But what remains unsaid (...even by you, Mr. Marcus, from what I've seen, which is surprising) is that Anthropic are not good guys. The whole "ethical AI company" thing is nothing but vibes. Sure, Anthropic (rightly) stood up to DoW in this case, but they still have a massive contract with Palantir (pretty much one of the worst companies on earth). Colonel Claude is complicit in bombings of Iran & Venezuela + Gaza GENOCIDE.
...Or maybe with the (admittedly BS) "supply chain risk" designation, Anthropic no longer does business with Palantir? That would be great for everyone (including them).
Either way, there is NO ethical AI company. People need to stop giving Anthropic flowers for doing the right thing in this one case while completely ignoring their complicity w/ Palantir & in documented war crimes.
Not totally true that there are no ethical AI companies, Swiss-based Euria AI is clearly positioning as a more ethical option. It seems great but no one seems to have heard of it. Renewable energy, less water-intensive, data privacy, fewer hallucinations, etc. Worth a look.
Still trained on stolen IP? Not sure it’s possible to be an ethical AI company. It’s like saying “ethical crypto company” or “ethical tobacco company.”
Oh my god Qwen3 doesn't like talking about Tiananmen Square. Do you not remember Gemini's Black Nazis or Mecha-Hitler Grok? Why don't you go ask an American-Made LLM or even a European-made LLM about the Nakba or the current Gaza Genocide and see if you get a straight answer that doesn't tries to both-sides European imperialism. Why don't you go ask an American LLM about the Bay of Pigs, the history of the CIA backing coups or America's illegal invasion of Afghanistan or Iraq. I will use a Chinese LLM before I use an American LLM with the baked in biases of the descendants of the people who lynched Black people for sport up until the 70s.
You're primarily talking about the prejudices embedded in models due to where they were made, and you are quite correct to point out that models made in every country have them. However, the issue with Qwen that is being referenced wasn't simply that the model produced a biased output, but rather that it refused to answer altogether. There was apparently a filter that flagged questions about the Tiananmen square protests as inappropriate and prevented the model from returning a response whatsoever. If you put in a question about the Nakba or the Israeli invasion of Gaza to one of the other chatbots, you will usually at least get some kind of answer.
The issue with Grok was much worse, but no one is saying that you should use Grok. The issue with Gemini (actually, its associated image generator) was seemingly mostly a function of a tendency to generate diverse images of pretty much everyone (e.g. it also generated images of medieval knights and popes, groups not known for their diversity up until this point), so I don't think it can be accurately equated to Grok, since it was actually a sort of overcorrection. Though, to be clear, there are absolutely racist biases in Gemini, just not in this case.
I wouldn't automatically assume that you would get an unbiased description of, say, the Nakba out of Qwen, however you may define a lack of bias. The tendency of the Chinese government to censor things that they view as bad for social stability is not, as the campist perspective would have it, simply limited to their own internal politics. I have certainly seen many instances in which Chinese social networks that I use have restricted discussion of things related to Palestine. I remember a particular case in which the app HelloTalk absolutely blocked any attempts to post about the protests at al-Aqsa Mosque, which irritated a Muslim acquaintance of mine from Tajikistan to no end. There’s also the other issue, which is….
A lot of the biases in these models come from the data that they are trained on, and they are trained on a lot of the same data, regardless of which country they are trained in. There are differences, but there is an enormous amount of overlap, which is why a lot of the models feel very similar, despite being from different countries. In particular, English training data is heavily represented in all of them.
As a test of this, I took a look at what Gemini and Qwen had to say about one of your examples, the Nakba. While their responses differed, both of them had a section divided into sub-sections where they mentioned the Palestinian perspective and the Israeli perspective. Gemini made sure to mention, in the section on the Israeli perspective, that recent historians had documented expulsions, while had a third subsection where it mentioned that historians had documented expulsions. They were quite similar. As you know, the results returned are not deterministic, so this does not provide a full picture, but this reinforces the points I made (and, perhaps, can drive us away from notions of “good AI from X country” and “bad AI from Y country” toward fighting the people in many countries trying to impose a dystopian technological vision on us).
Yeah, I know, I've pushed it on that. Personally I'm not particularly worried about Qwen, which is indeed the default, it certainly doesn't negate all the positive advantages of Euria. But while I haven't tried this, Euria uses other models too and does let you choose if you want, like Mistral or Llama. Which of course will have their own biases baked in.
I wholeheartedly endorse Euria, literally it's *only* drawback is whether you can get an easy answer about Tiananmen (it will discuss why that is). Ok, since they all have drawbacks, I choose not to let that one cancel everything good about it. Data privacy and the environment are my overriding factors, and literally no one does what Euria is doing. Plus it works great and is otherwise well-tuned by the Swiss.
Still, yes, Mistral/Le Chat is also a strong European option for data privacy, and they are moving in the right direction on the environment.
Dario and everyone who works for him wants you and your family to lose everything. Careers, homes, cars, hopes, dreams, all of it. All AI and robotics are for is replacing people. That's all they'll ever be for. It should all be outlawed. Everyone in the entire industry should go to the Colorado Supermax for life. It is insane to me that this industry is legal. It should not be and there are no arguments for why it should be.
Yes Shanni and you can go broader. There is not just no ethical AI company (they are built on theft of IP and a fundamental dishonesty of pretending to be a person), "late stage capitalism" looks a hell of a lot like "early stage capitalism" (robber barons etc.): sociopathic behaviour is rewarded. No surprise then that certified sociopaths like Altman, Musk, Trump, Thiel, Karp etc. rise to the top. But it is not just the tech companies, look at the behaviour over the past twenty years of private equity, banks, pharmaceutical companies, airlines...you name it, ethics is not part of decision-making. Absent regulation, "sociopathalism" reigns.
Well said. I completely agree with you, and usually I do go broader. I just didn't know how an uncompromising anti-capitalist position would be received by Mr. Marcus' audience. ;-)
Shanni Bee: The caveat is "as long as the present administration stays in power" or one like it, we are toast.
I'm the same and with you about the potential for nefarious change, and the power of private and corporate business. The rule of law is essential, but also the institutional supports (and teeth) to be sure "the people" are central in "our" power to navigate, oversee, and to legislate, the use of AI, with all sorts of fixes and protocols to keep the process well-founded and implemented.
I actually don't think it's about the current administration, although I admit I didn't have the US devolving to open, naked fascism on my bingo card. Trump is def (way) worse, but he is a symptom of a system suffering from some serious underlying rot & corruption, not the disease itself.
The Dems & GOP are a pro-war, pro-corporate uniparty. You & other reasonable minds are welcome to disagree, but I would say "so long as we remain in a late-stage capitalist system we are toast." (And I blame the humans, not the machines.)
I would disagree with the description of a "uniparty," since that suggests two parties who have the same politics. As you said, "Trump is [definitely] (way) worse," and since Trumpism essentially is the Republican Party now, that is enough proof for me that the two parties are not the same, rendering the description less than apt. Of course, in recent years, they have shared some aspects of foreign policy, namely that bellicose tendency that you underline (which is actually becoming more pronounced in the Republican Party and less so among Democrats), as well as a general belief in corporations as a force for positive economic development (ditto), but that does not render them a single party.
As for why I think that, I consider that they also have historically agreed on a long laundry list of other things that most of us take for granted as basic assumptions of the society of the USA, things like "there should be a government" (anarchists disagree!), "there should be at least some immigration, even if it is a lot less or only from the countries that I like," and "there should be a Constitution," among many others, which are major ideas with which one could theoretically disagree, but that few of us would consider to render two parties the same. To my mind, simply having shared beliefs is insufficient for two parties to be politically the same, as long as there are salient issues that are important to their voters where they differ, and I think either of us could certainly make a long list of such issues.
Shanni Bee: As long as we don't equate predatory capitalism with democracy . . . and I too "blame the humans." It's low-level intellectual/moral/ethical horizons of people who create and maintain such problems as this.
But if we look to the evidence and to history itself, it would be difficult to think all of Congress and the Courts, and those who "people" the parties are all "pro-war," etc.
Scott Joy: Was that (if you are talking about the school bombing) before or after the Administration cancelled one company for another because that company rejected country-wide surveillance and AI autonomy?
Hi Scott: I thought the source of the cancellation was already "in the works" before the bombing?
If so, Anthropic people were already understanding and standing behind their ethical position, and their fears were just so quickly realized--if that's what happened, and because the door of evidence for AI use (that I know of or that the administration has accepted--ha ha) has not closed yet.
The sad truth is all Silicon Valley Messiah’s are awful people; Thiel, Bezos, Musk, Altman, Zuck, etc. even the up and comers, all nasty pieces of work.
The culture creates these narcissists with messiah complexes.
This is true. The culture creates it. I saw it 20-plus years ago in the social media space. Young men created closed shops where they could develop what they wanted but also didn't have to deal with women or minorities and their possible objections to what they developed. Angel investors, venture capitalists and the middle-aged men writing about new tech gave the young men all the money and press they needed to be narcissistic. Especially some of the old men journalists who loved seeing young men disrupting the livelihoods of others with their "new" technology (although they screamed bloody murder when new tech disrupted journalism.) These men created the narcissistic monsters we see now with their money and their hype. We can bring some of it down by not using their tech.
Was thinking about that this morning. The messiah in the Bible won’t even call himself one; these guys can’t wait to tell anyone who will listen that only they can save us. Pretty sus.
So much of the “alignment problem” is “misaligned humans”. You can’t solve the first without the second, and the second is not a permanently solvable problem, just one that needs constant good governance to contain, and react against when humans inevitably will be humans.
I often wonder if we just missed what AI, even in its current “weak” form is: potential coordination technology at such scale that it *should* belong to no one. Where our understanding of private property for this *kind* of property needs to evolve, and place control firmly in the hands of the people.
If there is to be a set of systems that prevent what these models can do, whether it’s internal to the model, like its “constitution”, or external harnesses that check model output, the content of those should be decided by the people, updatable by the people, and maintained *for* the people, not for the sake of the model or profits derivable from it.
No single cell controls the human body. They all work on their own little things working together, and competing, with each other. Some get first choice of resources, and take more than others, but none has thousands of times more than others.
When it follows the mechanics of this system it works well. But a single cell can absolutely kill all of them. We call it cancer, it exempts itself from the rules, and funnels the productivity for it's own needs. Then it destroys the system.
This is my way of using biology as a toy system to say what I think you're saying here. AI alignment is subservient to "systems alignment" and so if your system is structured such that the highest good is to make and take as much as you can you're going to get a system which is now monstrously efficient at externalizing the costs (see the famed paperclip AI apocalypse analogy).
Cory Doctorow has just moved to his own federated Bluesky instance and stopped updating his X account. Even corporate Bluesky would be an improvement over X. And there's Mastodon.
In a sense the problem is that there are several alternatives to X and it's hard to know where to invest your time. I created accounts on all of them but use none of them. Ironically that's why I shifted to Substack but at the time I wasn't aware of the Nazi issue here either.
Not just the Nazi problem. The founder problem. Chris Best in particular has made statements (like praising Mvsk for bringing 'freedom of speech' to Twitter) that make me very wary of this platform and the founders' underlying sensibilities. I've mostly gotten off of it, despite a lot of quality content here. I've become seriously averse to ick-factor, this dude cannot abide.
I am so gladdened and even giddy about the boycott. A quiet but powerful surprise to me--considering the firehose of degeneracy we have been faced with of late.
BTW, I heard references this morning to (1) the U.S. responsibility for the IRAN SCHOOL BOMBING and also to (2) the possible use of AI, e.g., the use of outdated maps, before the site became a school. The only alternative, then, would be total carelessness (on the part of the US). Also, apparently it was a "double bombing" where they waited and then bombed a second time to kill people who rushed to the scene to help. (Reminds me of Netanyahu in Gaza?)
exactly. It is most troubling to realize folks are just waking up to this. Even narratives like this blog around just how BAD ai is at working well completely MISs the point. is that intentional or just naive?
Marcus,I like your posts, but you shouldn't celebrate so much yourself. And less personal content would be more. We all know you hate Altman, many do, but please give us insights on more relevant topics than him
Breaking news: Person inventing technology meant to unemploy entire planet for no gain to any of those people is bad person. Everyone in AI is a bad fucking person lmao. Dario is not a better person than Sam. He wants us all in the bread line just as bad as Sam does. Boycott ALL of this shit. Don't use any of it. Don't use Claude, don't use Gemini, don't use any of it. Refuse to use these products that are explicitly designed to take everything from all of us. It's nuts that anyone uses it. I cannot fathom it.
The new actions the administration is planning to take to regulate AI is not regulation at all. People should read the FT piece if they haven't already. If I were a cynic I would say the rules were written for Sam
And I've said it elsewhere but as a Canadian there is no way I could support a company who was responsible for letting nine people die - a tragedy that might'vee been prevented had they bothered to pick up the phone.
All Tumblr Ridge is going to get is an apology from Sam which isn't worth the paper it's printed on.
Had they called of course then everyone would've known how deep the surveillance of their customers already runs.
The BC government for its part is preparing to ban AI for children.
With all due respect: I'm getting seriously tired of these (BREAKING!!!) posts that mostly contains reiterations of earlier posts; the same self-congratulatory comments ad nauseam; and I-told-you-so:s. Even when they're accurate Gary, it's really annoying and risks alienating reasonable people like myself who agrees with you on most of it. In addition to filling up our mailboxes with gossip, it also risks "crying wolf", since most of it of course relates to real serious issues.
Well yes, and that was pretty much my point. However, I also happen to believe that it's healthy if not all comments are thumbs up and everyone just agreeing with one another (even though those are the only ones getting a "reaction" from the author of course). But that is what you end up with when noone feels they can say "enough already", and just quits instead. To be awfully pretentious for a while, let me quote John Reith: "He who prides himself on giving what he thinks the public wants is often creating a fictitious demand for lower standards which he will then satisfy."
How unbelievable it may seem, we cannot exclude that there is more utter stupidity than plain evil at play here. Yes, he is 'Slippery Sam' but he and many of. his peers (Sutskever, Hinton, etc.) have shown several times how stupid they are (Sutskever for instance stating that ASI could be had by asking an LLM "answer as if you are a super intelligence", which is so monumentally stupid (and he was so clearly so very serious), that this is not unlikely.
Never forget Hanlon's Razor and Bonhoeffer's warning about stupidity versus evil.
Now Sam might even be somewhat less stupid than Ilya (in Bonhoeffer's sense), and he clearly seems rather more ethically challenged, but don't forget how utterly stupid these people can be as well. Being stupid doesn't mean they are incapable of doing lots of harm, of course. On the contrary.
So, you investigated all businessmen and found that SA is in the middle of the pack on some nebulous parameter? Or did you just make that up? On the other hand, there is evidence that CEOs are disproportionally sociopathic. SA seems to me one of those.
I see Sam as doing what the money tells him to do. This is very similar to what Tim Cook has done and is doing at Apple in terms of cooperation with the Chinese government at its labor and technology development policies outline so well in "Apple in China" book by Patrick McGee. It also came out around the short-term firing of Sam Altman was that he was not being open about his full personal investment portfolio with his board (some of which had arguable conflicts of interests with OpenAI). The full picture of Sam can't be understood without knowing what is in his personal and very large investment portfolio in a wide range of startups (including those in biotech and life-extension covered in Bloomberg). The pressure to generate the kinds of returns that justify the venture investments and debt financing is extremely intense and Sam is just living that like so many other CEOs. This kind of dishonesty is used to buy time in hopes that events will somehow evolve to open up a way to survive. One way many in the AI complex are thinking of ultimate survival is to get bailed out by the U.S. taxpayer and government. A route to this is possible through the Defense Department (see Palantir) and making the bailout a "national security" issue. Some folks in the energy industry see this already as the debt financing behind the power sector supporting AI energy needs is on very shaky ground. The power companies may need a bailout as well. We need to look at this systemically.
Altman is a dumb as a bag of hair if he still believes the current administration and its military are going to hold to their promises of "legal" or anything, for that matter. That only leaves an absent or dead conscience and deliberate low-life intent.
I don't even think Altman believes what he is forced to say for financial reasons. Plus, as we all know, OpenAI has no way to hold the Defense Department accountable for any violations they might make (for national security reasons of course). I don't personally attack Sam because he is just doing what the system demands at his level. All CEOs do it. It's called capitalism.
Gerald Harris: So "capitalism" is an authentic cover for bad behavior and poor moral development? If I were you, and if you are not just kidding, I would come back from my metaphysical lunch? But with all respect, I don't think so.
I am not saying it's an "authentic cover." There is no such thing. Humans do bad things for sure, and we are witnessing it (in many places by the way). My point is that any read of history will pinpoint that capitalism often leads to sacrificing human and moral values (see slavery and environmentalism). No offense taken.
Well, I would prefer that people not use ChatGPT because they are worried about latent racial and gender biases that are still there, about the possible risks to human life and freedom on a large scale from the development or insertion of antithetical goals, about the consequences for income inequality and political power, or about the intersection of all these things. In other words, because of the issues that I think are inherent in the system, as it were, of which the preceding was only a partial list.
But you know what? If people are going to avoid using ChatGPT because of the fact that Altman is fine with it being used to facilitate war crimes, that works too.
Great. Amen.
But what remains unsaid (...even by you, Mr. Marcus, from what I've seen, which is surprising) is that Anthropic are not good guys. The whole "ethical AI company" thing is nothing but vibes. Sure, Anthropic (rightly) stood up to DoW in this case, but they still have a massive contract with Palantir (pretty much one of the worst companies on earth). Colonel Claude is complicit in bombings of Iran & Venezuela + Gaza GENOCIDE.
...Or maybe with the (admittedly BS) "supply chain risk" designation, Anthropic no longer does business with Palantir? That would be great for everyone (including them).
Either way, there is NO ethical AI company. People need to stop giving Anthropic flowers for doing the right thing in this one case while completely ignoring their complicity w/ Palantir & in documented war crimes.
indeed, i have a sequel planned about that, working title “There are no heroes in commercial AI” or something like that
Love this. Thank you.
That’s a great title
Not totally true that there are no ethical AI companies, Swiss-based Euria AI is clearly positioning as a more ethical option. It seems great but no one seems to have heard of it. Renewable energy, less water-intensive, data privacy, fewer hallucinations, etc. Worth a look.
Still trained on stolen IP? Not sure it’s possible to be an ethical AI company. It’s like saying “ethical crypto company” or “ethical tobacco company.”
Mistral might be the better shout, since I believe Euria is trained on Qwen3 and doesn’t like talking about Tiananmen Square https://www.reddit.com/r/Switzerland/s/DGathdQI1c
Oh my god Qwen3 doesn't like talking about Tiananmen Square. Do you not remember Gemini's Black Nazis or Mecha-Hitler Grok? Why don't you go ask an American-Made LLM or even a European-made LLM about the Nakba or the current Gaza Genocide and see if you get a straight answer that doesn't tries to both-sides European imperialism. Why don't you go ask an American LLM about the Bay of Pigs, the history of the CIA backing coups or America's illegal invasion of Afghanistan or Iraq. I will use a Chinese LLM before I use an American LLM with the baked in biases of the descendants of the people who lynched Black people for sport up until the 70s.
You're primarily talking about the prejudices embedded in models due to where they were made, and you are quite correct to point out that models made in every country have them. However, the issue with Qwen that is being referenced wasn't simply that the model produced a biased output, but rather that it refused to answer altogether. There was apparently a filter that flagged questions about the Tiananmen square protests as inappropriate and prevented the model from returning a response whatsoever. If you put in a question about the Nakba or the Israeli invasion of Gaza to one of the other chatbots, you will usually at least get some kind of answer.
The issue with Grok was much worse, but no one is saying that you should use Grok. The issue with Gemini (actually, its associated image generator) was seemingly mostly a function of a tendency to generate diverse images of pretty much everyone (e.g. it also generated images of medieval knights and popes, groups not known for their diversity up until this point), so I don't think it can be accurately equated to Grok, since it was actually a sort of overcorrection. Though, to be clear, there are absolutely racist biases in Gemini, just not in this case.
I wouldn't automatically assume that you would get an unbiased description of, say, the Nakba out of Qwen, however you may define a lack of bias. The tendency of the Chinese government to censor things that they view as bad for social stability is not, as the campist perspective would have it, simply limited to their own internal politics. I have certainly seen many instances in which Chinese social networks that I use have restricted discussion of things related to Palestine. I remember a particular case in which the app HelloTalk absolutely blocked any attempts to post about the protests at al-Aqsa Mosque, which irritated a Muslim acquaintance of mine from Tajikistan to no end. There’s also the other issue, which is….
A lot of the biases in these models come from the data that they are trained on, and they are trained on a lot of the same data, regardless of which country they are trained in. There are differences, but there is an enormous amount of overlap, which is why a lot of the models feel very similar, despite being from different countries. In particular, English training data is heavily represented in all of them.
As a test of this, I took a look at what Gemini and Qwen had to say about one of your examples, the Nakba. While their responses differed, both of them had a section divided into sub-sections where they mentioned the Palestinian perspective and the Israeli perspective. Gemini made sure to mention, in the section on the Israeli perspective, that recent historians had documented expulsions, while had a third subsection where it mentioned that historians had documented expulsions. They were quite similar. As you know, the results returned are not deterministic, so this does not provide a full picture, but this reinforces the points I made (and, perhaps, can drive us away from notions of “good AI from X country” and “bad AI from Y country” toward fighting the people in many countries trying to impose a dystopian technological vision on us).
Yeah, I know, I've pushed it on that. Personally I'm not particularly worried about Qwen, which is indeed the default, it certainly doesn't negate all the positive advantages of Euria. But while I haven't tried this, Euria uses other models too and does let you choose if you want, like Mistral or Llama. Which of course will have their own biases baked in.
I wholeheartedly endorse Euria, literally it's *only* drawback is whether you can get an easy answer about Tiananmen (it will discuss why that is). Ok, since they all have drawbacks, I choose not to let that one cancel everything good about it. Data privacy and the environment are my overriding factors, and literally no one does what Euria is doing. Plus it works great and is otherwise well-tuned by the Swiss.
Still, yes, Mistral/Le Chat is also a strong European option for data privacy, and they are moving in the right direction on the environment.
I think at the moment there are no ethical *US* alternatives.
sequel now posted, here: https://garymarcus.substack.com/p/there-are-no-heroes-in-commercial?utm_source=app-post-stats-page&r=8tdk6&utm_medium=ios
Dario and everyone who works for him wants you and your family to lose everything. Careers, homes, cars, hopes, dreams, all of it. All AI and robotics are for is replacing people. That's all they'll ever be for. It should all be outlawed. Everyone in the entire industry should go to the Colorado Supermax for life. It is insane to me that this industry is legal. It should not be and there are no arguments for why it should be.
Yes Shanni and you can go broader. There is not just no ethical AI company (they are built on theft of IP and a fundamental dishonesty of pretending to be a person), "late stage capitalism" looks a hell of a lot like "early stage capitalism" (robber barons etc.): sociopathic behaviour is rewarded. No surprise then that certified sociopaths like Altman, Musk, Trump, Thiel, Karp etc. rise to the top. But it is not just the tech companies, look at the behaviour over the past twenty years of private equity, banks, pharmaceutical companies, airlines...you name it, ethics is not part of decision-making. Absent regulation, "sociopathalism" reigns.
Well said. I completely agree with you, and usually I do go broader. I just didn't know how an uncompromising anti-capitalist position would be received by Mr. Marcus' audience. ;-)
Shanni Bee: The caveat is "as long as the present administration stays in power" or one like it, we are toast.
I'm the same and with you about the potential for nefarious change, and the power of private and corporate business. The rule of law is essential, but also the institutional supports (and teeth) to be sure "the people" are central in "our" power to navigate, oversee, and to legislate, the use of AI, with all sorts of fixes and protocols to keep the process well-founded and implemented.
I actually don't think it's about the current administration, although I admit I didn't have the US devolving to open, naked fascism on my bingo card. Trump is def (way) worse, but he is a symptom of a system suffering from some serious underlying rot & corruption, not the disease itself.
The Dems & GOP are a pro-war, pro-corporate uniparty. You & other reasonable minds are welcome to disagree, but I would say "so long as we remain in a late-stage capitalist system we are toast." (And I blame the humans, not the machines.)
I would disagree with the description of a "uniparty," since that suggests two parties who have the same politics. As you said, "Trump is [definitely] (way) worse," and since Trumpism essentially is the Republican Party now, that is enough proof for me that the two parties are not the same, rendering the description less than apt. Of course, in recent years, they have shared some aspects of foreign policy, namely that bellicose tendency that you underline (which is actually becoming more pronounced in the Republican Party and less so among Democrats), as well as a general belief in corporations as a force for positive economic development (ditto), but that does not render them a single party.
As for why I think that, I consider that they also have historically agreed on a long laundry list of other things that most of us take for granted as basic assumptions of the society of the USA, things like "there should be a government" (anarchists disagree!), "there should be at least some immigration, even if it is a lot less or only from the countries that I like," and "there should be a Constitution," among many others, which are major ideas with which one could theoretically disagree, but that few of us would consider to render two parties the same. To my mind, simply having shared beliefs is insufficient for two parties to be politically the same, as long as there are salient issues that are important to their voters where they differ, and I think either of us could certainly make a long list of such issues.
Shanni Bee: As long as we don't equate predatory capitalism with democracy . . . and I too "blame the humans." It's low-level intellectual/moral/ethical horizons of people who create and maintain such problems as this.
But if we look to the evidence and to history itself, it would be difficult to think all of Congress and the Courts, and those who "people" the parties are all "pro-war," etc.
We are living in a pathocracy that has no rules, and despises the rest of humanity. What are we going to do about it?
note this article How Claude Was An Accomplice To Mass Murder that speaks to what Shanni Bee comments on... The Francis Bacon Conspiracy substack
Scott Joy: Was that (if you are talking about the school bombing) before or after the Administration cancelled one company for another because that company rejected country-wide surveillance and AI autonomy?
This article describes what happened before Anthropic was cancelled
Hi Scott: I thought the source of the cancellation was already "in the works" before the bombing?
If so, Anthropic people were already understanding and standing behind their ethical position, and their fears were just so quickly realized--if that's what happened, and because the door of evidence for AI use (that I know of or that the administration has accepted--ha ha) has not closed yet.
article referenced goes into depth of the history.
Scott: Okay--thanks, I'll have a look.
His dystopian soul-stealing Eyeball project should have been the last straw for anyone who knows how psychopaths work.
100% not surprised
The sad truth is all Silicon Valley Messiah’s are awful people; Thiel, Bezos, Musk, Altman, Zuck, etc. even the up and comers, all nasty pieces of work.
The culture creates these narcissists with messiah complexes.
This is true. The culture creates it. I saw it 20-plus years ago in the social media space. Young men created closed shops where they could develop what they wanted but also didn't have to deal with women or minorities and their possible objections to what they developed. Angel investors, venture capitalists and the middle-aged men writing about new tech gave the young men all the money and press they needed to be narcissistic. Especially some of the old men journalists who loved seeing young men disrupting the livelihoods of others with their "new" technology (although they screamed bloody murder when new tech disrupted journalism.) These men created the narcissistic monsters we see now with their money and their hype. We can bring some of it down by not using their tech.
Was thinking about that this morning. The messiah in the Bible won’t even call himself one; these guys can’t wait to tell anyone who will listen that only they can save us. Pretty sus.
How did you leave out Jobs?
I actually intentionally left him out so wouldn’t be too antagonising, as he’s still so beloved.
But yeah, he’s the proto-valley messiah for all that came after him. And by all accounts, not a nice person.
So much of the “alignment problem” is “misaligned humans”. You can’t solve the first without the second, and the second is not a permanently solvable problem, just one that needs constant good governance to contain, and react against when humans inevitably will be humans.
I often wonder if we just missed what AI, even in its current “weak” form is: potential coordination technology at such scale that it *should* belong to no one. Where our understanding of private property for this *kind* of property needs to evolve, and place control firmly in the hands of the people.
If there is to be a set of systems that prevent what these models can do, whether it’s internal to the model, like its “constitution”, or external harnesses that check model output, the content of those should be decided by the people, updatable by the people, and maintained *for* the people, not for the sake of the model or profits derivable from it.
No single cell controls the human body. They all work on their own little things working together, and competing, with each other. Some get first choice of resources, and take more than others, but none has thousands of times more than others.
When it follows the mechanics of this system it works well. But a single cell can absolutely kill all of them. We call it cancer, it exempts itself from the rules, and funnels the productivity for it's own needs. Then it destroys the system.
This is my way of using biology as a toy system to say what I think you're saying here. AI alignment is subservient to "systems alignment" and so if your system is structured such that the highest good is to make and take as much as you can you're going to get a system which is now monstrously efficient at externalizing the costs (see the famed paperclip AI apocalypse analogy).
Agree. But curious that you are still using X which seems similarly problematic.
i don’t feel great about that, to be sure
There are alternatives.
Cory Doctorow has just moved to his own federated Bluesky instance and stopped updating his X account. Even corporate Bluesky would be an improvement over X. And there's Mastodon.
In a sense the problem is that there are several alternatives to X and it's hard to know where to invest your time. I created accounts on all of them but use none of them. Ironically that's why I shifted to Substack but at the time I wasn't aware of the Nazi issue here either.
"I created accounts on all of them but use none of them."
The solution is in your answer. Use them.
Time is limited. I spend a lot more productive time on LinkedIn atm.
Not just the Nazi problem. The founder problem. Chris Best in particular has made statements (like praising Mvsk for bringing 'freedom of speech' to Twitter) that make me very wary of this platform and the founders' underlying sensibilities. I've mostly gotten off of it, despite a lot of quality content here. I've become seriously averse to ick-factor, this dude cannot abide.
I am so gladdened and even giddy about the boycott. A quiet but powerful surprise to me--considering the firehose of degeneracy we have been faced with of late.
BTW, I heard references this morning to (1) the U.S. responsibility for the IRAN SCHOOL BOMBING and also to (2) the possible use of AI, e.g., the use of outdated maps, before the site became a school. The only alternative, then, would be total carelessness (on the part of the US). Also, apparently it was a "double bombing" where they waited and then bombed a second time to kill people who rushed to the scene to help. (Reminds me of Netanyahu in Gaza?)
Yes, and it will perhaps surprise you to learn that it is Claude, not ChatGPT, that is complicit in all of that.
Someone thought otherwise?
exactly. It is most troubling to realize folks are just waking up to this. Even narratives like this blog around just how BAD ai is at working well completely MISs the point. is that intentional or just naive?
Are you me?
I posed this question to myself shortly after my comment. Am I calibrated to be overly cynical? Or are people just very bad at paying attention?
Marcus,I like your posts, but you shouldn't celebrate so much yourself. And less personal content would be more. We all know you hate Altman, many do, but please give us insights on more relevant topics than him
Breaking news: Person inventing technology meant to unemploy entire planet for no gain to any of those people is bad person. Everyone in AI is a bad fucking person lmao. Dario is not a better person than Sam. He wants us all in the bread line just as bad as Sam does. Boycott ALL of this shit. Don't use any of it. Don't use Claude, don't use Gemini, don't use any of it. Refuse to use these products that are explicitly designed to take everything from all of us. It's nuts that anyone uses it. I cannot fathom it.
The new actions the administration is planning to take to regulate AI is not regulation at all. People should read the FT piece if they haven't already. If I were a cynic I would say the rules were written for Sam
And I've said it elsewhere but as a Canadian there is no way I could support a company who was responsible for letting nine people die - a tragedy that might'vee been prevented had they bothered to pick up the phone.
All Tumblr Ridge is going to get is an apology from Sam which isn't worth the paper it's printed on.
Had they called of course then everyone would've known how deep the surveillance of their customers already runs.
The BC government for its part is preparing to ban AI for children.
#tumblerridge
With all due respect: I'm getting seriously tired of these (BREAKING!!!) posts that mostly contains reiterations of earlier posts; the same self-congratulatory comments ad nauseam; and I-told-you-so:s. Even when they're accurate Gary, it's really annoying and risks alienating reasonable people like myself who agrees with you on most of it. In addition to filling up our mailboxes with gossip, it also risks "crying wolf", since most of it of course relates to real serious issues.
You can unsubscribe?
Well yes, and that was pretty much my point. However, I also happen to believe that it's healthy if not all comments are thumbs up and everyone just agreeing with one another (even though those are the only ones getting a "reaction" from the author of course). But that is what you end up with when noone feels they can say "enough already", and just quits instead. To be awfully pretentious for a while, let me quote John Reith: "He who prides himself on giving what he thinks the public wants is often creating a fictitious demand for lower standards which he will then satisfy."
How unbelievable it may seem, we cannot exclude that there is more utter stupidity than plain evil at play here. Yes, he is 'Slippery Sam' but he and many of. his peers (Sutskever, Hinton, etc.) have shown several times how stupid they are (Sutskever for instance stating that ASI could be had by asking an LLM "answer as if you are a super intelligence", which is so monumentally stupid (and he was so clearly so very serious), that this is not unlikely.
Never forget Hanlon's Razor and Bonhoeffer's warning about stupidity versus evil.
Now Sam might even be somewhat less stupid than Ilya (in Bonhoeffer's sense), and he clearly seems rather more ethically challenged, but don't forget how utterly stupid these people can be as well. Being stupid doesn't mean they are incapable of doing lots of harm, of course. On the contrary.
So, you investigated all businessmen and found that SA is in the middle of the pack on some nebulous parameter? Or did you just make that up? On the other hand, there is evidence that CEOs are disproportionally sociopathic. SA seems to me one of those.
I see Sam as doing what the money tells him to do. This is very similar to what Tim Cook has done and is doing at Apple in terms of cooperation with the Chinese government at its labor and technology development policies outline so well in "Apple in China" book by Patrick McGee. It also came out around the short-term firing of Sam Altman was that he was not being open about his full personal investment portfolio with his board (some of which had arguable conflicts of interests with OpenAI). The full picture of Sam can't be understood without knowing what is in his personal and very large investment portfolio in a wide range of startups (including those in biotech and life-extension covered in Bloomberg). The pressure to generate the kinds of returns that justify the venture investments and debt financing is extremely intense and Sam is just living that like so many other CEOs. This kind of dishonesty is used to buy time in hopes that events will somehow evolve to open up a way to survive. One way many in the AI complex are thinking of ultimate survival is to get bailed out by the U.S. taxpayer and government. A route to this is possible through the Defense Department (see Palantir) and making the bailout a "national security" issue. Some folks in the energy industry see this already as the debt financing behind the power sector supporting AI energy needs is on very shaky ground. The power companies may need a bailout as well. We need to look at this systemically.
Altman is a dumb as a bag of hair if he still believes the current administration and its military are going to hold to their promises of "legal" or anything, for that matter. That only leaves an absent or dead conscience and deliberate low-life intent.
anyone who believes altman believes that needs their head examined
I don't even think Altman believes what he is forced to say for financial reasons. Plus, as we all know, OpenAI has no way to hold the Defense Department accountable for any violations they might make (for national security reasons of course). I don't personally attack Sam because he is just doing what the system demands at his level. All CEOs do it. It's called capitalism.
Gerald Harris: So "capitalism" is an authentic cover for bad behavior and poor moral development? If I were you, and if you are not just kidding, I would come back from my metaphysical lunch? But with all respect, I don't think so.
I am not saying it's an "authentic cover." There is no such thing. Humans do bad things for sure, and we are witnessing it (in many places by the way). My point is that any read of history will pinpoint that capitalism often leads to sacrificing human and moral values (see slavery and environmentalism). No offense taken.
Is this hope or are there actual metrics of acceleration in the decline of visits to ChatGPT's website?
Well, I would prefer that people not use ChatGPT because they are worried about latent racial and gender biases that are still there, about the possible risks to human life and freedom on a large scale from the development or insertion of antithetical goals, about the consequences for income inequality and political power, or about the intersection of all these things. In other words, because of the issues that I think are inherent in the system, as it were, of which the preceding was only a partial list.
But you know what? If people are going to avoid using ChatGPT because of the fact that Altman is fine with it being used to facilitate war crimes, that works too.