Hi Jack! This is slightly random, but I'm a new reader of your blog, and I just wanted to say that I love your writing. The "Parable of the Drowned" was haunting, lonely, and yet felt quintessentially human. If you ever write a book (without ever mentioning AI), I would read it just for the prose
thanks so much for this amazing praise! I have written several hundred stories for the newsletter at this point (there's been one in each issue for the past few years) and I do have ambitions to collect the ones I especially like into a collection at some point. Of course, AI and my work relating to it keeps getting in the way. Some people read the newsletter just for the stories and others read it just for the AI stuff and some read it for both : )
Hi Jack I also really enjoyed the Google doc discussion. I was wondering about how it was implemented? We're there any rules about order of contributions or anything else? Were you allowed to go back and edit?
Regulation shouldn’t be an afterthought or crisis response. The Red Queen effect is real: as models and capabilities sprint ahead, policymakers and safety frameworks need to run just as fast
I love the Google Doc Roundtable format. One point came up though that constantly comes up when talking about any version of a "post-work-world"...
Chris: "Matt is right that the “generate and then redistribute surpluses” approach has never really been implemented, but I don’t see how it could ever work anyway. It entirely ignores the role that productive work plays in a person’s sense of dignity and in a society’s civic health."
...which, in turn, entirely ignores that *looks around on the internet at gathering spaces like X, Reddit, Imgur, Facebook, really basically *gestures vaguely all around**... great majority of people (of all skill levels and points in their career...) who are *productive* but also hate it, hate the thing that they're good at, hate the people they are productive for, hate the company mission for which they are productive, but are also locked in to sunk costs.
Man, how much would I love being able to claw back into my own life, minute for minute, the *time* I save my company through productivity gains from using advanced tools. Every US W-2 (and globally similar structured) employee *sells the minutes of their very lives* for some fixed rate of compensation with no regard to how productive each of those minutes is.
Absolutely! Generate all those surpluses, and then to ahead and redistribute them - and see what people do when they're given the freedom to be productive *on their own terms*. Not because they need to spend 2080 hours per year driving a forklift or writing some code or manning the Self-Checkout lane at the store or reading x-rays of broken fingers or whatever they're doing because that's what they *need* to do because their tweens need braces in 2027 or university tuition in 2031.
Maybe if you redistribute those surpluses and let people decide where to direct their own productivity, we'll reach something resembling a society that values art and creativity and science. I'd trade Grooming Agile/Scrum Stories for Becoming a World-Class Luthier in a heartbeat. But - without those redistributed surpluses, the core reality is that one of those things pays way better and has great health insurance and is generally something that people will pay for - but if I automated away 100% of my job, but kept my Late Career IT Guy Salary through productivity gains surplus redistribution - I could do that other thing, instead - rather than continuing to do something I've learned to be good at, but do not love at all, just to stay out of abject poverty as the alternative.
"Why this matters - for AI to go well, we need AI to police AI: AI systems are on a trajectory to think better and faster than humans. Along with this, AI systems are going to take many, many, many consequential actions, often at such a rate that no human or team of humans could hope to analyze each action." What is the crux of constitutional AI working in the long term?
Hi Jack! This is slightly random, but I'm a new reader of your blog, and I just wanted to say that I love your writing. The "Parable of the Drowned" was haunting, lonely, and yet felt quintessentially human. If you ever write a book (without ever mentioning AI), I would read it just for the prose
thanks so much for this amazing praise! I have written several hundred stories for the newsletter at this point (there's been one in each issue for the past few years) and I do have ambitions to collect the ones I especially like into a collection at some point. Of course, AI and my work relating to it keeps getting in the way. Some people read the newsletter just for the stories and others read it just for the AI stuff and some read it for both : )
I am excited to read your collection when that happens! I am sure you will find the time when the time is right :) until then, your newsletter it is
Hi Jack I also really enjoyed the Google doc discussion. I was wondering about how it was implemented? We're there any rules about order of contributions or anything else? Were you allowed to go back and edit?
The safety research coverage is important. More people should be paying attention to this work.
I really enjoyed "The Parable of the Drowned", sorrowful imagery, but also not despairing.
Regulation shouldn’t be an afterthought or crisis response. The Red Queen effect is real: as models and capabilities sprint ahead, policymakers and safety frameworks need to run just as fast
Exceptional selection, thank you!
I love the Google Doc Roundtable format. One point came up though that constantly comes up when talking about any version of a "post-work-world"...
Chris: "Matt is right that the “generate and then redistribute surpluses” approach has never really been implemented, but I don’t see how it could ever work anyway. It entirely ignores the role that productive work plays in a person’s sense of dignity and in a society’s civic health."
...which, in turn, entirely ignores that *looks around on the internet at gathering spaces like X, Reddit, Imgur, Facebook, really basically *gestures vaguely all around**... great majority of people (of all skill levels and points in their career...) who are *productive* but also hate it, hate the thing that they're good at, hate the people they are productive for, hate the company mission for which they are productive, but are also locked in to sunk costs.
Man, how much would I love being able to claw back into my own life, minute for minute, the *time* I save my company through productivity gains from using advanced tools. Every US W-2 (and globally similar structured) employee *sells the minutes of their very lives* for some fixed rate of compensation with no regard to how productive each of those minutes is.
Absolutely! Generate all those surpluses, and then to ahead and redistribute them - and see what people do when they're given the freedom to be productive *on their own terms*. Not because they need to spend 2080 hours per year driving a forklift or writing some code or manning the Self-Checkout lane at the store or reading x-rays of broken fingers or whatever they're doing because that's what they *need* to do because their tweens need braces in 2027 or university tuition in 2031.
Maybe if you redistribute those surpluses and let people decide where to direct their own productivity, we'll reach something resembling a society that values art and creativity and science. I'd trade Grooming Agile/Scrum Stories for Becoming a World-Class Luthier in a heartbeat. But - without those redistributed surpluses, the core reality is that one of those things pays way better and has great health insurance and is generally something that people will pay for - but if I automated away 100% of my job, but kept my Late Career IT Guy Salary through productivity gains surplus redistribution - I could do that other thing, instead - rather than continuing to do something I've learned to be good at, but do not love at all, just to stay out of abject poverty as the alternative.
"Why this matters - for AI to go well, we need AI to police AI: AI systems are on a trajectory to think better and faster than humans. Along with this, AI systems are going to take many, many, many consequential actions, often at such a rate that no human or team of humans could hope to analyze each action." What is the crux of constitutional AI working in the long term?