Gary Marcus proposes generative AI boycott to push for regulation, tame Silicon Valley Gary Marcus, professor emeritus at New York University and serial entrepreneur, is the author of several books, the latest of which takes the tech industry to task for irresponsibly developing generative AI and putting democracy at risk. The Register recently corresponded with Marcus to learn more about his work and his …
Don't you know that Belgium
embodies [a concept so] revolting that the publication or broadcast of the word is utterly forbidden in all parts of the galaxy except one - where they don't know what it means1.Alright... I'll get my towel.
_________________
1 Adams, D., The Hitch-Hiker's Guide to the Galaxy (Radio Series), Fit the Tenth, Scene 4.
Democracy, in all of its many perverted and corrupted forms, needs to be shown the error of its ways by other than humans who clearly so evidently fail to be up to the task of ensuring it isn't misused and abused to the greater advantage and excessive exclusive benefit of a self-serving delusional and sociopathic chosen few ....... because (a) they have lost the plot, and gone a very long way from "Don't be evil," and (b) they are becoming increasingly powerful, with almost no constraint.
AI, both Generative and Degenerative, can fix that, Gary Marcus ..... of that you can be assured ..... so rest easy, take a chill pill or two, there's nothing to be worried about in that great nothing burger you be peddling ..... that is until such times as any proposed future human actions may give you great cause to be gravely concerned at the possible consequences and likely repercussions.
quote: (a) they have lost the plot, and gone a very long way from "Don't be evil," and (b) they are becoming increasingly powerful, with almost no constraint.
An accurate description of every nation state regime out there.
quote: clear, persistent evidence of harm.
That has been claimed of every new concept or technology from the printing press and Galileo to Darwin, electricity and Elvis gyrating. It just takes a while for people to get used to it. It is not a reason to ban it.
It is lazy and crazy to compare smoking with technology.
quote: The threat to democracy from automatically generated misinformation and deepfakes.
There are far greater threats to democracy from extreme and populist politicians. Nothing Russia or China can cook up compares to the Democrat/Republican toxicity in the US and nothing Russia or China could do would damage a G7 country as much as Brexit did the UK. Stop diverting attention from the real dangers to 'anything foreign'. It just fuels xenophobia.
quote: Should free, global, instantaneous content distribution have stronger safeguards that make it easier to hold bad actors accountable.
ie. Shouldn't nation state regimes be allowed to operate blanket censorship over the internet? No. Because there is no halfway house. That level of control is Orwellian, straight from the wishlist of every dictatorship.
quote: limiting the reach of individual accounts to no more than X number of people?
Only the government can reach everyone, whilst the rest of us remain isolated in their own tribes, or our own towns? Again, on the wishlist of every dictator.
quote: In 2024 I see a lot of disillusionment.
Much more coming in 2025. AI will stutter, the bubble will deflate, it will find niches, and everyone will just carry on. It might not vanish entirely like the metaverse, but the more thoughtful will look back on the moral panic and feel embarrassed by it all. Perhaps moral panics are the new prolefeed.
quote: What do you think of Google using AI to provide search results.
Google search has been declining for years due to state censorship, commercial bias and restrictions on serving results. The short summaries of click through web pages are a minor improvement, but outweighed by the years-long reduction in quality.
quote: to allow wealth to access skill while removing from skill the ability to access wealth.
Basically a definition of the industrial revolution and mass production, which has allowed ordinary people to have things, instead of just the rich. It has also restricted famines and supported economic and social development. So not a bad thing offline. It may well be beneficial online too. Work changes.
quote: creative work is essentially being stolen at scale
Copied and cited, not stolen (albeit not always with a citation). Picasso said 'good artists copy, great artists steal'. All artists take from a range of sources and produce something new. None work in a vacuum. We do not view that as theft.
quote: The conflation of prediction with intelligence?
Yes. The solution to that is better education. As required by all new technologies until they bed down. When a rumour spreads about somebody that leads to them being killed by a mob for blasphemy, the problem is not the tech, but the people, their educational limitations and their superstitions, usually promoted by their lousy, inadequate government.
In most cases, especially social media, the tech is blamed for human failings and human behaviour. Time to be a bit more honest about that. Don't blame tech that is merely highlighting human shortcomings.
quote: I think people are in for disappointment. ... current technology is too flaky to make that work well.
Absolutely.
quote: Should AI models that affect the public be required to disclose their training data?
Yes. Cite sources like kids do in school essays and researchers do. No need to include 'that affect the public'. All AI systems should.
European privacy? Amusing. Receive a parcel from the EU and you have to offer up everything but your inside leg measurement, all of which goes on the paperwork and label. European privacy my arse.
"Basically a definition of the industrial revolution and mass production, which has allowed ordinary people to have things, instead of just the rich. It has also restricted famines and supported economic and social development. So not a bad thing offline. It may well be beneficial online too. Work changes."
People that tend to quote the industrial revolution, usually have zero knowledge of how it actually affected people.
For a very large percentage of people, the standard of living plummeted, poverty increased, life expectancy dropped and disease increased.
Often Luddite is banded around without, once again, the person using the term having absolutely no idea what they were about.
Luddites, contrary to popular belief we're NOT anti-industrialsation, far from it. All they originally wanted was time to adapt,, government support for those that lost their jobs and a fair wage. Think them more as the first trade union, rather than anti-technology. It's only when the government and company owners ignored them, did they start taking revenge.
Technology and it's lawyers does a pretty efficient job of smacking down any bioform that "scrapes" artistic material from the internet for their own consumption. I more or less understand the necessity of that, although not with artists being threatened with legal for displaying their own work (https://petapixel.com/2016/11/22/1-billion-getty-images-lawsuit-ends-not-bang-whimper/ https://petapixel.com/2016/08/04/getty-images-sued-accused-misusing-47000-photos/).
Why should training data for AI get a blanket exception?
.... Ladies and Gentlemen, Start urEngines. Let the Greater IntelAIgent Games Begin, Expand and Prosper.
Your worlds are rotten ripe ready for a’changing fundamentally, and all the basic and complicated tools that you need to proceed and accomplish any radical and/or popular task are virtually free and universally available to you. Who Dares Win Wins laying Waste the Blind See of Sore and Sad Serial Losers.
.... but now it's just more bollocks in a sea of AI slop. ..... sabroni
That would be in your not so humble opinion, sabroni, a popular simple view of AI which guarantees its stealthy progress generally practically unnoticed and virtually invisible and therefore physically impossible for anyone or anything to prevent it from just doing its GenAI/DeGenAI thing.
AI thanks you for your service. Keep up the good work.
quote: (a) they have lost the plot, and gone a very long way from "Don't be evil," and (b) they are becoming increasingly powerful, with almost no constraint.
An accurate description of every nation state regime out there.
===> So, a whataboutism that does not change the comments meaning or context !!! <===
quote: clear, persistent evidence of harm.
That has been claimed of every new concept or technology from the printing press and Galileo to Darwin, electricity and Elvis gyrating. It just takes a while for people to get used to it. It is not a reason to ban it.
It is lazy and crazy to compare smoking with technology.
===> Yet more whataboutism. The fact it has been claimed before in other instances does not invalidate the claim now. Also no request to ban it has been made ... JUST it needs control before it is too big/powerful to be controlled. <===
quote: The threat to democracy from automatically generated misinformation and deepfakes.
There are far greater threats to democracy from extreme and populist politicians. Nothing Russia or China can cook up compares to the Democrat/Republican toxicity in the US and nothing Russia or China could do would damage a G7 country as much as Brexit did the UK. Stop diverting attention from the real dangers to 'anything foreign'. It just fuels xenophobia.
===> Yet more whataboutism. ALL threats are a danger, the point is that they ALL have to be assessed and handled appropriately. Ignoring threats has never been a successful strategy EVER. Misinformation is a danger at whatever scale it happens and computer generated misinformation is the worse because it is so easy to generate, so fast to propagate and so difficult to control after the fact ... like putting an evil djinn back into the bottle !!! <===
quote <from question by The Register>: Should free, global, instantaneous content distribution have stronger safeguards that make it easier to hold bad actors accountable.
ie. Shouldn't nation state regimes be allowed to operate blanket censorship over the internet? No. Because there is no halfway house. That level of control is Orwellian, straight from the wishlist of every dictatorship.
===> This from a 'question' by the Register to Gary Marcus *NOT* a statement by Gary Marcus !!! Misdirection to set a mindset against the author by using <quote:> out of context. <===
quote <from question by The Register>: limiting the reach of individual accounts to no more than X number of people?
Only the government can reach everyone, whilst the rest of us remain isolated in their own tribes, or our own towns? Again, on the wishlist of every dictator.
===> This from a 'question' by the Register to Gary Marcus *NOT* a statement by Gary Marcus !!! Misdirection to set a mindset against the author by using <quote:> out of context. <===
quote: In 2024 I see a lot of disillusionment.
Much more coming in 2025. AI will stutter, the bubble will deflate, it will find niches, and everyone will just carry on. It might not vanish entirely like the metaverse, but the more thoughtful will look back on the moral panic and feel embarrassed by it all. Perhaps moral panics are the new prolefeed.
===> Mainly concur apart from the 'Moral Panic' nonsense. There is no 'Moral Panic' & there will be little to be embarrassed about unless the early warnings are ignored and GenAI is allowed to grow in an uncontrolled way. <==
quote: What do you think of Google using AI to provide search results.
Google search has been declining for years due to state censorship, commercial bias and restrictions on serving results. The short summaries of click through web pages are a minor improvement, but outweighed by the years-long reduction in quality
===> Mainly concur. <===
quote: to allow wealth to access skill while removing from skill the ability to access wealth.
Basically a definition of the industrial revolution and mass production, which has allowed ordinary people to have things, instead of just the rich. It has also restricted famines and supported economic and social development. So not a bad thing offline. It may well be beneficial online too. Work changes.
===> Addressed in https://forums.theregister.com/forum/all/2024/10/21/gary_marcus_ai_interview/#c_4952261 <===
quote: creative work is essentially being stolen at scale
Copied and cited, not stolen (albeit not always with a citation). Picasso said 'good artists copy, great artists steal'. All artists take from a range of sources and produce something new. None work in a vacuum. We do not view that as theft.
===> As the entity is not an 'Artist' but a machine/computer any nonsense regarding deriving 'New Art' from old is just noise. The original art is being taken, for no fee or via no agreement with the original artist. This taking is nothing more than 'Theft' without this 'taken art' there would not be anything 'new' as produced by the GenAI. To prove the need for the 'Taken art' remove all the input that has been taken and task the GenAI to produce new art of its own, the result would be Zero/Zilch/Nada. Derivative art from the mind of a real 'meatsack based' artist may be argued as new as it involves some new artistic input from the artist. New art from a GenAI is 100% derived from the 'taken art' and cannot be argued as New or influenced by in any meaningful way. I am afraid theft is theft no matter how you try to wriggle out of it !!! <===
quote: The conflation of prediction with intelligence?
Yes. The solution to that is better education. As required by all new technologies until they bed down. When a rumour spreads about somebody that leads to them being killed by a mob for blasphemy, the problem is not the tech, but the people, their educational limitations and their superstitions, usually promoted by their lousy, inadequate government.
In most cases, especially social media, the tech is blamed for human failings and human behaviour. Time to be a bit more honest about that. Don't blame tech that is merely highlighting human shortcomings.
===> The solution to that is better education. The rest of your diatribe says more about *you* than anything useful about this article etc. <===
quote: I think people are in for disappointment. ... current technology is too flaky to make that work well.
Absolutely.
===> Concur <===
quote: Should AI models that affect the public be required to disclose their training data?
Yes. Cite sources like kids do in school essays and researchers do. No need to include 'that affect the public'. All AI systems should.
===> Concur <===
European privacy? Amusing. Receive a parcel from the EU and you have to offer up everything but your inside leg measurement, all of which goes on the paperwork and label. European privacy my arse.
===> !!!??? <===
:)
'European privacy? Amusing. Receive a parcel from the EU and you have to offer up everything but your inside leg measurement, all of which goes on the paperwork and label. European privacy my arse.'
Can you cite a source for this and show us your training data?
My concern with summarization is that it adds another layer of separation between end users and primary sources - and one that offers an oportunity for deliberate distortion and manipulation.
The guardrails built into existing systems - however well-intentioned they might currently be - demonstrate that it's (a) possible, and (b) being actively worked on.
Dumb question? How is the glut of ML models being monetised? Ineffective advertising? I don't believe that.
The cost of running, hardware and electricity is practically at an all-time high, yet none of us (at least, nobody I know) has purchased anything on this front. If not actively avoiding it.
The demand side of "Supply and demand" has ways of killing things that are bad ideas. So why hasn't this happened yet?
Dumb question? How is the glut of ML models being monetised? Ineffective advertising? I don't believe that.
It's not a dumb question, and they're not being monetised (in an effective or sustainable way). They're being propped up for the time being by Venture Cap hype (SoftBank's Vision Fund posted a record $32Bn loss last year. Slow clap).
As this piece discusses, OpenAI in particular has no route to profit. They're slated to lose $4-5Bn this year, but unlike social media or traditional cloud services there's no real economy of scale which will tip them to profit when they get enough customers.
Every prompt uses a certain amount of compute (electricity) to complete. This is a linear cost - if you upload 100 photos to Faceache, it doesn't cost them 100x more than if you uploaded 1 photo. Cloudy and collab services like OneDrive are built on storage (which is very cheap) and lighter users subsidise heavier users. Even inactive users have a value to social networks by filling in the social graph.
But serving LLMs and GenAI has a fixed (and predictable) cost per request. It's a traditional consumption model. And the underlying cost is probably more than people actually want to pay for a chatbot. Plus you need to recoup basic running costs (payroll) and the ruinous investment of training each model. Paid users are going to be using it more than free users, and probably costing it more than they're paying.
Sooner or later, people are going to catch up with this and either go bust or pivot hard into something useful (like classical AI/ML for protein folding, secondary analysis of medical imaging, etc).
The Register 
Biting the hand that feeds IT
Copyright. All rights reserved © 1998–2024
