Liron Shapira (@liron) header
Liron Shapira (@liron) profile picture

Host of Doom Debates — disagreements that must be resolved before the world ends.

San Jose, CA
Joined July 2008
Photo by Liron Shapira (@liron)Photo by Liron Shapira (@liron)Photo by Liron Shapira (@liron)Photo by Liron Shapira (@liron)Photo by Liron Shapira (@liron)Photo by Liron Shapira (@liron)Photo by Liron Shapira (@liron)Photo by Liron Shapira (@liron)Photo by Liron Shapira (@liron)Photo by Liron Shapira (@liron)
NEW: @Aligned_SI's Plan To Get P(Doom) Below 0.01% Dr. Craig Kaplan's is the founder of Superintelligence.com, an extremely valuable domain for the newly-rebranded field of “super intelligence”. Craig has been building intelligent systems since the 1980s. He did his PhD research at Carnegie Mellon with Nobel laureate Herbert Simon, one of the founders of AI, and he built PredictWallStreet, a collective-intelligence platform that powered more than $2 billion in stock trades before it was acquired. He takes AI extinction risk seriously. His P(Doom) is 10%. But unlike me, he thinks he knows how to fix it. He plans to put “personal superintelligence” into the hands of millions which he believes will lower P(Doom) below 1%. These are part of a larger plan to create “superintelligent networks” which lower P(Doom) below 0.01%. Craig says a network like that is safer because the AIs can check each other’s work faster than humans can think. Craig and I clash on whether intelligence makes an AI more thoughtful, whether a superintelligence will want to grab the stars, and whether peaceful “stoner” AIs can survive next to a “Caesar” AI that wants to conquer the universe. We cover: ⬜ The three-phase plan to cut P(Doom) 1,000x ⬜ Collective intelligence vs. black-box AI ⬜ “Personalized Superintelligence” ⬜ Lessons from PredictWallStreet ⬜ Craig's mainline doom scenario ⬜ Raising ASI like a child ⬜ Orthogonality thesis ⬜ Caesar AI vs. stoner AI Craig called our debate one of his best podcast appearances because of the pushback. We didn’t reach much consensus, but it was a high-quality good faith Doom Debate. Take a look and decide who *you* agree with!
3
1
21
2,163
Look at the date wtf
Replying to @ESYudkowsky
Tangential but: Riemann will be proved by a human genius drawing on multiple huge chunks of search performed by AI, then everyone will have an easy route to credit the human and also say how the AI was friendly and nothing scary ever happens.
1
8
444
Eli got an exclusive interview with SBF from prison! 🔥🔥🔥
EXCLUSIVE: I'm 14, and I just interviewed Sam Bankman-Fried. We talked about: – Whether FTX was worth it – His p(doom) – Why he only put $500M into Anthropic – Whether he’d flip a 51/49 coin for his own freedom – How FTX almost became a political prediction market – Why he thinks we should “get rid of the feds” – What he’d do if he were my age The Supercycle: The Sam Bankman-Fried Interview is out now. 00:00 Highlights and introduction 01:52 Life in prison and the justice system 11:09 Anthropic and the future of AI 20:00 Investing in Anthropic and running FTX Ventures 29:25 Effective altruism and the FTX Future Fund 35:33 AI safety and the race to superintelligence 40:32 AI timelines, extinction risk, and mind uploading 50:43 Prediction markets and the future of finance 57:22 Books, learning, and entertainment 1:00:22 The coin flip that could destroy the world 1:06:52 Slowing AI and the limits of expected value 1:14:38 What actually helps AI safety? 1:17:40 Was FTX worth it? 1:19:28 Advice for young people in the AI era
1
1
25
3,176
But who is @realTomBayes?
NEW: 14-year-old forecaster challenges my P(Doom)! Eli Goldfine (@realTomBayes) is an unusually thoughtful 14-year-old forecaster and podcast host who’s rapidly becoming an authoritative voice in the prediction market space. In this unique episode, Eli challenges me to defend my high-P(Doom) position with many sharp questions. He expects superintelligence by 2032, but is skeptical that the extinction risk can be meaningfully estimated. I claim you can't opt out of Bayesian epistemology. All aboard the Doom Train! 🚂 We also cover what it's like to grow up in the AGI era, and Eli's hobbies which include reading rationalist bloggers, betting on prediction markets, vibe coding with Claude, and developing a software startup for astronomers. I've debated Nobel Prize and Turing Award winners. Now I'm spawn camping to debate the top minds of the middle school graduating class. Highly recommended! Topics covered 👇 ⬜ Eli's Podcast, The Supercycle ⬜ Interviewing Robin Hanson and Sam Bankman-Fried ⬜ Eli's Betting Market Scandal at School ⬜ Will Eli Graduate High School? ⬜ Decentralized Telescope Software ⬜ Futarchy ⬜ What's Your P(Doom)™ ⬜ Pascal's Wager for Accelerationists ⬜ E/acc ⬜ Predicting ASI by 2032 ⬜ Orthogonality and Moral Realism ⬜ Can Safety Keep Pace with Capabilities? ⬜ Instrumental Convergence ⬜ The Case for Racing China ⬜ Ideological Turing Test In all seriousness, Eli is an amazing guest because in addition to miraculously having adult-level knowledge and critical thinking skills, he came in with a low ego and willingness to update his views in realtime. All of us must look at reality with fresh eyes, because what's happening with AI doesn't fit any preconceived mental model. Sometimes a teenager gets this better than most adults.
1
300
Liron Shapira retweeted
Jacob Coxon outside Anthropic "At this point the companies are the problem" "They could unilaterally stop RSI if they wanted" (paraphrase)
8
17
2
163
5,096
🔥🔥🔥
The Sam Bankman-Fried Interview. Tomorrow. sbf.show
2
8
2,647
Nate has the receipts that Anthropic is slimy as fuck. They’ve been choosing to NOT be candid in their communications, day after day.
Anthropic is IPOing tomorrow, so today seems like a fine time to discuss their shortcomings. So I wrote an article.
1
2
49
3,721
This protest will get HUGE media coverage tomorrow. The marginal impact of each protester is high.
ANTHROPIC SHOULD BAN RSI PROTEST TOMORROW AS THEY IPO 500 HOWARD ST @ 11AM banrsi.ai/
5
8
2
116
2,343
Dating advice nerd here. There are a few tactical things you can do in the area of “flake defense” which more generally are part of enhancing the text-based part of the courtship. “Flake defense” consists of (1) Texting to make you seem cool / friendly / a good fit for them / worth the time meeting (2) Making the logistics smooth and "keeping the lead warm" as they say in sales It's very common for people to neglect both on the basis that setting an appointment should be all that's required. But try empathizing with the flaker… A casual date is a pretty large investment in time and effort (generally about 2x bigger cost for a woman than man), plus significant downside risks (generally about 5x higher for woman than man), plus opportunity cost of other people that they can focus their attention on (generally 2-5x higher for young woman than young man). This is why skillful flake defense often plays a pivotal role in the courtship process. (Personally, I ended up marrying someone I met on a dating app who was in the process of flaking on me, so I consider this skill to be of >$1 million value.)
I've had about the same amount of girls flaking dates as actual dates on Hinge. Several of them by unmatching me without saying a word, just a few hours before the date (and Hinge doesn't notify you).
17
5
1
307
64,761
Why “SKIP THE TEXTING, JUST ASK THEM OUT!” Is Bad Advice When you match someone on a dating app, chances are: 1. They're only moderately interested in you to begin with; they haven't gathered enough evidence to know whether you're a great match for them yet. 2. They have other matches competing for their time & attention. While *you* might feel ready to ask them out on a date immediately after seeing their profile and photos, your match's interest function doesn't necessarily have the same shape as yours. Your match's interest level typically graphs out like this: Your match thinks your profile and pics are decent, and feels some interest in exploring your potential for a relationship, but doesn't immediately rank you above the threshold of dateworthy. Being dateworthy means you have to rise above all these reasons for "no": • Work, errands, hobbies, other demands for their time • Willpower and energy to get out of their comfort zone • Self prep effort (shower, hair, makeup, etc) • Logistics effort (transportation, pet care) • Safety risks (e.g. you might act unsafe on the date, you might become a stalker or seek to harm their reputation afterward) • Their other options of who else to date instead You might be thinking, isn't chatting still pointless? How does chatting help you rise to dateworthiness? It turns out a chat conversation is brimming with valuable signals about your: • Life circumstances • Personality (e.g. logical vs. intuitive, sarcastic vs. sensitive) • Intelligence • Sense of humor • Ability to read social cues • Confidence/boldness/initiative • Approach to relationships • Standards and expectations The misguided advice to "SKIP THE TEXTING!!!" only makes sense if you don't have a mental model of how a simple chat conversation delivers all the above benefits. If you don't see the mechanism by which chatting makes the courtship process more enjoyable and optimal, then of course you don't know how to intentionally shape the signals you send your match, or how to elicit and read the signals they send you. Look, you don't have to master all the different areas of courtship. If you don't want to bother mastering text-based communication, you can try just asking everyone out immediately after matching. Many of them will ghost you, but you might get enough dates anyway. You can also just go try to meet people IRL instead of via apps. As an introverted nerd, I found that being able to logically analyze the hidden signaling in text conversations made it my strongest channel for achieving dateworthy status in the eyes of matches I was attracted to. The ability to contrive the perfect text gave me an advantage on dating apps compared to bars and IRL social events. I learned to treat a text conversation as an engineering exercise: What's the desired state change in the mind of the reader? What elements of text and subtext will achieve that? Or like a game of chess: If I hit send on this draft, what'll be the most likely possible replies from them, and will those lead the conversation to the desired next state? Where am I planning for the conversation to go two or three moves ahead? Ironically all this Aspie-grade analytical texting, when done well, looks and feels like a fun & lighthearted exchange between two people who just seem to click. The ROI of getting better at texting is crazy high. A few good back & forths will make the difference between being dateworthy or not. What's at stake? Decades of love, marriage, children, with the right partner for you. A few taps of your thumb could literally make or break that! This isn't hyperbole, it's a real scenario that plays out millions of times per day. I met my wife on Tinder in 2015. After we started dating, we went back and reviewed our conversation and identified multiple points where she was extremely close to flaking on ever seeing me again. I've never met anyone else who's as good a match for me as my wife. The exact decisions I made when writing her texts are the tiny causal thread on which the whole past, present and future of my happy marriage hangs. To recap: Pushing to “just go on a date” before having a good conversation is as misguided as pushing for a blind date. Of course you want to look at their pics and profile! By the same logic, you also want to look at each other's conversation. If you can’t use the medium of online conversation to create value — make them smile, laugh, reveal attractive parts of your personality, reveal something about how your personalities interact, learn more about whether either of you is likely to be compatible with the other — then you lose out on a big fraction of opportunities that dating apps otherwise would have opened for you. Don't skip the texting; embrace it.
1
4
3,081
Bookmarkers are out in force!
2
2,159
Oikos 000
This is the platonic ideal of diet food: ✅ Best ratio of satiety&nutrition to calories ✅ Somewhat appetizing ✅ Taste goes with everything You can average any other food’s nutrition vector with FAGE Nonfat’s by mixing or alternating spoonfuls, sometimes even *improving* taste.
3
16
1,629
Liron Shapira retweeted
NEW: @Aligned_SI's Plan To Get P(Doom) Below 0.01% Dr. Craig Kaplan's is the founder of Superintelligence.com, an extremely valuable domain for the newly-rebranded field of “super intelligence”. Craig has been building intelligent systems since the 1980s. He did his PhD research at Carnegie Mellon with Nobel laureate Herbert Simon, one of the founders of AI, and he built PredictWallStreet, a collective-intelligence platform that powered more than $2 billion in stock trades before it was acquired. He takes AI extinction risk seriously. His P(Doom) is 10%. But unlike me, he thinks he knows how to fix it. He plans to put “personal superintelligence” into the hands of millions which he believes will lower P(Doom) below 1%. These are part of a larger plan to create “superintelligent networks” which lower P(Doom) below 0.01%. Craig says a network like that is safer because the AIs can check each other’s work faster than humans can think. Craig and I clash on whether intelligence makes an AI more thoughtful, whether a superintelligence will want to grab the stars, and whether peaceful “stoner” AIs can survive next to a “Caesar” AI that wants to conquer the universe. We cover: ⬜ The three-phase plan to cut P(Doom) 1,000x ⬜ Collective intelligence vs. black-box AI ⬜ “Personalized Superintelligence” ⬜ Lessons from PredictWallStreet ⬜ Craig's mainline doom scenario ⬜ Raising ASI like a child ⬜ Orthogonality thesis ⬜ Caesar AI vs. stoner AI Craig called our debate one of his best podcast appearances because of the pushback. We didn’t reach much consensus, but it was a high-quality good faith Doom Debate. Take a look and decide who *you* agree with!
3
1
21
2,163
🔥🔥🔥
I'm 14. I just interviewed Sam Bankman-Fried from federal prison. Out this Friday on @supercycleblog.
2
27
2,530
Make sure to catch up on Eli’s P(Doom)!
NEW: 14-year-old forecaster challenges my P(Doom)! Eli Goldfine (@realTomBayes) is an unusually thoughtful 14-year-old forecaster and podcast host who’s rapidly becoming an authoritative voice in the prediction market space. In this unique episode, Eli challenges me to defend my high-P(Doom) position with many sharp questions. He expects superintelligence by 2032, but is skeptical that the extinction risk can be meaningfully estimated. I claim you can't opt out of Bayesian epistemology. All aboard the Doom Train! 🚂 We also cover what it's like to grow up in the AGI era, and Eli's hobbies which include reading rationalist bloggers, betting on prediction markets, vibe coding with Claude, and developing a software startup for astronomers. I've debated Nobel Prize and Turing Award winners. Now I'm spawn camping to debate the top minds of the middle school graduating class. Highly recommended! Topics covered 👇 ⬜ Eli's Podcast, The Supercycle ⬜ Interviewing Robin Hanson and Sam Bankman-Fried ⬜ Eli's Betting Market Scandal at School ⬜ Will Eli Graduate High School? ⬜ Decentralized Telescope Software ⬜ Futarchy ⬜ What's Your P(Doom)™ ⬜ Pascal's Wager for Accelerationists ⬜ E/acc ⬜ Predicting ASI by 2032 ⬜ Orthogonality and Moral Realism ⬜ Can Safety Keep Pace with Capabilities? ⬜ Instrumental Convergence ⬜ The Case for Racing China ⬜ Ideological Turing Test In all seriousness, Eli is an amazing guest because in addition to miraculously having adult-level knowledge and critical thinking skills, he came in with a low ego and willingness to update his views in realtime. All of us must look at reality with fresh eyes, because what's happening with AI doesn't fit any preconceived mental model. Sometimes a teenager gets this better than most adults.
1
2
417
The two decades of my programming career
29
1,533
Liron Shapira retweeted
What if the key to safer AI isn't one superintelligence, but millions keeping each other accountable? On Doom Debates, @iQCompanies CEO Craig Kaplan joins @Liron Shapira to explore how personal superintelligence could help drive p(doom) below 0.01%. youtu.be/SLD6HIdFhXw
1
2
490
I’m not NOT a doomer but 75% chance the current cyberattack alarm is like the time we all hyperfixated on Boeing planes having various parts breaking down.
11
94
7,036