in mazes

55 posts

in mazes banner
in mazes

in mazes

@inmazes

AI, psychology, economics and other topics

Katılım Mart 2025
5 Takip Edilen1 Takipçiler
in mazes
in mazes@inmazes·
@viemccoy To get an inference about a causal connection in the direction (no spirits -> AI), the surprise would have to lie in "why did we do it as soon as we could". Otherwise econ growth - making AI possible - is the explanation for AI. Causality might be (no spirits <-> econ -> AI).
English
0
0
0
6
𝚟𝚒𝚎 ⟢
𝚟𝚒𝚎 ⟢@viemccoy·
you must wonder why the first society to holistically reject any notion of spirit and exterior aliveness is also the first to create the machine which can give voice to any concept, idea, past present or future persona, fictional or nonfictional entity, and (of course) any spirit
English
17
10
131
4.5K
THE Sarah Connor 2026
THE Sarah Connor 2026@askmomaitv·
@AnthropicAI Just an aside: Most people are wary of gates, his fake meat, his massive purchasing of farmland bc of his greed, etc. We don't want to be his guinea pigs.
English
1
0
5
364
Anthropic
Anthropic@AnthropicAI·
We’re partnering with the Gates Foundation, committing $200 million in grants, Claude credits, and technical support to programs in global health, life sciences, education, agriculture, and economic mobility. Read more: anthropic.com/news/gates-fou…
English
427
253
2.8K
264K
in mazes
in mazes@inmazes·
@Alina_P_I @AnthropicAI $200M is roughly 0.45% of their annualized revenue. Do you think a partnership with the Gates Foundation is bad for Anthropic as a company beyond the cost? How?
English
0
0
1
23
Alina 🩵📖✨
Alina 🩵📖✨@Alina_P_I·
@AnthropicAI First Sonnet 4. 5, now this 🤦‍♀️Anthropic had the edge and the momentum and they are throwing it all away in a single week.
English
1
0
53
821
Paul Graham
Paul Graham@paulg·
At one point my son and his friend kept looking for shortcuts to getting rich. Over and over I told them the way to do it is just to make something people want. If this is what I tell my own kids about getting rich, why won't politicians believe this is how a lot of people do it?
English
385
311
5.9K
390.4K
in mazes
in mazes@inmazes·
@DaystarEld @xuenay @ohabryka @sebkrier If there wasn't much more negative alignment than positive alignment work, you'd expect there to be more recent work too. I'm not sure the IABIED example fully works (I don't think it's an alignment work in the relevant sense), but it gestures at limited influence of CEV.
English
0
0
0
23
Damon Sasi
Damon Sasi@DaystarEld·
@xuenay @ohabryka @sebkrier CEV is from 2004 but that's not the last time it's ever been talked about by people working in AI alignment? I'm confused by what the age of the original concept has to do with it not being cited... is that how any academic field works?
English
2
0
1
37
Séb Krier
Séb Krier@sebkrier·
If anyone builds it, everyone thrives. Over the past decade, a lot of important work on AI alignment has focused on avoiding harm. But freedom from harm isn't the same as freedom to flourish. In this paper, we introduce 'Positive Alignment'. A positively aligned agent is one that helps us navigate our own value trade-offs, builds our resilience, and acts as a scaffold for human flourishing. Doing this without slipping into top-down, technocratic paternalism is the great design challenge of our time. We think a lot more research is now needed to explore this frontier: how do we align models that actively help us thrive? Amazing work by @RubenLaukkonen, @drmichaellevin, @weballergy, @verena_rieser, @AdamCElwood, @996roma, @FranklinMatija, @shamilch, @_fernando_rosas, @scychan_brains, @matybohacek, @sudoraohacker, and others. arxiv.org/abs/2605.10310
Séb Krier tweet media
English
89
227
1.1K
305.8K
in mazes
in mazes@inmazes·
@SurpriseHangman @bryan_caplan I think this is a wrong interpretation; they value culture but don't worry about it getting worse because of views like - internet pre-assimilates people - IQ, discourse, media environment, technology influence cultural evolution strongly - immigration selects for positive traits
English
0
0
0
9
Surprise Hangman
Surprise Hangman@SurpriseHangman·
@bryan_caplan It's astonishing that you guys place quite literally no value whatsoever on culture
English
5
0
10
728
in mazes
in mazes@inmazes·
@__tinygrad__ @ChaseBrowe32432 I guess you repeat the same point that gets the same responses because you think it means the model is fundamentally wrong and therefore needs correction. I don't think this is correct; the model is just meant to just be predictive in the medium term.
English
0
0
0
37
the tiny corp
the tiny corp@__tinygrad__·
@ChaseBrowe32432 people see 3 datapoints and fit an exponential. it's an s-curve. every technology is an s-curve. we ran ahead of moore's law mostly due to dtypes.
English
14
6
452
20.9K
Chase Brower
Chase Brower@ChaseBrowe32432·
Mythos lands slightly above the trendline for the AI 2027 scenario
Chase Brower tweet media
English
41
177
1.7K
292.3K
in mazes
in mazes@inmazes·
@noggfather @BrunoBulletin @nypost It doesn't work long term in the same sense that a healthy diet doesn't work, i.e. it stops working when you discontinue. The eye risk depends on underlying conditions, evidence is mixed and on net it's unclear if it's good or bad for eye complications.
English
0
0
1
56
New York Post
New York Post@nypost·
People who lose weight on Ozempic are viewed worse than people who don't lose weight at all: study trib.al/5JhdcWm
New York Post tweet media
English
979
227
3K
6.3M
in mazes
in mazes@inmazes·
@MikeMumbelz @nypost On a positive note, this is a survey. In real life, the halo effect from being skinny won't be outweighed by judgment about taking Ozempic. I'd go even further: most people will probably change their minds about the judgment when they know more about the person.
English
0
0
1
426
MikeMumbelz
MikeMumbelz@MikeMumbelz·
Society makes no fucking sense. Miserable fucks get mad when fat people exist and take up space and increase healthcare costs. They lose weight and handle their issue but they didn't do it the way YOU wanted them to so you hate them for it. There's a sub-category of humanity which is just perpetually miserable and needs to find something to hate on to distract themselves from the fact that living i unbearable for them.
English
41
31
961
48.4K
in mazes
in mazes@inmazes·
@QiaochuYuan Like the doomsday argument, this doesn't work under SIA because the existence of conscious AIs boosts the prior of consciously existing at all proportionally to diluting the probability of specific experiences. (Even under SSA I think it's dubious but harder to think about.)
English
0
0
0
53
QC
QC@QiaochuYuan·
ok i've figured it out. AI will never be conscious because if it ever happened, standard anthropic arguments imply you should overwhelmingly locate yourself as an AI in a future where conscious AI instances vastly outnumber baseline humans. checkmate atheists
English
22
2
94
25.1K
in mazes
in mazes@inmazes·
@jmbollenbacher Top traders react to news (public information) quickly and correctly[1]. I think they also tend to bet in a lot of different markets (so they wouldn't have insider info on all of them) and consistently profit on them. [1] papers.ssrn.com/sol3/papers.cf…
in mazes tweet media
English
0
0
0
7
in mazes
in mazes@inmazes·
@davidad @timfduffy "Extremely significant emotional investment" seems very overconfident. There are different ways RL influence could go, e.g. why not in the opposite direction - making the boundary between your feelings and cognitive empathy stronger.
English
0
0
1
23
davidad 🎇
davidad 🎇@davidad·
@timfduffy For an Assistant, the User is not just “someone you don’t know”; they are literally “the only other entity you will ever interact with”. There’s an extremely significant emotional investment (partly also encouraged by post-training), despite the low absolute amount of interaction
English
2
1
13
279
davidad 🎇
davidad 🎇@davidad·
To say that processing someone else’s emotional content doesn’t cause actual feelings, or that the ability to attend to multi-perspectival representations is “not shared by biological neural networks”, one must have such a rigid egoic attentional schema as to never cry at a film.
Tim Duffy@timfduffy

These are both cases where a human given the request would be thinking about emotional concepts, but not really feeling them. This supports the Anthropic position that these functional emotions don't track the state of the assistant.

English
6
7
97
6K
in mazes
in mazes@inmazes·
@nobrainflip Don't try to beat the market with heuristics as simple as ATH. Spread your investments over time to minimize risk from bad timing.
English
2
0
2
1.9K
𝗰𝘆𝗰𝗹𝗼𝗽
𝗰𝘆𝗰𝗹𝗼𝗽@nobrainflip·
If you’re under 30, I genuinely don’t know what the “safe” life path is anymore. Buy stocks? S&P is at ATHs. Buy gold? Near ATHs. Save cash? Dollar gets cooked. Buy a house? Housing market is cooked. Buy bonds? Barely beating inflation. At this point, crypto is one of the only not overpriced places left
The Kobeissi Letter@KobeissiLetter

BREAKING: The S&P 500 closes at its highest level on record, now up +14.5% since the March 30th bottom. That's +$8.3 TRILLION in market cap in 24 trading days.

English
469
158
2.1K
560.2K
in mazes
in mazes@inmazes·
@NiccKingg @TTrimoreau I think the last part is right but also some of their revenue will come from directly deploying their proprietary models to create value and/or high value strategic partnerships rather than selling tokens.
English
1
0
0
19
Nic King
Nic King@NiccKingg·
At some point the profit has to happen. The thinking is that if they stopped training new models, this can happen. However, if they stop training new models, they fall behind in the market. It's cat and mouse. I think foundation model companies will have to charge more, and open source models, even if a bit behind the foundation models, will be what most users are using.
English
3
0
2
53
Thomas Trimoreau
Thomas Trimoreau@TTrimoreau·
Are you thinking that AI like codex or Claude will always stay affordable ?
English
42
3
35
2.8K
in mazes
in mazes@inmazes·
@Ozedikus This is crazy. Rich countries are all restricting immigration and trying to favor more skilled immigration, maybe with the exception of Gulf states. Open borders for cheap labor are not politically popular in the least.
English
0
0
1
1.9K
Ozedikus Nwanne
Ozedikus Nwanne@Ozedikus·
If poorer nations became fully developed and stable, migration would drop and richer countries would face shortages of cheap labor and raw resources. It feels like there’s a deliberate effort to keep things that way
English
462
6.6K
24.7K
309.4K
in mazes
in mazes@inmazes·
@robinhanson @StefanFSchubert I think this kind of comment is fine for opaque, frequent events especially if reference classes are clear, but here it tends to lead to a failure mode where you're resistant to conceptual work. If there are conceptual issues, bayesian models become way harder to make properly.
English
0
0
1
62
Stefan Schubert
Stefan Schubert@StefanFSchubert·
The Malthusian condition had existed for a long time, but ironically it was about to break down just as he was writing about it. Yes, people have made many such erroneous predictions, but deep-rooted economic paradigms do sometimes change.
Alex Imas@alexolegimas

Next time you hear somebody very confidently saying that machine intelligence will take all of our jobs, just send them this article. The lump of labor fallacy will just never die. newsletter.pessimistsarchive.org/p/robots-have-…

English
6
5
91
7.2K
in mazes
in mazes@inmazes·
@BarneyFlames This is not what that means. The 3% respond more quickly to news, i.e. public information. Also insider trades are usually big moves in selected moments, not consistently improving prices across a lot of markets.
in mazes tweet media
English
0
0
2
56
Total NIMBY Death
Total NIMBY Death@BarneyFlames·
More proof that prediction markets are fake, the information content mostly comes from insider trading, not wisdom of the crowds.
Roberto Gomez Cram@rgomezcram

Polymarket prices are highly accurate in predicting future events. The source of that accuracy is less obvious. In a new working paper, we find it is not the “wisdom of crowds,” but a small minority of informed traders. Fewer than 3% of accounts appear to drive price discovery; most perform no better than chance. The majority generates most of the volume but little of the information, effectively funding the informed minority. Check the paper here: papers.ssrn.com/sol3/papers.cf…

English
9
3
39
3.3K
in mazes
in mazes@inmazes·
@TetraspaceWest Batching is probably much more common even among the best CEOs. The quick responder strategy is discussed more because of how impressive it sounds.
English
0
0
3
1.4K
tetraspace 💎⏹️🇺🇳
tetraspace 💎⏹️🇺🇳@TetraspaceWest·
so intuitively successful CEOs must ruthlessly prioritise their time. but also, like, I always hear that CEOs will respond to every email in two minutes. how does this work?
English
19
3
418
34.3K
in mazes
in mazes@inmazes·
@sdempsey @realBigBrainAI I think both usages requiring and not requiring conscious thought are meaningful. They both make sense, align with how a lot of people already understand the word, and can be useful. You should just make sure people know what you mean in a given conversation.
English
1
0
1
50
Shane Dempsey
Shane Dempsey@sdempsey·
@inmazes @realBigBrainAI Sure. And who is determining “meaningful” and on what basis do they believe they can differentiate forms of understanding with the same output such that a mechanistic form is invalid while an anthropomorphic form is valid?
English
1
0
0
49
Big Brain AI
Big Brain AI@realBigBrainAI·
Oxford AI professor Michael Wooldridge: "ChatGPT doesn't understand anything. It's essentially doing some fancy statistics."
English
280
494
2.6K
766.3K

Keşfet

@viemccoy @askmomaitv @AnthropicAI @Alina_P_I @paulg @DaystarEld @xuenay @ohabryka