big data "has big designs on a big cloud," reports NBC News:
A steady drumbeat from some of the most influential executives in the technology industry has emerged in recent months to push the idea that the U.S. government should invest in a "national research cloud" — a hub for U.S. research into artificial intelligence where researchers from academia and smaller tech companies could share data sets and other resources.
It's an idea that has been backed by a government commission led by ex-Google CEO Eric Schmidt and including executives from Amazon, Microsoft and Oracle, which recommended that the Biden administration create a hub for U.S. research into artificial intelligence. The White House has warmed up to the idea, ordering another report on it due next year with an eye toward competing with China on the development of artificial intelligence. "We should be able to stay ahead of China. We estimated that we are one to two years ahead of China, broadly speaking, in this area. I hope that's true," Schmidt said in an interview with NBC News. "Investments that are targeted in research — new algorithms — should be able to keep us ahead," he said.
The stakes could be enormous. Some experts in artificial intelligence believe it has the potential to transform the economy — automating some jobs, while creating new ones — and the potential military applications have spurred investment by the Pentagon.
But this month, the idea began getting fresh pushback. Research groups including New York University's AI Now Institute and Data & Society, a nonprofit technology research group based in New York, say the very tech companies pushing this idea stand to profit from it, because the national hub would likely be housed in the same companies' commercial cloud computing services. They say that's a conflict, and little more than a cash grab by what's effectively the next generation of military contractors. The plan also could entrench the very same tech companies that President Joe Biden's antitrust enforcers are working to rein in, these critics say.
It's an idea that has been backed by a government commission led by ex-Google CEO Eric Schmidt and including executives from Amazon, Microsoft and Oracle, which recommended that the Biden administration create a hub for U.S. research into artificial intelligence. The White House has warmed up to the idea, ordering another report on it due next year with an eye toward competing with China on the development of artificial intelligence. "We should be able to stay ahead of China. We estimated that we are one to two years ahead of China, broadly speaking, in this area. I hope that's true," Schmidt said in an interview with NBC News. "Investments that are targeted in research — new algorithms — should be able to keep us ahead," he said.
The stakes could be enormous. Some experts in artificial intelligence believe it has the potential to transform the economy — automating some jobs, while creating new ones — and the potential military applications have spurred investment by the Pentagon.
But this month, the idea began getting fresh pushback. Research groups including New York University's AI Now Institute and Data & Society, a nonprofit technology research group based in New York, say the very tech companies pushing this idea stand to profit from it, because the national hub would likely be housed in the same companies' commercial cloud computing services. They say that's a conflict, and little more than a cash grab by what's effectively the next generation of military contractors. The plan also could entrench the very same tech companies that President Joe Biden's antitrust enforcers are working to rein in, these critics say.
...Wut? (Score:4, Funny)
Re: ...Wut? (Score:3)
The point is it's a gated community, not GitHub. If we share all our datasets, but China doesn't, we are at a disadvantage unless of course we are poisoning those datasets...
Government focused AI research needs to be controlled in a sense of it's role in 5the generational warfare. Sharing some datasets or building some kind of treaty which may involve the former, is effectively the only reason to be open.
A treaty involving proportional retaliation by NATO should be of equal priority but something like non-p
Re: (Score:2)
The point is it's a gated community, not GitHub.
If that's the point then it's antithetical to the belief system of the most valuable contributors and is worthless anyway.
Even worse, it is ironic misuse of abundance (Score:2)
As I wrote here: https://pdfernhout.net/recogni... [pdfernhout.net]
"... Likewise, even United States three-letter agencies like the NSA and the CIA, as well as their foreign counterparts, are becoming ironic institutions in many ways. Despite probably having more computing power per square foot than any other place in the world, they seem not to have thought much about the implications of all that computer power and organized information to transform the world into a place of abundance for all. Cheap computing makes possibl
Expensive and ineffectual (Score:5, Insightful)
The point is it's a gated community, not GitHub. If we share all our datasets, but China doesn't, we are at a disadvantage unless of course we are poisoning those datasets...
Given the context, consider the number of researchers who will have access to the proposed AI cloud.
Now consider the likelihood that China will get access to that cloud, given the number of researchers with access?
Note the number of contractors and 3rd parties that have access to US secret information (such as Chelsea Manning), and various dumps of CIA computer assets.
The proposed system will not work as intended, and will only serve as an expensive barrier to our own AI researchers.
Re: Expensive and ineffectual (Score:2)
More than fair critique. There is no rebuttal to your comment.
Old habits die hard. Cannot teach an old dog, new tricks. All we can do is say why the behavior exists, despite its shortcomings.
Re: (Score:2)
consider.
a i applications currently are being created and used by only the faceless ones whose only common trait is that they are billionaires.
but the one glaring omission are simple a i applications that address specific issues for the 300 plus million americans.
a i applications that would help a person decide which credit card to obtain.
or an a i application that would suggest which health care program works best for that person.
or an a i application that can navigate the u s tax code for an employee doin
Re: (Score:2)
Compete with China, not cooperate with them!
Re: (Score:2)
easy.
make money obsolete.
that would be a nobel prize winning event
Re: (Score:2)
The animals have lived on this planet for millions of years without money.
Man is the only animal stupid enough to not have figured this out. /s
Re: (Score:2)
I hear Hong Kong, Tibet and Xinjiang are all cooperating with China. They're doing fine, right? And now China wants Taiwan to cooperate with them. What could go wrong?
Re: (Score:1)
Re: (Score:3)
Re: (Score:1)
National... Public-funded... ain't happening (Score:1)
The minute someone suggests a taxpayer-funded project, the private entrenched players start flexing their lobbying muscle and making noise about "unfair competition". Just look at the number of municipal ISP proposals that got killed off before even reaching city hall.
The only way anything non-profit ever happens in the US is through the military, who takes public money and fans it out generously to private contractors anyway.
Re: (Score:2)
The minute someone suggests a taxpayer-funded project, the private entrenched players start flexing their lobbying muscle and making noise about "unfair competition".
RTFS. That is exactly the opposite of what is happening here.
The entrenched players are the advocates. They see the federal dollars flowing into their trough.
Re: (Score:1)
In America socialism is only for corporations.
Re:National... Public-funded... ain't happening (Score:4, Informative)
Just to put the military portion in perspective, for FY 19, the total was about $700 Billion. Of that,
$147,287 -- Procurement
$95,253 -- RDT&E (Research, Development, Test & Evaluation)
So approx. 36% of the military's budget "fans out" "generously" to private contractors. Undoubtedly, there are private contractors supplying bases, but these are not "generous" and they'd need to be paid regardless of who was providing the service. The rest of the money goes for what amount to just about everything a private company must pay: salaries, health care, retirement, etc.
In 2019, about $4.4 Trillion was the total U.S. budget, about $3.5 Trillion was paid for via taxes, etc. Discretionary budget was about $1.3 Trillion (Defense is part of Discretionary). Which leave about $3 Trilion. So roughly 70% of the budget is non-discretionary. Of that $3 Trillion, $1 Trillion was Social Security. About $800 Billion was Medicare. So about $1.2 Trillion was left in non-discretionary such as interest on the national debt (8% of the total budget, not non-discretionary).
Re: (Score:3)
$1.9T of FY 2019 tax receipts were income taxes.
$1.2T of FY 2019 tax receipts were FICA taxes.
FICA taxes cover *only* SS and Medicare, and the deficit there came from the SS trust funds, and as such did not contribute to the national deficit.
Though it will when the trust funds are expended.
In an indirect way, the trust funds do contribute to the deficit- but only because the $3T in assets that they have accrue interest, and since the rest of the
Re: (Score:2)
This is the entrenched players suggesting a taxpayer funded project to funnel more money into the entrenched players without clear benefit above what the players would normally provide anyway. It is not about setting up a government funded alternative.
Big companies looking for a big payout (Score:5, Insightful)
There is obviously no need for any "national cloud". This is nothing but big tech companies looking for new and better ways to tap into governmental largess.
where researchers from academia and smaller tech companies could share data sets and other resources.
Researchers already do this, and have been doing so for decades. While I'm not active in research at the moment, a year or two ago I did need to dig up a data set that I used in research some years ago. Finding it was zero problem. Standard data sets are all online. The bigger problem comes with private data sets: researchers who publish results, but refuse to share they data they used. Hmmm.
a way to provide cutting-edge computing power to academic researchers.
Computing power is dirt cheap, and computers (and graphics cards) are so powerful that there are very few projects that require "cutting-edge" power. If you're working on new algorithms, new architectures, or some other new ideas - you develop your ideas on a small scale, and maybe later try them out with more data. In the rare cases where you do need more computing power, there are plenty of solutions. For example, I used to rsh into all the computers into one of the computer labs, and run learning tasks when the computer lab was closed.
tl;dr: There is no need for this. Note that the idea is coming from the companies that would provide the service, not from any potential users of the product.
Re: (Score:2)
There may or may not be use for a 'national cloud', but to define that as 'throw more money at existing cloud providers for nothing new or expanded' is ridiculous.
If there were a purpose to that it would be to create a public alternative at lower cost. It may not be a feasible goal, but it at least would have some logical point to it.
Re: (Score:1)
In the US if you fuck over the government on a grant you're eligible for a larger grant next time.
In CCP's China, they bill your family for the bullet.
In the US poor people paying inflated prices for groceries are buying yachts for tech executives using the government apparatus to disintermediate the transfer.
Let's have government stick to defending the borders.
No (Score:4, Insightful)
What the US should do is let Oracle, Microsoft, Amazon, Apple and Google fight it out.
That way, we can ensure the lack of any cohesive overall AI plan that would threaten US citizens. (or I guess other countries as well)
In other words, we don't need to fund any of the companies mentioned above. They all have more than enough money already.
Re: (Score:2)
I believe that's called a cage match. I don't think I've ever seen Bill with a shiv.
Re: (Score:2)
Re: (Score:1)
What a surprise... (Score:3)
A bunch of hosting providers says US government should throw a bunch of money at their companies to do what they would have done anyway, and probably make a requirement that government and government funded programs to throw more money at their companies.
Countries should definitely invest in this. (Score:2)
So, a "free" marketing platform? (Score:2)
Those of us who have actually worked with "AI" know that it's a marketing buzzword that really means "black box analytics." In other words, vendors package their proprietary statistical algorithms and call it AI. These academics who want the US to fund their "AI" projects really want free marketing in order to launch their own startups.
Let's say we want to give this free money to people doing creative things. The government has a really poor track record of picking worthy candidates to receive their funding
Fight fire with fire (Score:1)
There are other things you could do (Score:2)
My first thought was, who in their right mind in private industry is going to share datasets? Training datasets are the most valuable piece of intellectual property in this new world and most companies in this space spend huge sums of money curating those datasets. The exact contents of those datasets and (most importantly) the processes used to curate them are tightly held secrets.
I think a better approach would be to fund the development (probably via the existing NSF process) of relatively generic trai
Re: (Score:2)
"who in their right mind in private industry is going to share datasets?": You may have put your finger on the main problem with AI for the DoD so far.
An article entitled "Want better AI for the DOD? Stop treating data like currency" (https://www.c4isrnet.com/opinion/2021/05/07/want-better-ai-for-the-dod-stop-treating-data-like-currency/) speaks to this. Here's a snippet:
Many data brokers in the defense community still treat data like currency.
No (Score:2)
> Should the US Fund a 'National Cloud' for AI Research to Compete With China?
You mean like the MMC, which was set up in a panic after the Japanese formed their 5th Generation project?
How did MMC turn out again? Oh yeah...
Not crazy, but a meh idea as described (Score:2)
There are particular hubs for things like bioinformatics datasets that are really useful. They're great and have benefitted their fields.
If I were to want to tackle this problem for AI, I'd start smaller scale - find particular subfields - vision ground-truth datasets perhaps, and make public repositories for that marked as public domain (wouldn't be a national thing, but science isn't so national anyhow).
If a bunch of these later merged into something AI-as-a-field wide, great. But rushing to that would li
GIGO (Score:2)
It would have to be AWS (Score:2)
Of course, if they do fund a project, it will be awarded to a contractor who will hire a sub contractor who will hire a sub sub in India who will deliver half a project. The contractor will ask for more money, milk it until congress gets involved, will fire the CEO with a HUGE golden parachute, then milk the government for more. The project will never be delivered but the company will hire the congress people who investigated them to use thei
How about we tell companies to stop helping China? (Score:2)
US? (Score:2)