Sign in

titotal.bsky.social

@titotal.bsky.social
94 followers 369 following 32 posts

titotal.substack.com

PostsRepliesMedia
titotal.bsky.social @titotal.bsky.social · 21h
Speaking as someone who did computational quantum chemistry simulations for several years, it is not at all true that chemistry can be "reliably simulated". It takes days of supercomputing time to somewhat accurately simulate the orbital properties of a couple of atoms, even knowing the correct eqs.
010
titotal.bsky.social @titotal.bsky.social · 04/10/2026
I think it was great when it was advocating for more rigor in giving, but the movement has been shifting over the years to arguing for less rigor and more highly speculative decisions based on extremely unreliable expected value calculations.
010
titotal.bsky.social @titotal.bsky.social · 04/10/2026
"Wanting to contain buggy code through controls and security backups is impossible, we should just try and write completely perfect bugless code instead!"
070
titotal.bsky.social @titotal.bsky.social · 04/10/2026
An "effective" altruist should, actually, care about whether they are funneling wealth, power, and influence into the hands of horrible people.
050
titotal.bsky.social @titotal.bsky.social · 01/10/2026
About eighty articles against hard take-off in here: magnusvinding.com/2017/12/16/a... My favourite is reflectivealtruism.com/category/my-... "Diamondoid bacteria" is contested by me here: titotal.substack.com/p/diamondoid... "counting arguments" are debunked here: optimists.ai/2024/02/27/c...
030
titotal.bsky.social @titotal.bsky.social · 22/09/2026
The blog posts spread by AI safety advocates are in no way comparable in credibility to actual scientific articles, and I say this as someone who does both. The level of skeptical scientific thinking and academic rigour in the AI safety community is generally awful.
1281
titotal.bsky.social @titotal.bsky.social · 21/09/2026
I have been asking all these questions specifically to these doomer communities for like a decade now. They have responses to all of them but I find them unsatisfying or straight up wrong.
011
titotal.bsky.social @titotal.bsky.social · 16/09/2026
No, it's silly when talking about much better systems as well. These systems run on the electricity, society and infrastructure we provide to them. Extermination scenarios originally relied on beliefs about magical nanotech, they don't hold up today.
120
titotal.bsky.social @titotal.bsky.social · 14/09/2026
These are the Roko's Basilisk guys, they've been preaching this nonsense in badly written blog posts for decades. They very sincerely believe that AI will program itself into becoming a nigh-omnipotent machine god that can instantly invent magical nanomachines and strip the earth into atoms.
010
titotal.bsky.social @titotal.bsky.social · 12/09/2026
Also this analogy assumes that you and Magnus are fighting on a completely equal playing field. You could definitely beat Magnus if you took away all his pieces. In real life humans start with significant advantages in resources over some rogue AI: it relies on us for everything!
120
titotal.bsky.social @titotal.bsky.social · 08/09/2026
The rationalist apocalyptic AI narrative has barely changed in decades, despite AI developments not matching with their initial predictions. Yudkowsky here thought neural nets were a dead-end, and that AI would be built by a few guys in a basement. They believe in instant machine god as a dogma.
000
titotal.bsky.social @titotal.bsky.social · 24/08/2026
The average person does not actually know which individual companies are building the most datacenters. People are livid about AI datacenters, but don't care about google maps datacenters. This is plausibly because people like google maps and they are indifferent or angry about AI.
010
titotal.bsky.social @titotal.bsky.social · 21/04/2026
I would call them "attempted empiricists": They try to be empirical but do a poor job because they've tried to reinvent everything from the ground up from pop-sci blog posts.
010
titotal.bsky.social @titotal.bsky.social · 31/03/2026
Also telling that the chief prophet of race science crap (steve sailer) was just some guy with a finance and marketing degree. Or Charles murray, a political scientist who failed to explain what heritability is. Or the various intelligence researchers who committed fraud or terrible research.
000
titotal.bsky.social @titotal.bsky.social · 13/03/2026
I wouldn't pay for it, but: 1. if you specifically ask for links, it's often more useful than (enshittified) google for searching for things 2. For small programming tasks, it's sometimes more convenient than yoinking stuff from stackexchange 3. can do tech support slightly better than google
001
titotal.bsky.social @titotal.bsky.social · 05/03/2026
This isn't true, the "ai as normal technology" folks are the first example that comes to mind. The problem is that the two extremes of utopia or armeggedon make for more compelling stories.
000
titotal.bsky.social @titotal.bsky.social · 05/02/2026
This survey is way too generous to the Rationalist movement, you should be asking questions like "Should you quantify your beliefs even if you're just making the numbers up", and "should blog posts be on the same footing as peer reviewed scientific articles"
120
titotal.bsky.social @titotal.bsky.social · 14/01/2026
If either end of the confidence interval is dangerous, then it doesn't really matter if the confidence is wide: the conclusion is still that the concentration is dangerous, the uncertainty is only over how much. This is important information to know, and motivates future, more precise studies
020
titotal.bsky.social @titotal.bsky.social · 21/10/2025
It can generate text consisting of code, and the chatgpt developers can set up a system to actually run the code that it outputs. That code can theoretically set up an excel file and output it for download, if the developers set that up. In practice, it's often easier for it to make stuff up.
000
titotal.bsky.social @titotal.bsky.social · 05/09/2025
When you plop down in the airplane seat, the airport process is finally done. You don't have to do anything or go anywhere for hours, you can finally relax. I'd rather get to this state sooner than later.
020
titotal.bsky.social @titotal.bsky.social · 09/06/2025
Your "scientific papers" are 3 pages long and contain no equations. What is there to check?
020
titotal.bsky.social @titotal.bsky.social · 09/06/2025
These "full physics papers" are 3 and 2 pages long respectively. There's no equations to actually check in there. This is gibberish.
020
titotal.bsky.social @titotal.bsky.social · 23/05/2025
The reason people associate the basilisk with lesswrong is because that's where it was posted, and that's where people took it seriously enough to freak out about it. A dude publicising this doesn't take away from the fact that the flawed epistemology of rationalism lead them to nonsensical results.
120
titotal.bsky.social @titotal.bsky.social · 14/05/2025
Non-yanks are horrified that your service workers have to beg for scraps from every single customer.
0180
titotal.bsky.social @titotal.bsky.social · 09/05/2025
There is a mistake in the very first sentence of your report. It claims DOGE was established in executive order 14512, but EO numbers are not that high yet (the actual EO was 14158). If errors like this are common, then you are overestimating the AI capabilities here.
000
titotal.bsky.social @titotal.bsky.social · 05/05/2025
Please tell me you checked if the Q's and A's were actually correct
100
Reposted by @titotal.bsky.social
Randall Munroe @xkcd.com · 25/04/2025
PhD Timeline xkcd.com/3081
5865978020438
Reposted by @titotal.bsky.social
Mumbling Sage In Wisconsin @mumblingsage.bsky.social · 13/04/2025
I don't think any multiple-choice exams accused white people of causing earthquakes.
"Freedom of Speech" painting by Norman Rockwell
2722
titotal.bsky.social @titotal.bsky.social · 25/03/2025
He covers this in the second video: it's a fun theory but it just doesn't line up with the evidence: there's no way someone wouldn't have seen the accident.
020
titotal.bsky.social @titotal.bsky.social · 06/03/2025
They want to replace every job in the world with garbage enshittified AI.
040
titotal.bsky.social @titotal.bsky.social · 27/02/2025
They are a splinter group of the Rationalists, which are fairly friendly with Yarvins neoreactionaries but I wouldn't describe them as having the same beliefs. They are more intellectual cousins, united by a belief that academia should be replaced by blog posts.
110
titotal.bsky.social @titotal.bsky.social · 13/02/2025
Kinda crazy that a machine which works entirely by multiplying massive matrices of numbers together still hasn't been able to learn the rules for multiplying large numbers together
060
titotal.bsky.social @titotal.bsky.social · 05/02/2025
The global health and animal rights people are still operating and getting a decent chunk of the EA money: in exchange they get trotted out as a shield for the other useless half of EA and the scandalous incompetence of the EA "leadership"
020
titotal.bsky.social @titotal.bsky.social · 02/12/2024
Yudkowsky is like the platonic ideal of the Feynman bro stereotype. All he does is waltz into a field he's read one textbook in, declare himself smarter than all the experts, then waltz out.
060