Ergebnisse
119 ArtikelSortieren
42:08EN
Sequoia Capital· 30. Juli 2024PodcastMeta Trained a 405B Teacher on 16,000 GPUs — Then Told the World to Distill ItTwo days after Llama 3.1 405B shipped, Meta's generative-AI product lead walks Sequoia through why the most expensive model in the family is free, and why he thinks the frontier itself is heading toward commodity.
13:40ENTesting Llama 3.1 8B: Benchmark Gains and Mixed Practical ResultsMatthew Berman · Forward FutureMatthew Berman tests coding, reasoning and other prompts on the small model and finds several failures despite stronger reported benchmarks.
1:54:01EN
Cognitive Revolution "How AI Changes Everything"· 26. Juni 2024PodcastA Tiny Network Grew a Salt Crystal the Training Data Never ShowedModeration: Nathan LabenzTim Duignan trained tens of thousands of parameters on salt water. The model invented a crystal — and a continuum solvent theory he had spent a PhD failing to write by hand.
16:06ENAI News: Video Generation, Apple Intelligence, Robotics and DeepSeek Coder V2Matthew Berman · Forward FutureMatthew Berman reviews a week of launches and research, from Luma and Runway to NVIDIA, robotics and coding models.
9:53ENTesting Microsoft Phi-3 Medium Locally: Mixed Coding and Reasoning ResultsMatthew Berman · Forward FutureMatthew Berman runs a quantized Phi-3 Medium on a MacBook and explores coding, arithmetic and reasoning prompts.
12:50ENTesting Smaug: Hosted 70B Versus a Smaller Local ModelMatthew Berman · Forward FutureMatthew Berman compares two Smaug configurations on Snake, math and reasoning puzzles.
1:42:17ENCognitive Revolution "How AI Changes Everything"· 20. Apr. 2024PodcastThe Cheapest Inference Provider Is the Most DangerousModeration: Nathan LabenzFireworks AI cofounders Lin Qiao and Dmytro Ivchenko explain why subsidized GPU bills vanish, how LoRA serving can cut cost a hundredfold, and why OpenAI is the rival that actually matters.
15:01ENTesting Llama 3 at Launch: Code, Reasoning and Image GenerationMatthew Berman · Forward FutureMatthew Berman tries Meta AI’s new interface on a snake game, logic puzzles and rapid image generation.
14:23ENTesting DBRX: Coding Strengths and Reasoning GapsMatthew Berman · Forward FutureMatthew Berman examines Databricks’ open-weight model and puts it through coding, arithmetic and reasoning prompts.
14:11ENLM Studio in 2024: Local Models, Parallel Inference and a Local APIMatthew Berman · Forward FutureMatthew Berman demonstrates model downloads, hardware settings, JSON output and serving multiple local models.
2:47:17EN
Lex Fridman· 7. März 2024Podcast · 2 StimmenIf You Want Human-Level AI, Abandon Generative ModelsGast: Yann Lecun · Moderation: Lex FridmanYann LeCun tells Lex Fridman why autoregressive LLMs cannot plan, remember, or understand the physical world — and why open source is the only democratic answer.
23:00ENTesting Google Gemma 7B at Launch: Setup MattersMatthew Berman · Forward FutureMatthew Berman compares local and hosted tests of Google’s early open-weight model.
3:37ENMark Zuckerberg Announces Llama 3 Training and Meta’s AI PlansMatthew Berman, Mark Zuckerberg · Forward FutureIn January 2024, Matthew Berman reacts to Zuckerberg’s remarks on general intelligence, compute and AI devices.
19:10EN
No Priors: AI, Machine Learning, Tech, & Startups· 11. Jan. 2024Podcast · 2 StimmenThe World's Only Capped-Profit CompanyModeration: Sarah Guo, Elad GilNo Priors looks back at 2023 through eight conversations on money, intelligence, jobs, and who should put the guardrails on a model.
2:16:27ENCognitive Revolution "How AI Changes Everything"· 12. Dez. 2023PodcastSuperhuman Go Has a Blind Spot. Balaji Bets on a War of the GodsModeration: Nathan LabenzBalaji Srinivasan tells Nathan Labenz why adversarial inputs, missing actuators, and cryptography make a single runaway AI god less likely than a contest among many that still need priests.
32:57ENNo Priors: AI, Machine Learning, Tech, & Startups· 9. Nov. 2023Podcast · 2 StimmenA 7-billion-parameter model on a MacBook: Mistral's CEO reruns the cost of AIModeration: Sarah Guo, Elad GilArthur Mensch tells No Priors why Chinchilla made small models valuable, why open-sourcing Mistral 7B is not a bioweapon risk, and why guardrails belong on the application.
4:42ENConnecting AutoGen to a Local Model With LM StudioMatthew Berman · Forward FutureMatthew Berman demonstrates an early local-model setup and explains where agent behavior still needs work.
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.
1:04:38ENLex Fridman· 28. Sept. 2023Podcast · 2 StimmenHundreds of Miles Apart, Zuckerberg and Fridman Sat Close Enough to See the WrinklesGast: Mark Zuckerberg · Moderation: Lex FridmanLex Fridman sat in a dark room in Austin wearing a Quest headset; Mark Zuckerberg sat in another conference room, also wearing one, and it did not feel like a video call.
29:38ENNo Priors: AI, Machine Learning, Tech, & Startups· 21. Sept. 2023Podcast · 2 StimmenA Beginner Just Shipped a $250K Business. Professionals Got 30 Percent.Moderation: Sarah Guo, Elad GilReplit CEO Amjad Masad on Ghostwriter, why a 3-billion-parameter model was enough, and the agents he wants earning bounties while you sleep.
30:03ENNo Priors: AI, Machine Learning, Tech, & Startups· 7. Sept. 2023Podcast · 2 StimmenA Dinner-Plate Chip Cut Training From 60 Days to 3.5Moderation: Sarah Guo, Elad GilCerebras CEO Andrew Feldman tells No Priors why wafer-scale silicon, not another GPU, is the answer to the compute crunch — and why most production models will stay at 13 billion parameters or below.
52:48ENCognitive Revolution "How AI Changes Everything"· 25. Aug. 2023PodcastOpenAI Cut Fine-Tuning 90 Percent, and Founders Crossed Llama 2 Off the ListModeration: Nathan LabenzNathan Labenz and Erik Torenberg walk through GPT-3.5 Turbo fine-tuning, why self-hosting still loses on total cost, and what is actually working in code, school, and medicine.
12:20ENTesting UltraLM 13B: A Small Model’s Strengths and Reasoning ErrorsMatthew Berman · Forward FutureA 2023 local test finds useful writing and simple coding alongside failures in games, arithmetic and logic.
8:17ENUsing OpenLLM to Serve Open Models Through an APIMatthew Berman · Forward FutureMatthew Berman demonstrates a local model server, its interactive API and a Python client in a 2023 walkthrough.
Mehr laden24 von 119 angezeigt