Slashdot Mirror
The Impact of Artificial Intelligence on Innovation (nber.org)
Abstract of a paper [PDF] which was originally published last month: Artificial intelligence may greatly increase the efficiency of the existing economy. But it may have an even larger impact by serving as a new general-purpose "method of invention" that can reshape the nature of the innovation process and the organization of R&D. We distinguish between automation-oriented applications such as robotics and the potential for recent developments in "deep learning" to serve as a general-purpose method of invention, finding strong evidence of a "shift" in the importance of application-oriented learning research since 2009.
We suggest that this is likely to lead to a significant substitution away from more routinized labor-intensive research towards research that takes advantage of the interplay between passively generated large datasets and enhanced prediction algorithms. At the same time, the potential commercial rewards from mastering this mode of research are likely to usher in a period of racing, driven by powerful incentives for individual companies to acquire and control critical large datasets and application-specific algorithms. We suggest that policies which encourage transparency and sharing of core datasets across both public and private actors may be critical tools for stimulating research productivity and innovation-oriented competition going forward.
We suggest that this is likely to lead to a significant substitution away from more routinized labor-intensive research towards research that takes advantage of the interplay between passively generated large datasets and enhanced prediction algorithms. At the same time, the potential commercial rewards from mastering this mode of research are likely to usher in a period of racing, driven by powerful incentives for individual companies to acquire and control critical large datasets and application-specific algorithms. We suggest that policies which encourage transparency and sharing of core datasets across both public and private actors may be critical tools for stimulating research productivity and innovation-oriented competition going forward.
We don't have AI. All we have is automation. Stop please calling it AI.
No matter what you think, we won't end up with people working less and living better. We will end up with 1% exclusively benefiting from all the increases in productivity and 99% getting hit by even higher unemployment because even conspicuous consumption by 1% has a limit.
The problem with "AI" is that it isn't really intelligent. It relies on a library of existing data and existing solutions in order to give you an answer that already exists. It is almost the exact opposite of innovation.
Let's say you have a library of parts consisting of bits of wire, a lighbulb, a light switch, and a positive voltage source. The "AI" is free to use this toolkit to design its circuit however it likes. Let's consider how the AI would tackle creating a working lightbulb circuit:
1 - Randomly generate a circuit. This means taking a random number of each bit and randomly connecting it together.
2 - Run the circuit through a simulation (like SPICE maybe) to see what it does. Does it turn on a light? Great, we're done! But the circuit is randomly generated so it's likely that the circuit doesn't do anything at all. Oh well, better generate another circuit - go back to step 1, do not pass go, and do not collect $200!
It takes an enormous amount of computing power to get an answer, and the computer has to know something about the physics behind the operation in order to properly simulate the generated circuit. This is a problem if you're doing research about something that we know nothing about, like advanced physics.
Perhaps more importantly is the fact that a successful generated circuit won't look like a circuit that you and I would draw. It would likely have a bunch of extra junk sub-circuits that do absolutely nothing - dead pathways, extra wire segments all over the place, lightbulbs connected to nothing but lightbulbs, batteries connected in direct shorts, etc. Don't believe me? Fire up your google machine and look up "AI-Created Nude Portraits" - all the requisite parts are there to make it pornographic, but the assembled picture doesn't match reality. Maybe not a problem with generating lightbulb circuits, but imagine trying to build a CPU with a "AI" machine.
But IN ORDER FOR IT TO BE GOOD, we either need to collectively start ignoring IP laws within individual nations, or enough of us need to just ship from existing nations, while performing industrial espionage to create an independent nation or international organization that flaunts those laws and disseminates the information for the betterment of the public. Think the LoTeks from Johnny Mnemonic, whose storyline parallels these concerns, as well as the ones the Goldman Sachs analyst just reiterated, very succinctly.
The latest and greatest gizmo you bought in the morning will be obsolete by evening.
I'd like to live to see that era, it seems like it will be very interesting — in a good way.
In Soviet Washington the swamp drains you.
I'd like to picture you jumping up in the air matrix style, legs spread wide...pants just exploding off your body, before going in for the ultimate face rape of yours truly, cock all the way down my throat...violently gyrating your pelvis as you attempt to get enough friction for you to climax without losing your tight grip around my face. legs wrapped around my back, you ejaculate hot ropes of your seed down my throathole like a facehugger alien
Just because you can crunch #s doesn't mean you can invent the new Zippo lighter, Polaroid camera, or iPhone that doesn't already exist.
A lot of the innovation that occurs is what results from "fixing" what already exists but is not efficient enough in some way. It is innovation that no one except the designers themselves appreciate, because it is hidden innovation inside of products.
AI is going to be so amazing! It is really going to transform our lives. I'm sure it is right around the corner.
If true artificial intelligence existed, there'd be cause for concern. As it is, lazy humans would be the issue, not software.
is them making a discovery but not telling us.
Won't for a while ... hopefully!
Definition of Artificial Intelligence:
"the capability of a machine to imitate intelligent human behavior"
We have many instances of precisely that in many problem domains.
You are flatly wrong, and therefore the rest of us will not stop calling it "AI" no matter how much you implore.
Both of the "we suggest" paragraphs are coma inducing, so you get past that and what remains? This:
Strunkian synopsis: Imminent ruthless-genius land-grab race to the bottom.
Then the rest of the words could have been devoted to explaining precisely what they mean by "application-specific algorithms", because specificity roams wide.
Are we talking 120 or twelve thousand?
Redisribute more of my wealth & I will join the unemployed. Demonrat Party will only survive on other people's $, not mine.
Economists and many - Left or Right - seem to think that peoples' ONLY motivation is money.
I for one do not have to work but choose to. (I got my money the old fashioned way: I chose my parents well.)
But I work to have something to do. Get out and see people - real people. People of all walks of life.
And the learning experience!
Frankly, the middle/upper class Liberals/Conservatives don't have a fucking clue about the poor.
How do I?
I do their taxes and hear their stories.
Pretty much ALL of you got it wrong. And so do the poor; funny enough - but that's a whole book to describe.
The problem is the copyright and patent laws which force us to be slaves of corporations. The same types of laws make it illegal to raise capital unless you are rich. The solution is to realize that all these forces are run by communists and to assassinate them all.
go into biotech to replace the jobs NAFTA sent to Mexico. At least once a week I read how the sort of rank and file jobs that were supposed to have been created (the biotech equivalent to a code monkey) were replaced by some form of automation.
Oh, and the point of being in the 1% isn't conspicuous consumption, it's power. The good thing about being rich is the poor have to do what you say or they starve to death.
Hi! I make Firefox Plug-ins. Check 'em out @ https://addons.mozilla.org/en-US/firefox/addon/youtube-mp3-podcaster/
those are products that used innovations (metallurgy & oil refinery, chemistry and plastics, microcomputers, advanced radios and LCD screens).
Most of that tech was done by the government under the auspices of "military" spending.
Hi! I make Firefox Plug-ins. Check 'em out @ https://addons.mozilla.org/en-US/firefox/addon/youtube-mp3-podcaster/
It won't be any better than successive approximation, albeit fast, so might appear to be intelligent.
And yes it will make capitalism more efficient.
But that will regress our society
Go well
We suggest that this is likely to lead to a significant substitution away from more routinized labor-intensive research towards research that takes advantage of the interplay between passively generated large datasets and enhanced prediction algorithms.
It is cruel to have this (AI generated?) stuff just after another story telling "Papers today are longer than ever and full of jargon and symbols"
I consider it AI if it uses some form of non-trivial statistical analysis either in data preparation or during usage. Neural nets are just statistical approximators.
I imagine a scenario where distributed software exists, kind of a cross between seti@home and cryptocurrency mining, that performs processing for a deep-learning algorithm. The amount of total memory and processing power devoted to this could be as large as the larger cryptocurrencies, let's say. Now let's say that this algorithm is dedicated to creating new inventions. It's been said that a machine that's smarter than humans is the last thing we'd ever need invent, and this would be like that. Now, the algorithm and its results wouldn't be owned by any one entity, but it'd patent its ideas through a corporation it effectively controls (people would be needed legally, but the corporate charter could say that they have to do what the algorithm says). The fees for licensing its patents would then be distributed proportionately to those performing its processing.
If people dogpiled A.I. to the degree they're jumping on cryptocurrency, we'd have the singularity already :)
Corruption is convincing someone that the selfless ideal is the same as their selfish ideal.
"We suggest that this is likely to lead to a significant substitution away from more routinized labor-intensive research towards research that takes advantage of the interplay between passively generated large datasets and enhanced prediction algorithms"
What is a passively generated data set? You can have data without creating data. Passive data is an oxymoron.
What does enhanced prediction algorithms mean? Enhanced how? the thing reads like it came from a journal paper generator.
putting the 'B' in LGBTQ+
https://becominghuman.ai/forget-silicon-valley-innovation-is-happening-in-china-now-c6cfdbd74bc4
If AI is so bad, how come China is experiencing a boom due to AI?
>We suggest that policies which encourage transparency and sharing of core datasets across both public and private actors may be critical tools for stimulating research productivity and innovation-oriented competition going forward.
Good luck with that considering the compitition in academica is viscious as government funding is reduced.
general-purpose method of invention
I've already patented that.