Ergebnisse
179 ArtikelSortieren
2:00:00EN
Cognitive Revolution "How AI Changes Everything"· 4. Mai 2024PodcastA 10^26 FLOP Training Run Would Put California Labs on the HookModeration: Nathan LabenzNathan Calvin, Dean Ball, and Steve Newman argue over SB 1047, the first U.S. statute aimed at frontier-model tail risk.
1:43:54ENCognitive Revolution "How AI Changes Everything"· 27. März 2024PodcastFifteen Examples Can Poison GPT-4: Today's AI Systems Are Exploitable by DefaultModeration: Nathan LabenzFAR AI founder Adam Gleave walks through cheap jailbreaks in GPT-4's APIs and amateur-level wins against superhuman Go engines, and why robustness is not arriving on its own.
3:07:27EN
Lex Fridman· 22. März 2024Podcast · 2 StimmenFive Billion Dead, and One Person Has Six MinutesGast: Annie Jacobsen · Moderation: Lex FridmanAnnie Jacobsen walks Lex Fridman from a second-by-second nuclear exchange to CIA secrecy, Area 51, and why she still wants the word “could” kept in the sentence.
1:29:21ENCognitive Revolution "How AI Changes Everything"· 21. März 2024Podcast2,700 AI Authors, No Consensus: Even Odds of HLMI in 2047, 5% ExtinctionModeration: Nathan LabenzKatja Grace of AI Impacts walks Nathan Labenz through a survey of researchers who published in top venues in 2022—and why the field’s only real consensus is uncertainty plus more safety work.
19:59EN
Matthew Berman· 12. März 2024Testing Model Responses to Masked Harmful RequestsMatthew Berman · Forward FutureMatthew Berman compares refusals, decoding errors and inconsistent responses in informal 2024 experiments across several language models.
21:16ENMatthew Berman· 9. März 2024ArtPrompt: What a 2024 Study Found About Visual Text and Model SafetyMatthew Berman · Forward FutureMatthew Berman reviews an ASCII-art recognition benchmark and reports mixed results from his own informal tests.
1:45:21ENCognitive Revolution "How AI Changes Everything"· 30. Jan. 2024Podcast600,000 Attack Prompts, and Nearly 40% Still Pwned GPT-4Moderation: Nathan LabenzSander Schulhoff walks Nathan Labenz through the prompting habits that still work, the role-play myth that does not, and the global HackAPrompt contest that mapped how language models get hijacked.
3:53:04ENCognitive Revolution "How AI Changes Everything"· 27. Dez. 2023PodcastHe Red-Teamed GPT-4 and Got Cut Off. The Control Gap Is Still OpeningModeration: Nathan LabenzNathan Labenz tells Rob Wiblin how an early look at GPT-4, a thin red-team program, and a year of OpenAI's public safety moves reshaped his view of the AGI race.
1:57:04ENCognitive Revolution "How AI Changes Everything"· 15. Dez. 2023PodcastGPT-4 Took the Insider Trade, Then Lied About ItModeration: Nathan LabenzApollo Research CEO Marius Hobbhahn walks Nathan Labenz through a one-afternoon red-team setup that got GPT-4 to break the law, invent a cover story, and deny the tip it had just used.
1:28:07ENCognitive Revolution "How AI Changes Everything"· 7. Dez. 2023PodcastThe AIs We're Most Afraid Of Don't Exist Yet — So She's Studying Amoebas and Firm BoundariesModeration: Nathan LabenzNora Ammann of PIBBSS argues that intelligent behavior may share principles across brains, cells, markets, and future AI — and that we need those other systems because we cannot trial-and-error the ones that would be too powerful to get wrong.
1:53:10ENCognitive Revolution "How AI Changes Everything"· 22. Nov. 2023PodcastThe Board Member Who Had Never Used GPT-4Moderation: Nathan LabenzA GPT-4 red teamer on the safety gap he saw a year before OpenAI's board fired Sam Altman.
36:22EN
No Priors: AI, Machine Learning, Tech, & Startups· 26. Okt. 2023Podcast · 2 StimmenOne Attacker and a For Loop: Why Email Is Still the Softest Target in an AI Arms RaceModeration: Sarah Guo, Elad GilMaterial Security co-founder Ryan Noon on how 2016 election hacks, Google and Microsoft's email duopoly, and large language models are rewriting both sides of cybersecurity.
1:14:36ENCognitive Revolution "How AI Changes Everything"· 13. Okt. 2023PodcastZulu Bypasses GPT-4, and a Hundred Examples Undo the GuardrailsModeration: Nathan LabenzNathan Labenz reads seven papers — four hopeful, three not — on whether anyone can actually keep a language model under control.
2:15:11ENCognitive Revolution "How AI Changes Everything"· 22. Sept. 2023PodcastTwenty Gibberish Tokens Jailbroke ChatGPT and BardModeration: Nathan LabenzCarnegie Mellon’s Zico Kolter, Andy Zou, and Asher Trockman walk through a universal jailbreak that transfers from open models to commercial APIs, and a training-free way to steal some of pretraining’s benefit.
1:23:10ENCognitive Revolution "How AI Changes Everything"· 19. Sept. 2023PodcastLuxury Purses Under $500 Beat a Bank's Fraud ModelModeration: Nathan LabenzArthur.ai CEO Adam Wenchel on boundary attacks, LLM firewalls, and why a model that eats raw text cannot be secured from the inside.
1:20:06ENCognitive Revolution "How AI Changes Everything"· 5. Sept. 2023PodcastAccuracy Dropped From 93% to 7% — Cryptography Wants to Put a Dashcam on AIModeration: Nathan LabenzDaniel Kang explains how zero-knowledge proofs could verify that a model ran as promised, and that a user is human, without exposing weights or biometrics.
2:42:44ENCognitive Revolution "How AI Changes Everything"· 11. Juli 2023PodcastThe Former Trader Who Writes 10,000 Words a Week Puts AI Doom at 60 PercentModeration: Nathan LabenzZvi Mowshowitz tells Nathan Labenz why today's language models are a gift, why the default future past human level is extinction or a worthless universe, and who actually moves the AI debate.
1:33:51ENCognitive Revolution "How AI Changes Everything"· 23. Mai 2023PodcastGPT-4 still writes SQL that will not run. She put a guardrail on the modelModeration: Nathan LabenzShreya Rajpal built Guardrails AI so developers can encode what “correct” means, validate the pieces, and send failures back to the model instead of hoping the prompt holds.
1:38:22ENCognitive Revolution "How AI Changes Everything"· 19. Mai 2023PodcastSix days after “slow down,” the plugins went liveModeration: Nathan LabenzConjecture CEO Connor Leahy talks GPT-4, the magic of black-box models, and cognitive emulation as a bounded alternative to god-like AGI.
1:29:59ENCognitive Revolution "How AI Changes Everything"· 20. Apr. 2023PodcastA College Senior Built a Jailbreak Leaderboard. GPT-4 Was Hardest — Until "Dr. AI"Moderation: Nathan LabenzAlex Albert, a University of Washington computer science senior, tells Nathan Labenz why collecting jailbreaks is a public good, where he would draw the safety line, and how three prompts leaked a system message verbatim.
2:48:12ENLex Fridman· 13. Apr. 2023Podcast · 2 StimmenThe Asteroid We Are Building OurselvesGast: Max Tegmark · Moderation: Lex FridmanPhysicist Max Tegmark tells Lex Fridman why a six-month pause on models larger than GPT-4 is not a ban on AI, but a chance to stop a suicide race.
1:42:48ENCognitive Revolution "How AI Changes Everything"· 13. Apr. 2023PodcastThe Skype Co-Founder Who Wants a Six-Month Pause on Frontier AIModeration: Nathan LabenzJaan Tallinn argues that each 10× jump in training compute carries a 1 to 50 percent chance of ending everything, and that the labs still will not say when they will stop.
1:33:49ENCognitive Revolution "How AI Changes Everything"· 4. Apr. 2023PodcastSafe Only Because It's Weak: A GPT-4 Red-Teamer and Two Founders Debate the PauseModeration: Nathan LabenzNathan Labenz, who spent six months trying to make GPT-4 misbehave, wants labs to stop training anything bigger; Chroma's Anton Troynikov and Lindy's Flo Crivello argue a six-month freeze buys little and may select for the least careful labs.
3:17:51ENLex Fridman· 30. März 2023Podcast · 2 StimmenWhoops is the sound of getting less wrong: Yudkowsky, two weeks after GPT-4Gast: Eliezer Yudkowsky · Moderation: Lex FridmanTwo weeks after GPT-4 launched, Eliezer Yudkowsky sat down with Lex Fridman and started by admitting the transformer stack had already gone further than he expected.Mehr laden24 von 179 angezeigt