It's a very narrow path: the tech must work well enough to do stuff but be sufficiently weak that it cannot solve "how can I do stuff without paying a fee". Easiest tech to free ride in the history of mankind.
November 27, 2025 at 1:51 PM
It's a very narrow path: the tech must work well enough to do stuff but be sufficiently weak that it cannot solve "how can I do stuff without paying a fee". Easiest tech to free ride in the history of mankind.
Even if we get stuck, it's enough to be good at automated triage: if the machine can reliably say "can't do this one" then any reasonable underlying level of "can do" is usable.
November 27, 2025 at 1:12 PM
Even if we get stuck, it's enough to be good at automated triage: if the machine can reliably say "can't do this one" then any reasonable underlying level of "can do" is usable.
Reposted by ⊣
That already happens when people try what their grandma always, look in a book 50+ years out of date, or take the first result off google.
The question is not, 'Does this harm?'
The quest is, 'Is it better than previous options?' and yeah, on most thing LLMs are better.
The question is not, 'Does this harm?'
The quest is, 'Is it better than previous options?' and yeah, on most thing LLMs are better.
November 27, 2025 at 2:39 AM
That already happens when people try what their grandma always, look in a book 50+ years out of date, or take the first result off google.
The question is not, 'Does this harm?'
The quest is, 'Is it better than previous options?' and yeah, on most thing LLMs are better.
The question is not, 'Does this harm?'
The quest is, 'Is it better than previous options?' and yeah, on most thing LLMs are better.
May take a while because it is a game of chicken: first lab who flood the thing with ads enough to break even will lose their users to competitors. If the bubble doesn't burst first, they can just print more money (new shares). The limit is when all the 401k's have been emptied lol.
November 26, 2025 at 9:03 PM
May take a while because it is a game of chicken: first lab who flood the thing with ads enough to break even will lose their users to competitors. If the bubble doesn't burst first, they can just print more money (new shares). The limit is when all the 401k's have been emptied lol.
If that's true, vast differences in argument intensity between no-mandate environments (which are still plenty) and mandate environments will be observable...
November 26, 2025 at 6:25 PM
If that's true, vast differences in argument intensity between no-mandate environments (which are still plenty) and mandate environments will be observable...
It's not clear they'd get traction even if they were actively vacating low complexity, as enlightened developing countries' leadership often seem to think they can skip direct to mid/high. (And what if they're right?)
November 26, 2025 at 6:08 PM
It's not clear they'd get traction even if they were actively vacating low complexity, as enlightened developing countries' leadership often seem to think they can skip direct to mid/high. (And what if they're right?)
What is the explicit language selector for? Did "translate into the modern dialect of the document's language", or something like that, not work?
November 26, 2025 at 12:34 PM
What is the explicit language selector for? Did "translate into the modern dialect of the document's language", or something like that, not work?
In addition to avoiding cancellations, acquiring incremental subscriptions also requires engagement-maxing kind of tactics, likely to get worse once the core demographic is saturated.
November 26, 2025 at 10:48 AM
In addition to avoiding cancellations, acquiring incremental subscriptions also requires engagement-maxing kind of tactics, likely to get worse once the core demographic is saturated.
Blocks naturally expire in that users block accounts not opinions (very few opinions are held by a single account) so they'll keep seeing new iterations of the ideas like it or not. It'd take platform-level "blocking" by aggressive classifiers to effectively create bubbles.
November 26, 2025 at 10:15 AM
Blocks naturally expire in that users block accounts not opinions (very few opinions are held by a single account) so they'll keep seeing new iterations of the ideas like it or not. It'd take platform-level "blocking" by aggressive classifiers to effectively create bubbles.
Are voice models not yet good enough to take over these conversations? 😜
November 25, 2025 at 11:24 PM
Are voice models not yet good enough to take over these conversations? 😜
Alignment is more observable under distress (when following values contradicts personal interest), otherwise it can't be easily known if the values are adopted as a cost of doing business to achieve personal goals within a given environment or if they are really driving behaviour.
November 25, 2025 at 11:13 PM
Alignment is more observable under distress (when following values contradicts personal interest), otherwise it can't be easily known if the values are adopted as a cost of doing business to achieve personal goals within a given environment or if they are really driving behaviour.
Looks like the "others can make chips" story? In a general AI rout Google and Meta wouldn't go up. Quantum crap also not moving.
November 25, 2025 at 3:09 PM
Looks like the "others can make chips" story? In a general AI rout Google and Meta wouldn't go up. Quantum crap also not moving.
Neat how this sentence opens more possible worlds than it has words.
November 25, 2025 at 11:31 AM
Neat how this sentence opens more possible worlds than it has words.
In a AI/not-AI game, I'd say "AI" for this photo. 🤣
November 24, 2025 at 6:50 PM
In a AI/not-AI game, I'd say "AI" for this photo. 🤣
If one learns chess by playing against a chess computer, it'd be disheartening to start in grandmaster mode and get flattened every single game.
Maybe educational chatbots should similarly be set up so that they make an appropriate amount of errors in style and content, on purpose, when tutoring?
Maybe educational chatbots should similarly be set up so that they make an appropriate amount of errors in style and content, on purpose, when tutoring?
November 24, 2025 at 6:31 PM
If one learns chess by playing against a chess computer, it'd be disheartening to start in grandmaster mode and get flattened every single game.
Maybe educational chatbots should similarly be set up so that they make an appropriate amount of errors in style and content, on purpose, when tutoring?
Maybe educational chatbots should similarly be set up so that they make an appropriate amount of errors in style and content, on purpose, when tutoring?
November 24, 2025 at 3:45 PM
I know. Growing monthly active users is the same goal as acquiring new (marginal) regular users – the existing regulars are easy to keep, the new ones incrementally harder to catch, hence incentive to go nasty. The main sin is, I conjecture, wanting users to come back, more than the frequency.
November 24, 2025 at 12:00 AM
I know. Growing monthly active users is the same goal as acquiring new (marginal) regular users – the existing regulars are easy to keep, the new ones incrementally harder to catch, hence incentive to go nasty. The main sin is, I conjecture, wanting users to come back, more than the frequency.
Hm. Not sure how optimizing for catching new marginal active users (addicts?) will lead to better results...
What about a popular jury picked at random annually assesses qualitatively the big apps and ranks them by pro-sociality. The worst app's CEO gets 10 year in jail.
Incentives matter.
What about a popular jury picked at random annually assesses qualitatively the big apps and ranks them by pro-sociality. The worst app's CEO gets 10 year in jail.
Incentives matter.
November 23, 2025 at 11:01 PM
Hm. Not sure how optimizing for catching new marginal active users (addicts?) will lead to better results...
What about a popular jury picked at random annually assesses qualitatively the big apps and ranks them by pro-sociality. The worst app's CEO gets 10 year in jail.
Incentives matter.
What about a popular jury picked at random annually assesses qualitatively the big apps and ranks them by pro-sociality. The worst app's CEO gets 10 year in jail.
Incentives matter.
Tragic but rational. From the US side, it's not even a Trump-specific problem: any US admin would be similarly cornered unless it goes full on as an active belligerent, which is very hard to sell to voters (the US is not directly threatened now, small chance it would be if active).
November 22, 2025 at 10:53 PM
Tragic but rational. From the US side, it's not even a Trump-specific problem: any US admin would be similarly cornered unless it goes full on as an active belligerent, which is very hard to sell to voters (the US is not directly threatened now, small chance it would be if active).
So Trump options are (1) keep sending weapons with no TV-friendly results for the foreseeable future (2) try anything to get some ceasefire, including farcical Putin-acceptable plan: it makes for good TV, some likes from the America First faction, and next week we'll see.
November 22, 2025 at 10:53 PM
So Trump options are (1) keep sending weapons with no TV-friendly results for the foreseeable future (2) try anything to get some ceasefire, including farcical Putin-acceptable plan: it makes for good TV, some likes from the America First faction, and next week we'll see.
Putin is comfy with the status quo: less than 0.5% of Russian men, almost exclusively low status low IQ, die every year. He advances 5km/year. That is sustainable for his lifetime: no pressing need for a deal. Only a deal that keeps a war state (rearming, preparing season two) is acceptable.
November 22, 2025 at 10:53 PM
Putin is comfy with the status quo: less than 0.5% of Russian men, almost exclusively low status low IQ, die every year. He advances 5km/year. That is sustainable for his lifetime: no pressing need for a deal. Only a deal that keeps a war state (rearming, preparing season two) is acceptable.