Chaînes mises à jour récemment

Choisissez-en une pour ne voir qu’elleToutes les chaînes →

Derniers articles

4 462 articles
Trier
3:20:00EN
Lex Fridman· 4 mars 2024Podcast · 2 voixThe last empire is still falling: Serhii Plokhy on Ukraine, Russia, and 1991Invité : Serhii Plokhy · Anime : Lex FridmanHarvard historian Serhii Plokhy tells Lex Fridman why the war in Ukraine is the unfinished collapse of the Russian Empire — not a sudden break with history.GeopoliticsPolitical PhilosophyHistory
46:13EN
20VC with Harry Stebbings· 4 mars 2024Podcast · 2 voixNicolai Tangen on Long-Term Investing and Leading Norway’s Sovereign FundInvité : Nicolai Tangen · Anime : Harry StebbingsThe fund chief discusses market uncertainty, transparency, climate risk and the difference between patient investing and fast decisions.InvestingLeadershipDecision Making+1
5 diapositives40:19EN
Matthew Berman· 3 mars 2024Musk's 2024 OpenAI Lawsuit: The Allegations, History and Later ProceedingsMatthew Berman · Forward FutureMatthew Berman examines Musk's complaint about OpenAI's mission and governance. Its claims about AGI and the 2023 board crisis were allegations, not court findings.OpenAIAI SafetyAI History+2
15:58EN
Matthew Berman· 2 mars 2024Elon Musk on Self-Driving, AI Compute and Electricity ConstraintsElon Musk, Matthew Berman · TeslaA 2024 Bosch interview and Matthew Berman’s commentary explore autonomous vehicles, robots and the infrastructure needed for AI.Autonomous DrivingAGI TimelinesRobotics+2
1:08:17EN
20VC with Harry Stebbings· 1 mars 2024Podcast · 2 voixFrank Quattrone on Technology M&A, IPOs and the Amazon Pricing DebateInvité : Frank Quattrone · Anime : Harry StebbingsThe Qatalyst founder recounts deals involving Qualtrics, LinkedIn and Amazon, and explains how markets, financing and regulation shape transactions.Acquisition StrategyVenture CapitalInvesting+2
5 diapositives9:15EN
Matthew Berman· 1 mars 2024Figure 01 and Other Robot Demos: What the 2024 Clips ShowMatthew Berman · Forward FutureMatthew Berman reviews a humanoid making coffee, inexpensive robot arms and other robotics demonstrations. The clips show specific tasks, not proof of artificial general intelligence.RoboticsAI agentsReinforcement Learning+1
0:27EN
20VC with Harry Stebbings· 1 mars 2024Podcast · 2 voixFrank Quattrone on Founder Payments in AcquisitionsInvité : Frank Quattrone · Anime : Harry StebbingsA banker explains cash closings, incentives and stock restrictions.
Diapositives1:26:51EN
All-In Podcast· 1 mars 2024Podcast · 4 voixGoogle’s AI Challenge, Klarna’s Support Automation and Reddit’s IPOAnime : Jason Calacanis, Chamath Palihapitiya, David Sacks, David FriedbergThe All-In hosts discuss Gemini’s image controversy, markets for training data and early customer-service automation, before reviewing Reddit’s 2023 filing.AI StrategyEntreprise et stratégieInvesting
1:32:19EN
Cognitive Revolution "How AI Changes Everything"· 29 févr. 2024PodcastHis 2018 Prompting Paper Was Rejected. His Search API Now Beats GoogleAnime : Nathan LabenzRichard Socher traces You.com from a rejected unified NLP model to an LLM-native web index, three assistant modes, and a very low P(doom).Produit et UXGénération augmentée par récupérationSearch & retrieval+3
3 diapositives11:07EN
Jane Street· 29 févr. 2024How OCaml Locals Use a Separate StackRichard Eisenberg · Jane StreetA list example shows how region-scoped local values can avoid repeated garbage-collected allocations.Programming LanguagesSoftware engineering
39:03EN
No Priors: AI, Machine Learning, Tech, & Startups· 29 févr. 2024Podcast · 2 voixMI300 Leads Inference by 2x: AMD's CTO on Winning Chip Competition Without a Walled GardenAnime : Sarah Guo, Elad GilWhen Mark Papermaster became AMD's CTO in fall 2011, nearly all of the company's revenue sat on PCs and gaming, Moore's Law was slowing, and everyone knew it needed a turnaround.Infrastructure et déploiementInferenceAI cost & performance+3
7 diapositives16:27EN
Matthew Berman· 29 févr. 2024Matthew Berman Reviews EMO’s Audio-Driven Portrait Video ResearchMatthew Berman, Jensen Huang · Forward FutureAlibaba researchers animate a portrait from an image and vocal audio; Berman examines the method, its limitations and the implications for trust in video.Generative mediaVision et vidéoDeep Learning+1
1:13:37EN
20VC with Harry Stebbings· 28 févr. 2024Podcast · 2 voixSami Inkinen on Success, Family and a Founder’s IdentityInvité : Sami Inkinen · Anime : Harry StebbingsThe Trulia co-founder and Virta Health CEO reflects on the anxiety that followed an IPO, rowing across the Pacific with his wife and building a life beyond work.FoundersPersonal GrowthRelationships+2
7 diapositives21:41EN
Matthew Berman· 28 févr. 2024Google Genie and the Idea of Generated Interactive WorldsMatthew Berman · Forward FutureMatthew Berman explores early world-model research, generated media and simulated agents while imagining a future of personalized games.World modelsGenerative mediaRobotics+2
3:04:22EN
Lex Fridman· 27 févr. 2024Podcast · 2 voixA Ten-Page Memo Told Tucker Carlson Not to Interview PutinInvité : Tucker Carlson · Anime : Lex FridmanAfter eight days in Moscow, Carlson tells Lex Fridman why he let Vladimir Putin talk, what Navalny’s death does and does not prove, and why he thinks Washington is blocking a settlement in Ukraine.GeopoliticsFree SpeechPolitical Philosophy+3
1:50:09EN
Cognitive Revolution "How AI Changes Everything"· 27 févr. 2024PodcastNine Hundred Occupations, Twenty Years, and Still No AI RevolutionAnime : Nathan LabenzEconomist Robin Hanson tells Cognitive Revolution why each new AI demo looks like the last one, and why wages and headcount are the tests that matter.Entreprise et stratégieAGI TimelinesAI Strategy+3
5 diapositives13:57EN
Matthew Berman· 27 févr. 2024Mistral Large: A 2024 Review of Coding, Reasoning and Safety ResponsesMatthew Berman · Forward FutureMatthew Berman reviews Mistral’s launch claims and tests the model with a small set of prompts.Modèles et entraînementBenchmarksAI Safety+1
1:29:20EN
20VC with Harry Stebbings· 26 févr. 2024Podcast · 2 voixJustin Ishbia on Private Equity, Building Teams and the Benefits of ScaleInvité : Justin Ishbia · Anime : Harry StebbingsThe Shore Capital founder explains his buyout strategy, how he develops leaders and why he treats operations as a system.InvestingCompany BuildingLeadership+2
10 diapositives30:39EN
Matthew Berman· 25 févr. 2024NVIDIA’s GEAR Research, Gemini 1.5 and Fast AI InferenceMatthew Berman · Forward FutureMatthew Berman reviews embodied-agent research and a February 2024 set of AI releases and demonstrations.RoboticsAI researchBenchmarks+2
4 diapositives55:29EN
Matthew Berman· 24 févr. 2024Building a CrewAI Newsletter Team: Scraping, Summarizing and DebuggingMatthew Berman, Joe · Forward FutureMatthew Berman builds a two-agent newsletter prototype, works through input and scraping errors, and shows what still needs to be fixed.Multi-agent systemsIngénierie d’agentsWorkflow automation+2
2:00:03EN
Cognitive Revolution "How AI Changes Everything"· 23 févr. 2024PodcastPatients Rated GPT-4 More Empathetic Than DoctorsAnime : Nathan LabenzAzeem Azhar of Exponential View maps the AI S-curve for Nathan Labenz: why the past always looks flat, why empathy went first, and what a million times more compute would actually change.Entreprise et stratégieAI StrategyProduct Strategy+3
2 diapositives18:40EN
Matthew Berman· 23 févr. 2024Rabbit R1 Before Release: Questions About Cost, Actions and PrivacyMatthew Berman, Jesse Lyu, Jason Calacanis · Forward FutureIn a Rabbit-sponsored February 2024 video, Matthew Berman relays the company's answers and interviews its founder through excerpts.Personal agentsAI interface designPrivacy & data governance+1
0:51EN
20VC with Harry Stebbings· 23 févr. 2024PodcastDoug Leone’s Interview Question About Self-AwarenessInvité : Doug LeoneA personal interviewing practice and its limitations.HiringLeadership
31:28EN
No Priors: AI, Machine Learning, Tech, & Startups· 22 févr. 2024Podcast · 2 voixBolt Memory onto GPT-4 and Cut Hallucinations by up to 50%Anime : Sarah Guo, Elad GilPinecone CEO Edo Liberty tells Sarah Guo and Elad Gil how RAG, serverless vectors, and keeping knowledge outside the model make LLMs more reliable.RAG et connaissancesGénération augmentée par récupérationVector search+3
Charger plus24 sur 4 462 affichés