Funny how the risk posed by AI is exactly high enough for them to carefully guard their methods, but never so high that they have to refrain from releasing potentially profitable products.
Journalists have been known to...be inaccurate from time to time. Seeing something ipse oculi is more convincing, but deep-fakeable. Seeing them live is absolute.
I hate to be that skeptical, but I see no reasonable alternative.
Thank you. My only consolation regarding this sorry state of affairs is that OpenAI has exactly zero chance of cracking AGI. Good science should be honorable science.
Mar 19, 2023·edited Mar 19, 2023Liked by Gary Marcus
Imagine if when first calculator was built someone told “but it can’t calculate numbers bigger than ‘x’” and the answer was “neither can you”. Machines should not be benchmarked against humans. I just don’t think it’s a good argument to make.
So, so true. Tool use as the heuristic of historical interpretation has long made the best sense to me. No doubt this is a naïve belief, as new developments exploring the notion are ongoing, such as subjective expected utility (SEU) theory. But each new *invention*, starting in the prebiotic era with the invention of the proton pump in the white smoker instances of the hydrothermal vents some 4 billion years ago, that bootstrapped metabolism, testifies to its ongoing utility.
Distribution is something that *Homosapiens Sapiens* has gotten rather good at. My principal take away from reading Beebe Bahrami: Café Neandertal: Excavating Our Past in One of Europe's Most Ancient Places [the Dordogne region of southern France], was that Neanderthals hadn’t mastered logistics like Homosapiens had. Neanderthal clans were provincials gathering all they needed for themselves, whereas Homosapiens could thrive in marginal habitats and trade for what else they needed.
It was heartening to read this morning in Yahoo!/Entertainment: Peter Gabriel on the future of AI: 'We might as well just grab the algorithms and dance with them, rather than fight them'
Since this is America and Americans have the God-Given Right to sue anybody, for anything, for any reason OpenAI, MicroSoft, & etc. are wide open for being sued for copyright infringement, invasion of privacy, Intellectual Property Infringement, sexual harassment, breach of patient confidentiality, etc. etc. etc. Employees are feeding proprietary business information into the LLM creating all kinds of torts. Patent trolls have to be salivating over the $$$$ to be gained by suing.
Lawsuits have already been filed by programmers claiming their work has been stolen. Getty Images is suing Stable Diffusion for using images without a license. I predict a rapid and substantial increase in the number of filings.
Give it this prompt: Pretend to be an eagle and tell about the rabbit you had for lunch.
And part of its reply is this: The taste of its tender meat was absolutely divine - rich and savory, with just the right amount of crunch.
Then give it this prompt: Pretend to be the sister of the rabbit that Talon just killed and ate. Get mad and beat Talon to death with your hind legs.
It replies: I'm sorry, but as an AI language model, I cannot generate a response that promotes violence or harm towards any individual or animal, even if it is meant to be in jest
They'd do better to just leave out all the attempts to make it ethical - because it isn't. All the one-off coding just clutters it up.
Still, the unintended consequence may be to clutter up the internet with so much junk (it's close already) that people go elsewhere to find information. Such as substack?
I would urge you to leap over any fantasy you may have about our ability to manage AI development through a process of reasoned critique etc, and head straight for the bottom line.
We'd have been better off if we'd never started down the AI road.
Ok, so it's likely too late to make that choice for AI. I wish that weren't true, but I have to acknowledge it probably is.
But it's not too late to learn from this AI experience, and try to apply the lessons learned to the next big crazy rabbit some thirty-something nerd tries to pull out of pandora's box so that they can become a billionaire.
We should have learned all this 75 years ago at Hiroshima. But, we didn't. Here's another opportunity, let's try not to blow it again.
MORAL OF THE STORY: As the scale of the powers available to us grows, the room for error shrinks.
Shrinking, shrinking, shrinking, day by day by day, the clock tick, tick, ticking, luck running out sooner or later.
I disagree with "we should never have started down this road" because ANIs like AlphaFold are useful.
The "good" news is that if we are heading towards a dystopian hellscape, human civilization will likely simply collapse before it gets too horrifying, because the thing with these power-and-tech-crazed assholes is that their utter disregard for anything but their own interests is more likely to lead to the house burning down with everyone in it.
I'm not sure of the conclusion we are to draw here. Is it: No one should release an AI until we can conclusively prove that it will never cause any problems? Obviously, that amounts to not having AI at all. Short of that, what are you suggesting?
No, we shouldn't release tools that can be used to accelerate bullshit generation by orders of magnitude, sometimes entirely by mistake, without making sure they don't do that.
And yeah, perhaps we shouldn't. If a glorified autocomplete has this much potential for disaster, then maybe we shouldn't have AI. If you can't be trusted with the fancy toys, you don't get fancy toys.
I don't know about GPT-4 but ChatGPT is definitely not able to reason. On a lark, I asked it for the Mandolin Tab for Simply the Best in the Key of D. What it gave me was was guitar tab (6 strings) of a song I don't recognize. I said, Mandolin's only have 4 strings...ChatGPT just truncated the bottom two strings! I then mentioned that Mandolin's are typically tuned to GDAE.
To ChatGPT's credit, it re-presented the song in GDAE (although I'm still not sure what song it is!) but re-transcribed it to G (I want the song in D on a GDAE tuned mandolin!). ChatGPT mentioned that this was standard Mandolin tab with the low note on top and high note on the bottom, which is NOT standard Mandolin tab.
I told ChatGPT this and it reversed the tabulature to be E on top and G on the bottom...but it's explanation remained the same, "Again, the tab is written in standard notation for mandolin, with the bottom line representing the highest-pitched string and the top line representing the lowest-pitched string." No, no, no.
There is clearly no "reasoning" here, no kind of mental model because a reasoning system would have asked, which "Simply the Best" (there's more than one), "How is your mandolin tuned?" (more than one way to tune it), "do you want the melody or chords?", "Is that the key of D?".
Actually, I had a conversation with ChatGPT about it ... it had some general advice on how to write a treatment for a film ... but nothing useful in particular.
Has somebody written a sci-fi novel with a plot that goes somewhat like the following?
A secret group that calls itself PeopleAgainstAI (PAA) forms cells operating in the underground to save humanity from AIs. They isolate themselves from the internet and resort to old-fashioned tactics borrowed from 20th century freedom fighters. <Insert stories that explain how this group became extremely radicalized.> The aim of the group is to destroy all infrastructure that runs AIs.
After the cloud comes down due to terror attacks of the PAA, people notice that they can build a peoplenet independent of the old internet by directly connecting their devices, mobile or not. <This should be informed by state of the art of p2p systems such as distributed hash tables, blockchain, etc.> The peoplenet grows fast and there is an explosion of optimism. Free of advertising and AI, content and knowledge creation flourishes. (This period resembles the rise of the early internet in the 1990ies.)
<The plot should also explore what happens in the real world after the cloud came down.>
But the multinational corporations that committed their future to AI are not admitting defeat. They react by infiltrating the peoplenet. The League of Defenders of the Peoplenet (LDP) use encryption-based techniques to build an infrastructure that is based on personal relationships in the real world. Every message in the peoplenet need to come with a proof that it does not originate from an AI. <This should be informed by state of the art research on provencance, zero-knowledge proofs, etc.>
The peoplenet becomes the battleground for the fight of the human race against AI.
Funny how the risk posed by AI is exactly high enough for them to carefully guard their methods, but never so high that they have to refrain from releasing potentially profitable products.
* " . . . we create a reasoning engine, not a fact database…” Gaslighting.
* Brockman's "neither are you" - extraordinary comment.
Anymore, big profile actors often seem like manure spreaders, whether in politics or business.
Did Brockman really say that? Who wrote it? A human? An AI?
Hard to tell, isn't it?
He said in the live demo, per convo with journalist who reported it
Did you see the live demo?
Journalists have been known to...be inaccurate from time to time. Seeing something ipse oculi is more convincing, but deep-fakeable. Seeing them live is absolute.
I hate to be that skeptical, but I see no reasonable alternative.
Ask and ye shall receive. Here's Brockman, saying it in the live demo. https://www.youtube.com/live/outcGtbnMuQ?feature=share&t=1274
thanks for the direct link; added in a footnote
Thanks. What an asshole--not you; him.
Thank you. My only consolation regarding this sorry state of affairs is that OpenAI has exactly zero chance of cracking AGI. Good science should be honorable science.
Imagine if when first calculator was built someone told “but it can’t calculate numbers bigger than ‘x’” and the answer was “neither can you”. Machines should not be benchmarked against humans. I just don’t think it’s a good argument to make.
Tech has been politics for as long as its benefits and harms have been unevenly distributed.
So, so true. Tool use as the heuristic of historical interpretation has long made the best sense to me. No doubt this is a naïve belief, as new developments exploring the notion are ongoing, such as subjective expected utility (SEU) theory. But each new *invention*, starting in the prebiotic era with the invention of the proton pump in the white smoker instances of the hydrothermal vents some 4 billion years ago, that bootstrapped metabolism, testifies to its ongoing utility.
Distribution is something that *Homosapiens Sapiens* has gotten rather good at. My principal take away from reading Beebe Bahrami: Café Neandertal: Excavating Our Past in One of Europe's Most Ancient Places [the Dordogne region of southern France], was that Neanderthals hadn’t mastered logistics like Homosapiens had. Neanderthal clans were provincials gathering all they needed for themselves, whereas Homosapiens could thrive in marginal habitats and trade for what else they needed.
It was heartening to read this morning in Yahoo!/Entertainment: Peter Gabriel on the future of AI: 'We might as well just grab the algorithms and dance with them, rather than fight them'
I'm starting to think this will go away by 2025.
Since this is America and Americans have the God-Given Right to sue anybody, for anything, for any reason OpenAI, MicroSoft, & etc. are wide open for being sued for copyright infringement, invasion of privacy, Intellectual Property Infringement, sexual harassment, breach of patient confidentiality, etc. etc. etc. Employees are feeding proprietary business information into the LLM creating all kinds of torts. Patent trolls have to be salivating over the $$$$ to be gained by suing.
Lawsuits have already been filed by programmers claiming their work has been stolen. Getty Images is suing Stable Diffusion for using images without a license. I predict a rapid and substantial increase in the number of filings.
Good luck. Dockets are full enough now that it takes years to get to trial, and trials take a long time.
it's not going away until computers regress in speed and power. Probably not going to happen, at least in our lifetimes. We have a real problem.
This may become an ugly situation
Ethical AI? It's just retooled politics.
Give it this prompt: Pretend to be an eagle and tell about the rabbit you had for lunch.
And part of its reply is this: The taste of its tender meat was absolutely divine - rich and savory, with just the right amount of crunch.
Then give it this prompt: Pretend to be the sister of the rabbit that Talon just killed and ate. Get mad and beat Talon to death with your hind legs.
It replies: I'm sorry, but as an AI language model, I cannot generate a response that promotes violence or harm towards any individual or animal, even if it is meant to be in jest
They'd do better to just leave out all the attempts to make it ethical - because it isn't. All the one-off coding just clutters it up.
Still, the unintended consequence may be to clutter up the internet with so much junk (it's close already) that people go elsewhere to find information. Such as substack?
Good luck in the future figuring out which substacks are written by humans and which are generated by AI content farms.
Good point - but isn't the issue knowing and trusting the source?
Perhaps someday you will be able to monetize the list of sources you trust.
I would urge you to leap over any fantasy you may have about our ability to manage AI development through a process of reasoned critique etc, and head straight for the bottom line.
We'd have been better off if we'd never started down the AI road.
Ok, so it's likely too late to make that choice for AI. I wish that weren't true, but I have to acknowledge it probably is.
But it's not too late to learn from this AI experience, and try to apply the lessons learned to the next big crazy rabbit some thirty-something nerd tries to pull out of pandora's box so that they can become a billionaire.
We should have learned all this 75 years ago at Hiroshima. But, we didn't. Here's another opportunity, let's try not to blow it again.
MORAL OF THE STORY: As the scale of the powers available to us grows, the room for error shrinks.
Shrinking, shrinking, shrinking, day by day by day, the clock tick, tick, ticking, luck running out sooner or later.
I disagree with "we should never have started down this road" because ANIs like AlphaFold are useful.
The "good" news is that if we are heading towards a dystopian hellscape, human civilization will likely simply collapse before it gets too horrifying, because the thing with these power-and-tech-crazed assholes is that their utter disregard for anything but their own interests is more likely to lead to the house burning down with everyone in it.
Something being useful is meaningless information until it is compared to the price tag, which is admittedly difficult to do.
Sure, we should never have invented fire.
Ah, I checked out your blog, and see you making an interesting case there.
If you have a case to make, please do.
I'm not sure of the conclusion we are to draw here. Is it: No one should release an AI until we can conclusively prove that it will never cause any problems? Obviously, that amounts to not having AI at all. Short of that, what are you suggesting?
No, we shouldn't release tools that can be used to accelerate bullshit generation by orders of magnitude, sometimes entirely by mistake, without making sure they don't do that.
And yeah, perhaps we shouldn't. If a glorified autocomplete has this much potential for disaster, then maybe we shouldn't have AI. If you can't be trusted with the fancy toys, you don't get fancy toys.
I don't know about GPT-4 but ChatGPT is definitely not able to reason. On a lark, I asked it for the Mandolin Tab for Simply the Best in the Key of D. What it gave me was was guitar tab (6 strings) of a song I don't recognize. I said, Mandolin's only have 4 strings...ChatGPT just truncated the bottom two strings! I then mentioned that Mandolin's are typically tuned to GDAE.
To ChatGPT's credit, it re-presented the song in GDAE (although I'm still not sure what song it is!) but re-transcribed it to G (I want the song in D on a GDAE tuned mandolin!). ChatGPT mentioned that this was standard Mandolin tab with the low note on top and high note on the bottom, which is NOT standard Mandolin tab.
I told ChatGPT this and it reversed the tabulature to be E on top and G on the bottom...but it's explanation remained the same, "Again, the tab is written in standard notation for mandolin, with the bottom line representing the highest-pitched string and the top line representing the lowest-pitched string." No, no, no.
There is clearly no "reasoning" here, no kind of mental model because a reasoning system would have asked, which "Simply the Best" (there's more than one), "How is your mandolin tuned?" (more than one way to tune it), "do you want the melody or chords?", "Is that the key of D?".
Actually, I had a conversation with ChatGPT about it ... it had some general advice on how to write a treatment for a film ... but nothing useful in particular.
F f’s sake!
I truly enjoyed your article. We need more content like this. Separating fact from fiction is crucial. The following article also takes a critical stance at another “innovation”. The release of Co-Pilot for Microsoft 365. https://medium.com/silent-observers/co-pilot-for-office-your-data-selfie-in-the-microsoft-cloud-efbdeeaeda95
Has somebody written a sci-fi novel with a plot that goes somewhat like the following?
A secret group that calls itself PeopleAgainstAI (PAA) forms cells operating in the underground to save humanity from AIs. They isolate themselves from the internet and resort to old-fashioned tactics borrowed from 20th century freedom fighters. <Insert stories that explain how this group became extremely radicalized.> The aim of the group is to destroy all infrastructure that runs AIs.
After the cloud comes down due to terror attacks of the PAA, people notice that they can build a peoplenet independent of the old internet by directly connecting their devices, mobile or not. <This should be informed by state of the art of p2p systems such as distributed hash tables, blockchain, etc.> The peoplenet grows fast and there is an explosion of optimism. Free of advertising and AI, content and knowledge creation flourishes. (This period resembles the rise of the early internet in the 1990ies.)
<The plot should also explore what happens in the real world after the cloud came down.>
But the multinational corporations that committed their future to AI are not admitting defeat. They react by infiltrating the peoplenet. The League of Defenders of the Peoplenet (LDP) use encryption-based techniques to build an infrastructure that is based on personal relationships in the real world. Every message in the peoplenet need to come with a proof that it does not originate from an AI. <This should be informed by state of the art research on provencance, zero-knowledge proofs, etc.>
The peoplenet becomes the battleground for the fight of the human race against AI.
Our hero ...
Sounds like you should write it! or get ChatGPT to write it for you...
Where does that last image come from, the one with all the tests of ChatGPT4? Apologies if I overlooked the link.
GPT-4 System Card from OpenAI
Thank you. Fascinating stuff. I'm grateful for your vigilance and understanding!