Here’s his accompanying twitter post: https://xcancel.com/BernieSanders/status/2095542398084415952 I kind of feel like the post itself, ironically, has an AI-generated feel to it. It kind of...
Literal quote. I think Bernie's great, but this proposed legislation, which seems to be an awareness campaign rather than a genuine attempt at lawmaking, is really misguided. Step 1) Halt US AI...
But what happened next, nobody expected.
Literal quote. I think Bernie's great, but this proposed legislation, which seems to be an awareness campaign rather than a genuine attempt at lawmaking, is really misguided.
Step 1) Halt US AI development, likely crashing stock markets globally, but worse in US markets. Arguably OK, but if the goal is to tank financial markets through legislation, put that in the title. At the very least acknowledge that you understand that's what will happen!
Step 2) Beg the rest of the world to stop too. Which they absolutely won't do, pure fantasy.
Step 3) Watch from the sidelines as China becomes the leader in AI.
I'm not saying that US tech bros are the right shepherds of this technology, but I don't think China is going to be any better.
Bernie's populist message resonates with a lot of people, I agree wholeheartedly with his core platform. I don't understand the anti-AI pivot. Using AI as a proxy for wealth inequality isn't going to make any progress on the real problems.
I'm pretty tired of the "But China is going to make the torment nexus" argument. How about for once we let someone else screw up instead of insisting that the US needs to be the leading screw-up...
Step 3) Watch from the sidelines as China becomes the leader in AI.
I'm pretty tired of the "But China is going to make the torment nexus" argument. How about for once we let someone else screw up instead of insisting that the US needs to be the leading screw-up of the world?
China is already beating us IMO. Because their approach to AI is much more sustainable and future proof than this multi-trillion dollar rollout of data centers. in 20 years, well after this bubble pops, the most important AI models will be the ones that can run on your phone, not cloud into some data center that hikes up energy costs for everyone in the area.
Not to mention that this current approach to data centers only severs to enrich a very select few people in the country, not the country as a whole. Believe it or not, we don't need a billionaire to compete in technology.
I don't understand the anti-AI pivot.
Data center rollouts are one of the very few bi-partisan issues right now in this otherwise very divided country. the pivot makes perfect sense, and in one way or another we are going to deal with the fallout of this huge gamble in a few years. I completely understand why Bernie is getting on this now.
To be fair, the stock market is gonna crash anyway once the circular money vortex picks up a metaphorical bull and chucks it through the largest metaphorical antique shop in the world. And...
To be fair, the stock market is gonna crash anyway once the circular money vortex picks up a metaphorical bull and chucks it through the largest metaphorical antique shop in the world.
And something about "starting from an unreasonable position and compromising to reasonable" being better than "starting reasonable and then compromising to worthlessness."
See also: ACA. In the long term, it basically just got a ban on preexisting conditions. Which was good, but also did not solve the problem it set out to.
Especially since IME AI videos are noticeable for long, uncut segments without errors. I got sent a video about cuts of meat by what turned out to be a fake AI butcher character. It was his lack...
Especially since IME AI videos are noticeable for long, uncut segments without errors. I got sent a video about cuts of meat by what turned out to be a fake AI butcher character. It was his lack of stumbling without frequent cuts that made it as noticeable as it was
Hard to say. Bernie likely has professional writers and AI was ultimately trained on this exact kind of rhetoric. This newpaper-y slightly click-baity feel to reel people in.
Don’t have any evidence to make that claim concretely though.
Hard to say. Bernie likely has professional writers and AI was ultimately trained on this exact kind of rhetoric. This newpaper-y slightly click-baity feel to reel people in.
Doesnt this look a bit amateurish for two long term professionals? Like, Bernie Sanders has been an elected official for like 50 years. If anyone is qualified to take an actual stab at writing...
Doesnt this look a bit amateurish for two long term professionals?
Like, Bernie Sanders has been an elected official for like 50 years. If anyone is qualified to take an actual stab at writing real legislation it would be him.
If this document with this wording actually got passed into law, would you be satisfied with that?
Granted, thats probably not the point here, this is just a statement piece thats probably not intended to go anywhere. But maybe it should be? These guys arent random activists who need to enact change by shouting into the ether, they are directly in a position to be doing the things they want to see done.
Bernie’s got his heart in the right place and always has. But recently his videos make me think he’s been duped a bit into thinking we’re about to see Terminator come true. Edit: I have recently...
Bernie’s got his heart in the right place and always has. But recently his videos make me think he’s been duped a bit into thinking we’re about to see Terminator come true.
Edit: I have recently gotten a feeling that in 12 months we’ll have AI services that can replace many people working office jobs. We need to get ahead of that shitshow. Not super intelligence. Mere general intelligence.
We're already there. I know the common take on recent mass layoffs is that AI is just an excuse to do them, and that companies would have done them anyway. Some of that is certainly true, but AI...
I have recently gotten a feeling that in 12 months we’ll have AI services that can replace many people working office jobs.
We're already there. I know the common take on recent mass layoffs is that AI is just an excuse to do them, and that companies would have done them anyway. Some of that is certainly true, but AI tools have also greatly increased productivity in certain areas, reducing the need for labor. This is most obvious in the tech sector, but it's also very prevalent in basically any industry that requires dealing with large volumes of language, whether that language is formal, like programming, or natural, like copywriting, paralegal work, marketing, and sales.
Its accelerating, but LLMs have been public, and useful for four years now. The economy's transformation has been well underway for a while, and the average worker is not coming out ahead.
That said, theres no legislation that a single country could pass to curtail that transformation. The cat is out of the bag. The only way to control it is multilateral treaties that the major industrial powers are signatories to, along with inspections, penalties, and so on. Similar to nuclear non-proliferation. The only thing that any one country can do alone is enact polices that dampen the negative impacts of the transformation on the average citizen. They can't meaningfully stop or delay the transformation alone.
No. People who do not work on the ground fell for the snake oil that AI tools are massively increasing productivity, and are reacting in kind because it feeds into their justifications for actions...
AI tools have also greatly increased productivity in certain areas, reducing the need for labor.
No. People who do not work on the ground fell for the snake oil that AI tools are massively increasing productivity, and are reacting in kind because it feeds into their justifications for actions they always wanted to do anyway. All formal studies on AI productivity maxes out at 10% productivity gains. Many seem to place it more conservatively around 3%.
That's "I can make emails faster" levels of productivity, not "I'm ready for an non-human white collar future" levels. We come around to this argument every single year and then I read new studies and it's creeping alone logarithmically at most. Not exponentially like what the headlines want us to believe.
The economy's transformation has been well underway for a while, and the average worker is not coming out ahead.
Well it's too bad the average worker is the one who has to vote on such issues, huh?
The only thing that any one country can do alone is enact polices that dampen the negative impacts of the transformation on the average citizen.
Governments regulate companies, not the other way around. The governments can 100,000% entail stronger worker protections, if not outright ban mass scale AI if they wished. Don't let billionaire propaganda tell you otherwise.
This entirely hinges on your thesis of AI not actually being that useful, and companies being sold a complete lie. That's increasingly becoming harder of a position to defend as the tools get more...
Governments regulate companies, not the other way around. The governments can 100,000% entail stronger worker protections, if not outright ban mass scale AI if they wished.
This entirely hinges on your thesis of AI not actually being that useful, and companies being sold a complete lie. That's increasingly becoming harder of a position to defend as the tools get more capable though. I've seen studies similar to what you were referencing, and most of the ones that I've seen have been early in the technologies infancy. Companies have continually gotten better at implementing and are becoming less reckless with their token spend as they've stared dialing things in.
There's no universe where AI somehow gets less productive.
So if it's a case that we see even slow growth in AI productivity, no country can unilaterally ban AI outright, because they'll get completely outcompeted on a global market.
That's the whole issue at stake here. Its a Mexican standoff. Even if a country would like to protect their labor force, they're worried that putting measures in place that are too aggressive will cause that work to go elsewhere, and they'll be left behind. That's why we still haven't seen significant anti AI legislation anywhere, despite us thinking about these problems for the past four years.
You really need international agreements in place to assuage those fears.
Not at all. Governments put bans/embargoes on useful things more often than you'd expect. This is more talking about possibility than viability. But back to reality: Yes. And those will be the...
This entirely hinges on your thesis of AI not actually being that useful
Not at all. Governments put bans/embargoes on useful things more often than you'd expect. This is more talking about possibility than viability.
But back to reality:
Companies have continually gotten better at implementing and are becoming less reckless with their token spend as they've stared dialing things in.
Yes. And those will be the companies to survive, maybe thrive post-bubble. As of now, the largest contributors haven't really slowed down.
Even if a country would like to protect their labor force, they're worried that putting measures in place that are too aggressive will cause that work to go elsewhere, and they'll be left behind.
America has given away decades of labor and technology, hence why we became so behind on very necessary resources that could have mitigated the current bubble. So I don't really think that's a justification the US truly believes in.
You really need international agreements in place to assuage those fears.
Only if you 1) lack other kinds of leverage and 2) genuinely think you will be left behind. I don't think the US has neither. And that's really what all this comes down to: even if this was genuinely a transformative technology, the ones heading the charge do not have the interests the country nor humanity in mind.
Ed Zitron pointed out in one recent article how modern AI use is a lot like people tricking out their terminals. Spending endless time 'optimizing their workflow' and spending 4 hours to automate...
Ed Zitron pointed out in one recent article how modern AI use is a lot like people tricking out their terminals.
Spending endless time 'optimizing their workflow' and spending 4 hours to automate something that takes 5 minutes once a year. Building gigantic Rube Goldberg machines that have an amorphous black blob in the middle. Like Kubernetes with less documentation.
You have to sit there and verify almost all of the output yourself. With enough expertise to know when it goes wrong.
Related: Just the other day, Opus 5 on Max, it silently converted 7005 to 7505 while doing some SQL queries, and completely derailed itself. To a commoner that was using this, it would have gone completely unnoticed, and then subsequently would have tried having Claude fix a hallucinated problem and probably murder something else in the process.
Well clearly we don't "have" to. That assumes companies or shareholders care about the quality of the output. If they are fine with garbage out and are willing to ignore customers/stakeholders (or...
You have to sit there and verify almost all of the output yourself. With enough expertise to know when it goes wrong.
Well clearly we don't "have" to. That assumes companies or shareholders care about the quality of the output.
If they are fine with garbage out and are willing to ignore customers/stakeholders (or trick them) into consuming slop, then the slop factory wins out. The most profitable model requiring the lowest (eventually) expenses. Will that prevail? Doesn't look good so far for slop, but I wouldn't quite say they lost yet.
Worst case, the rich get away with the big con instead of the long con.
I mean, pretty much any sizable org can make cuts to IT and be fine in the short term. Otherwise no IT staff could ever take vacation. It's only once the tech debt starts spiraling out of control...
I mean, pretty much any sizable org can make cuts to IT and be fine in the short term. Otherwise no IT staff could ever take vacation. It's only once the tech debt starts spiraling out of control that it becomes a crisis.
Of course everyone remaining will mask how well they are/aren't struggling to get everything done with a reasonable work/life balance until the layoff window is passed.
And, as my boss put it "the Claude bill is currently annoying, but not yet problematic."
There will be a tipping point where all but the most irresponsible companies will notice that it's cheaper and more effective to pay for 3 junior staffers than pay that Claude bill.
The funny thing is that's basically just the Silicon Valley ouroboros coming around the cycle again. Once it was capable microcomputers disrupting mainframe companies, then Internet companies more...
There will be a tipping point where all but the most irresponsible companies will notice that it's cheaper and more effective to pay for 3 junior staffers than pay that Claude bill.
The funny thing is that's basically just the Silicon Valley ouroboros coming around the cycle again. Once it was capable microcomputers disrupting mainframe companies, then Internet companies more efficiently eating industries that relied heavily on face to face dealings by automating the low hanging fruit.
Now we're in something like the dotcom bubble, only Pets.com is trying really hard to convince every reasonably stable company that their doomsday weapon will destroy their business if they don't get on board and give them one million billion dollars (taps face with pinky finger).
Next up, some enterprising startups will determine they can do things the companies that got duped do, saving money by paying a few people instead of the billion dollar rent. +/- robust and reliable infrastructure after some high profile slop piles cause something nasty.
I'm talking about you literally just upload the same instructions you give to your human employees and software will spin up a Windows VM and click around and do exactly the same work faster and...
I'm talking about you literally just upload the same instructions you give to your human employees and software will spin up a Windows VM and click around and do exactly the same work faster and cheaper. Not just certain parts of a job or a few fields.
Honestly, I don't really see that ever becoming a thing at scale, just because it's not a very efficient way to spend tokens or compute power, and it's not really how we've seen technology develop...
Honestly, I don't really see that ever becoming a thing at scale, just because it's not a very efficient way to spend tokens or compute power, and it's not really how we've seen technology develop in the past. Like, when tractors were invented, they weren't shaped like mules and slotted into a yoke with a person driving them.
I suspect what work looks like in the future looks similar to the evolution you see today. Instead of having, say, 5 engineers on a team, you have one or two, and those engineers just prompt most of the time. Instead of having 5 layers of management, you flatten it out and have maybe 2, because you don't need as many people to do the same work, so you don't need as many people to manage them.
Headcounts shrink, but people don't necessarily get 1:1 replaced with an AI agent, and the people running the agents are doing different kinds of work than they once were.
I doubt that's the real document being proposed. it looks more like a summary for proper legislation to be written up if this ever gets more support. I can imagine they never try to sway congress...
I doubt that's the real document being proposed. it looks more like a summary for proper legislation to be written up if this ever gets more support. I can imagine they never try to sway congress members with a massive 20+ page documentation until it's near the support needed to potentially pass a vote.
Bernie discovering the Dutch method of writing a proposal meant to be shared on social media, but doomed to be unsuccessful in every other way.
Here’s his accompanying twitter post: https://xcancel.com/BernieSanders/status/2095542398084415952
I kind of feel like the post itself, ironically, has an AI-generated feel to it. It kind of reminds me of those LinkedIn posts
That talk.
Just like this.
Don’t have any evidence to make that claim concretely though.
Literal quote. I think Bernie's great, but this proposed legislation, which seems to be an awareness campaign rather than a genuine attempt at lawmaking, is really misguided.
Step 1) Halt US AI development, likely crashing stock markets globally, but worse in US markets. Arguably OK, but if the goal is to tank financial markets through legislation, put that in the title. At the very least acknowledge that you understand that's what will happen!
Step 2) Beg the rest of the world to stop too. Which they absolutely won't do, pure fantasy.
Step 3) Watch from the sidelines as China becomes the leader in AI.
I'm not saying that US tech bros are the right shepherds of this technology, but I don't think China is going to be any better.
Bernie's populist message resonates with a lot of people, I agree wholeheartedly with his core platform. I don't understand the anti-AI pivot. Using AI as a proxy for wealth inequality isn't going to make any progress on the real problems.
I'm pretty tired of the "But China is going to make the torment nexus" argument. How about for once we let someone else screw up instead of insisting that the US needs to be the leading screw-up of the world?
China is already beating us IMO. Because their approach to AI is much more sustainable and future proof than this multi-trillion dollar rollout of data centers. in 20 years, well after this bubble pops, the most important AI models will be the ones that can run on your phone, not cloud into some data center that hikes up energy costs for everyone in the area.
Not to mention that this current approach to data centers only severs to enrich a very select few people in the country, not the country as a whole. Believe it or not, we don't need a billionaire to compete in technology.
Data center rollouts are one of the very few bi-partisan issues right now in this otherwise very divided country. the pivot makes perfect sense, and in one way or another we are going to deal with the fallout of this huge gamble in a few years. I completely understand why Bernie is getting on this now.
To be fair, the stock market is gonna crash anyway once the circular money vortex picks up a metaphorical bull and chucks it through the largest metaphorical antique shop in the world.
And something about "starting from an unreasonable position and compromising to reasonable" being better than "starting reasonable and then compromising to worthlessness."
See also: ACA. In the long term, it basically just got a ban on preexisting conditions. Which was good, but also did not solve the problem it set out to.
That's how he talks though. All of his speeches have this cadence
Especially since IME AI videos are noticeable for long, uncut segments without errors. I got sent a video about cuts of meat by what turned out to be a fake AI butcher character. It was his lack of stumbling without frequent cuts that made it as noticeable as it was
Hard to say. Bernie likely has professional writers and AI was ultimately trained on this exact kind of rhetoric. This newpaper-y slightly click-baity feel to reel people in.
Doesnt this look a bit amateurish for two long term professionals?
Like, Bernie Sanders has been an elected official for like 50 years. If anyone is qualified to take an actual stab at writing real legislation it would be him.
If this document with this wording actually got passed into law, would you be satisfied with that?
Granted, thats probably not the point here, this is just a statement piece thats probably not intended to go anywhere. But maybe it should be? These guys arent random activists who need to enact change by shouting into the ether, they are directly in a position to be doing the things they want to see done.
Bernie’s got his heart in the right place and always has. But recently his videos make me think he’s been duped a bit into thinking we’re about to see Terminator come true.
Edit: I have recently gotten a feeling that in 12 months we’ll have AI services that can replace many people working office jobs. We need to get ahead of that shitshow. Not super intelligence. Mere general intelligence.
We're already there. I know the common take on recent mass layoffs is that AI is just an excuse to do them, and that companies would have done them anyway. Some of that is certainly true, but AI tools have also greatly increased productivity in certain areas, reducing the need for labor. This is most obvious in the tech sector, but it's also very prevalent in basically any industry that requires dealing with large volumes of language, whether that language is formal, like programming, or natural, like copywriting, paralegal work, marketing, and sales.
Its accelerating, but LLMs have been public, and useful for four years now. The economy's transformation has been well underway for a while, and the average worker is not coming out ahead.
That said, theres no legislation that a single country could pass to curtail that transformation. The cat is out of the bag. The only way to control it is multilateral treaties that the major industrial powers are signatories to, along with inspections, penalties, and so on. Similar to nuclear non-proliferation. The only thing that any one country can do alone is enact polices that dampen the negative impacts of the transformation on the average citizen. They can't meaningfully stop or delay the transformation alone.
No. People who do not work on the ground fell for the snake oil that AI tools are massively increasing productivity, and are reacting in kind because it feeds into their justifications for actions they always wanted to do anyway. All formal studies on AI productivity maxes out at 10% productivity gains. Many seem to place it more conservatively around 3%.
That's "I can make emails faster" levels of productivity, not "I'm ready for an non-human white collar future" levels. We come around to this argument every single year and then I read new studies and it's creeping alone logarithmically at most. Not exponentially like what the headlines want us to believe.
Well it's too bad the average worker is the one who has to vote on such issues, huh?
Governments regulate companies, not the other way around. The governments can 100,000% entail stronger worker protections, if not outright ban mass scale AI if they wished. Don't let billionaire propaganda tell you otherwise.
This entirely hinges on your thesis of AI not actually being that useful, and companies being sold a complete lie. That's increasingly becoming harder of a position to defend as the tools get more capable though. I've seen studies similar to what you were referencing, and most of the ones that I've seen have been early in the technologies infancy. Companies have continually gotten better at implementing and are becoming less reckless with their token spend as they've stared dialing things in.
There's no universe where AI somehow gets less productive.
So if it's a case that we see even slow growth in AI productivity, no country can unilaterally ban AI outright, because they'll get completely outcompeted on a global market.
That's the whole issue at stake here. Its a Mexican standoff. Even if a country would like to protect their labor force, they're worried that putting measures in place that are too aggressive will cause that work to go elsewhere, and they'll be left behind. That's why we still haven't seen significant anti AI legislation anywhere, despite us thinking about these problems for the past four years.
You really need international agreements in place to assuage those fears.
Not at all. Governments put bans/embargoes on useful things more often than you'd expect. This is more talking about possibility than viability.
But back to reality:
Yes. And those will be the companies to survive, maybe thrive post-bubble. As of now, the largest contributors haven't really slowed down.
America has given away decades of labor and technology, hence why we became so behind on very necessary resources that could have mitigated the current bubble. So I don't really think that's a justification the US truly believes in.
Only if you 1) lack other kinds of leverage and 2) genuinely think you will be left behind. I don't think the US has neither. And that's really what all this comes down to: even if this was genuinely a transformative technology, the ones heading the charge do not have the interests the country nor humanity in mind.
Ed Zitron pointed out in one recent article how modern AI use is a lot like people tricking out their terminals.
Spending endless time 'optimizing their workflow' and spending 4 hours to automate something that takes 5 minutes once a year. Building gigantic Rube Goldberg machines that have an amorphous black blob in the middle. Like Kubernetes with less documentation.
You have to sit there and verify almost all of the output yourself. With enough expertise to know when it goes wrong.
Related: Just the other day, Opus 5 on Max, it silently converted 7005 to 7505 while doing some SQL queries, and completely derailed itself. To a commoner that was using this, it would have gone completely unnoticed, and then subsequently would have tried having Claude fix a hallucinated problem and probably murder something else in the process.
Well clearly we don't "have" to. That assumes companies or shareholders care about the quality of the output.
If they are fine with garbage out and are willing to ignore customers/stakeholders (or trick them) into consuming slop, then the slop factory wins out. The most profitable model requiring the lowest (eventually) expenses. Will that prevail? Doesn't look good so far for slop, but I wouldn't quite say they lost yet.
Worst case, the rich get away with the big con instead of the long con.
Yes, this is why antitrust is important.
If we can't do anything to actually foster competition, then nobody has a reason to seek out quality.
I mean, pretty much any sizable org can make cuts to IT and be fine in the short term. Otherwise no IT staff could ever take vacation. It's only once the tech debt starts spiraling out of control that it becomes a crisis.
Of course everyone remaining will mask how well they are/aren't struggling to get everything done with a reasonable work/life balance until the layoff window is passed.
And, as my boss put it "the Claude bill is currently annoying, but not yet problematic."
There will be a tipping point where all but the most irresponsible companies will notice that it's cheaper and more effective to pay for 3 junior staffers than pay that Claude bill.
The funny thing is that's basically just the Silicon Valley ouroboros coming around the cycle again. Once it was capable microcomputers disrupting mainframe companies, then Internet companies more efficiently eating industries that relied heavily on face to face dealings by automating the low hanging fruit.
Now we're in something like the dotcom bubble, only Pets.com is trying really hard to convince every reasonably stable company that their doomsday weapon will destroy their business if they don't get on board and give them one
millionbillion dollars (taps face with pinky finger).Next up, some enterprising startups will determine they can do things the companies that got duped do, saving money by paying a few people instead of the billion dollar rent. +/- robust and reliable infrastructure after some high profile slop piles cause something nasty.
I'm talking about you literally just upload the same instructions you give to your human employees and software will spin up a Windows VM and click around and do exactly the same work faster and cheaper. Not just certain parts of a job or a few fields.
Honestly, I don't really see that ever becoming a thing at scale, just because it's not a very efficient way to spend tokens or compute power, and it's not really how we've seen technology develop in the past. Like, when tractors were invented, they weren't shaped like mules and slotted into a yoke with a person driving them.
I suspect what work looks like in the future looks similar to the evolution you see today. Instead of having, say, 5 engineers on a team, you have one or two, and those engineers just prompt most of the time. Instead of having 5 layers of management, you flatten it out and have maybe 2, because you don't need as many people to do the same work, so you don't need as many people to manage them.
Headcounts shrink, but people don't necessarily get 1:1 replaced with an AI agent, and the people running the agents are doing different kinds of work than they once were.
I doubt that's the real document being proposed. it looks more like a summary for proper legislation to be written up if this ever gets more support. I can imagine they never try to sway congress members with a massive 20+ page documentation until it's near the support needed to potentially pass a vote.