Caleb Parikh

900 posts

Caleb Parikh

Caleb Parikh

@caleb_parikh

Trying to make the future go well by funding impactful projects.

Berkeley Katılım Ekim 2016
690 Takip Edilen468 Takipçiler
Sabitlenmiş Tweet
Caleb Parikh
Caleb Parikh@caleb_parikh·
I increasingly feel like transparency should be the main demand/frame that the AIS community tries to meme. It's broadly popular, will help keep the AIS community and public institutions relevant during takeoff, and is more justified than pausing (imo).
English
3
3
58
3.4K
Caleb Parikh
Caleb Parikh@caleb_parikh·
@AaronBergman18 @kepe__ Which imo points to the use of the thought experiment mostly playing on people’s biases. I don’t feel like I fully understand your reasoning (e.g. you said somewhere that your argument was setup to avoid biases which confused me) so maybe I’m missing the core of your argument.
English
1
0
1
13
Caleb Parikh
Caleb Parikh@caleb_parikh·
@AaronBergman18 @kepe__ You said “… would you accept 70 years of the worst conceivable torture in exchange for any amount of happiness?” if you’d used 1s instead of 70 years, or said an extremely short duration of [extreme] suffering for a 70 years of [pleasure] that would shift people’s intuitions
English
1
0
1
11
kepe ✨🇫🇮
kepe ✨🇫🇮@kepe__·
for a million dollars, you'll experience the worst, most horrific torture you can imagine. once it starts, you can't back out. after it's done, it'll stop immediately, and you won't have lasting effects like PTSD what's the maximum you would take the deal for?
English
40
3
46
27.1K
Caleb Parikh
Caleb Parikh@caleb_parikh·
@jachiam0 DeepMind is owned by Alphabet, a publicly traded company. Meta is also publicly traded.
English
0
0
1
122
Joshua Achiam
Joshua Achiam@jachiam0·
It's worth observing that the public already owns about 26% of OpenAI stock, embodied in the OpenAI Foundation. The nonprofit Foundation also controls the for-profit part. OpenAI is the only AI company that is set up with public ownership at all.
Sen. Bernie Sanders@SenSanders

AI is built on humanity’s collective knowledge. The wealth it generates must benefit humanity — not just Elon Musk, Sam Altman and other AI oligarchs. That’s why I’ll be introducing the American AI Sovereign Wealth Fund Act — to give the public a direct ownership stake.

English
48
3
120
95.9K
Caleb Parikh
Caleb Parikh@caleb_parikh·
@AaronBergman18 @kepe__ I reread this post and I don’t really get it. Unlike you, I think IHE is biased and my intuitions run in the opposite direction, which undermines all of the asymptote stuff in part 2 (iiuc).
English
1
0
0
35
Aaron Bergman 🔍
Aaron Bergman 🔍@AaronBergman18·
This is heterodox but (1) “infinite” has various possible meanings but the one relevant here, I claim, plausibly (and imo irl) corresponds to just a normal set of physical circumstances. And then you can get worse than this definition of “infinite” (and likewise there are various sizes of infinity in math) I agree I can’t imagine experiencing it, but I’m not sure this is the case for all physically possible minds forum.effectivealtruism.org/posts/je5TiYES…
English
2
0
1
69
Caleb Parikh
Caleb Parikh@caleb_parikh·
@kepe__ @AaronBergman18 I’m not sure there’s a coherent notion of “infinitely” bad torture, and I’m confident that you can’t imagine experiencing it.
English
1
0
2
62
Caleb Parikh
Caleb Parikh@caleb_parikh·
@AaronBergman18 @kepe__ Why? Seems unlikely that pain vs pleasure is so asymmetric that it’s better just to die than live - particularly if you’ll be relatively well off.
English
1
0
1
49
Aaron Bergman 🔍
Aaron Bergman 🔍@AaronBergman18·
@kepe__ It’s better just to die Trying to understand “worst possible torture” using your intuitions based on >99.9% of actual human experience is like trying to viscerally comprehend distance to the edge of the observable universe
English
3
1
12
304
Caleb Parikh retweetledi
Bull Theory
Bull Theory@BullTheoryio·
BREAKING: Google is planning to release 32 million mosquitoes across Florida and California. The company has asked the EPA for permission to proceed, with the public given until June 5 to respond. The mosquitoes are infected with Wolbachia bacteria, which stops them from reproducing and slowly collapses the wild population from within. Google's previous Debug Project trial in California's Central Valley nearly eliminated mosquitoes from three test sites entirely. A separate trial in Singapore cut dengue cases by 70% within 12 months. Google has now released over 1 billion mosquitoes across four continents. This new proposal is the largest deployment in US history.
Bull Theory tweet mediaBull Theory tweet media
English
2.7K
2.2K
25.2K
14.4M
Caleb Parikh
Caleb Parikh@caleb_parikh·
@ciphergoth My impression is that passkeys address the phishing threat model about as well as security keys. IIUC, they provide the same cryptographic guarantees, with the main differences being how much you trust the secure enclave, and whether you accept the tradeoff of cloud sync?
English
0
0
0
116
Caleb Parikh
Caleb Parikh@caleb_parikh·
@MScarince @LinchZhang I’m not sure the ocr thing is important. It’s just one weird case on a small amount of text that moved the pangram prediction enough to see something.
English
1
0
1
38
Linch
Linch@LinchZhang·
I've seen dozens of people trying to discredit Pangram analysis via reference to either "AI detectors" in general, or clearly different AI detectors specifically. It's like going to a Thai restaurant and complaining about your Big Mac because "restaurants suck"
Agostinho Serrano@EducatingwithAI

@freed_dfilan My own PHD thesis (1999), many paragraphs are 100% AI Generated per those detectors. 🤣

English
5
2
64
6.8K
Caleb Parikh
Caleb Parikh@caleb_parikh·
@LinchZhang @MScarince When I ran this using the OCR'd text from a screenshot, it gave 100% human, then I found the same paragraph online, and it said AI-assisted. Afaict, the only difference are the ASCII quotes used, and some newlines at the bottom.
Caleb Parikh tweet mediaCaleb Parikh tweet mediaCaleb Parikh tweet media
English
1
0
1
58
Linch
Linch@LinchZhang·
@MScarince I do think this is a real false positive! Though in context it says "lightly AI assisted" with "low confidence" and it's a single passage. This is much less aggressive than anything in the Humanitas.
English
1
0
2
89
Caleb Parikh retweetledi
Aidan O’Gara
Aidan O’Gara@aidanogara_·
New funding opportunity to work on AI-enabled concentration of power! Apply by July 2 for up to $2M/year for individuals or organizations. Areas include model specs, mass surveillance, autonomous weapons, AI tools for democratic debate, and much more. Link below.
English
6
21
117
10.2K
Nate Soares ⏹️
Nate Soares ⏹️@So8res·
I'm at a book festival in Wales right now and it's doing its best to compete with Arizona weather-wise
English
2
1
38
3K
Caleb Parikh
Caleb Parikh@caleb_parikh·
@So8res This feels like saying Eliezer/Nate called their org MI[research]I -> they thought research was going to be much more impactful than comms. They called their book IABIED -> they think alignment is impossible. IMO there are better sources of evidence, but sure you CAN deduce.
English
0
0
4
216
Nate Soares ⏹️
Nate Soares ⏹️@So8res·
Early EAs branded their organizations with names like "80,000 hours", from which you can deduce that they were pretty bad at internalizing the possibility of AI.
English
18
8
417
22.2K
Caleb Parikh
Caleb Parikh@caleb_parikh·
@AndyMasley Idk what's going on here; it's happened twice in the last two days for me. Half the reason I booked the Uber was to have AC rather than walk in 33°C heat. If you ask them politely to put the windows up and the AC on, they are normally pretty chill and just do it ime.
English
0
3
9
799
Andy Masley
Andy Masley@AndyMasley·
I’ve taken a few Ubers in the UK so far and none of them use AC, they all just drive with the windows down. What’s the story?
English
39
1
105
14.7K
Dwarkesh Patel
Dwarkesh Patel@dwarkesh_sp·
Who should I interview on my podcast? Open to more AI, but also to random history/econ/etc professors that I might not have heard of before.
English
1.3K
43
1.9K
406.2K
Caleb Parikh
Caleb Parikh@caleb_parikh·
@ShakeelHashim Could you share a prompt or chat transcript that produced substantially better writing than most humans by your lights? I could see lots of people being interested in taking a look.
English
0
0
1
105
Caleb Parikh
Caleb Parikh@caleb_parikh·
@AdriGarriga @LinchZhang Idk, in Genesis God only promises not to send a global flood, and x-risk was averted only by Noah and co.'s actions.
English
3
0
4
95
Adrià Garriga-Alonso
Adrià Garriga-Alonso@AdriGarriga·
@LinchZhang I don't think existential risk is thinkable in a universe where God created humanity in His image and is constantly watching over us.
English
2
0
4
137
Linch
Linch@LinchZhang·
Kinda sad that the recent papal encyclical on AI didn't mention existential risk even once, given that my boy Claude was a major contributor to ghostwriting it. Really thought that Anthropic/Claude cared more about extinction and takeover risk!
English
2
0
27
1.5K
Caleb Parikh
Caleb Parikh@caleb_parikh·
@idavidrein @joel_bkr new METR uplift study when? It would help society keep tabs on spiritual def/acc - though it's pretty decorrelated from common-sense strategies like having your AI corporation tithe, spinning up praying subagents, or burning tokens on reading scripture.
Linch@LinchZhang

Significant fractions of Magnifica Humanitas, the Pope's new encyclical on AI, appears AI-generated or AI-assisted. Maybe Claude specifically, judging from the repeated references to 'genuinely." Kinda interesting!

English
0
0
4
359
Leo Gao
Leo Gao@nabla_theta·
@kerckhove_ts buffers are hugely useful if there are large efficiency returns to batching, which Factorio mostly doesn't have
English
1
0
22
962
Tom Sydney Kerckhove
Tom Sydney Kerckhove@kerckhove_ts·
There's a rule of thumb in Factorio: "avoid item/fluid buffers" which you only tend to learn after quite a while. (Buffers don't really save you, but do prolong the time before you find a broken process) I think about this a lot in the software industry.
English
55
58
1.8K
99.8K
Peter Wildeford🇺🇸🚀
Peter Wildeford🇺🇸🚀@peterwildeford·
Someone tried to do the calendar scam on me again, attempting to takeover my account. Here's the anatomy of the scam - don't fall for it.
Peter Wildeford🇺🇸🚀 tweet media
English
3
8
117
8.5K

Keşfet

@AaronBergman18 @kepe__ @jachiam0 @ciphergoth @MScarince @LinchZhang @elonmusk @BarackObama