#221 - OpenAI Codex, Gemini in Chrome, K2-Think, SB 53

Our 141st episode with a summary and discussion of last week's big AI news, now back with the usual hosts! Read out our text newsletter and comment on the podcast at https://lastweekin.ai/ Email us your questions and feedback at contact@lastweekin.ai Check out our sponsor, the SuperDataScience podcast. You can listen to SDS across all major podcasting platforms (e.g., Spotify, Apple Podcasts, Google Podcasts) plus there’s a video version on YouTube. Timestamps + Links: (00:00) Intro / Banter (01:00) SuperDataScience podcast ad Tools & Apps (02:55) Adobe is upgrading Photoshop’s generative AI model — and releasing more for Illustrator and Express (07:52) Character.AI introduces group chats where people and multiple AIs can talk to each other (14:15) Kendall Jenner becomes "Billie": Meta's AI celebrities face more resistance than enthusiasm (18:22) Dropbox redesigns its web interface and releases AI-powered Dash in open beta Applications & Business (20:22) Google Cloud pledges a ‘shared fate,’ offering legal indemnification for customers (25:00) Chinese search engine company Baidu unveils Ernie 4.0 AI model, claims that it rivals GPT-4 (31:05) Key Arm China Staff Quit to Create Government-Backed Startup (33:30) AMD to acquire AI software startup in effort to catch Nvidia (34:47) Microsoft Will Pay You $15,000 If You Get Bing AI to Go Off the Rails (37:20) ChatGPT’s mobile app hit record $4.58M in revenue last month, but growth is slowing (38:46) TSMC: ecosystem for 2nm chip development is nearing completion (41:50) OpenAI has quietly changed its core values, and being 'thoughtful' and 'audacious' no longer makes the cut Research & Advancements (46:35) LLark: A Multimodal Foundation Model for Music (50:20) TimeGPT-1 (55:25) How FaR Are Large Language Models From Agents with Theory-of-Mind? (01:00:17) HyperAttention: Long-context Attention in Near-Linear Time (01:04:33) PaLI-3 Vision Language Models: Smaller, Faster, Stronger (01:05:42) A Long Way to Go: Investigating Length Correlations in RLHF (01:08:20) Understanding the Effects of RLHF on LLM Generalisation and Diversity Policy & Safety (01:12:05) No Fakes Act wants to protect actors and singers from unauthorized AI replicas (01:15:03) Ukrainian AI attack drones may be killing without human oversight (01:19:59) How a billionaire-backed network of AI advisers took over Washington (01:28:38) China targets 50% boost in computing power as AI race with U.S. ramps up (01:31:03) China proposes blacklist of sources used to train generative AI models (01:34:44) Biden to cut China off from more Nvidia chips, expand curbs to more countries (01:37:51) TSMC expects permanent U.S. approval to supply chip tools to its China factory (01:40:34) Southeast Asia eyes hands-off AI rules, defying EU ambitions Synthetic Media & Art (01:44:05) AI Images Detectors Are Being Used to Discredit the Real Horrors of War (01:47:03) Outro See Privacy Policy at https://art19.com/privacy and California Privacy Notice at https://art19.com/privacy#do-not-sell-my-info.

26 Okt 20231h 48min

Our 140th episode with a summary and discussion of last week's big AI news! Read out our text newsletter and comment on the podcast at https://lastweekin.ai/ Email us your questions and feedback at contact@lastweekin.ai Note: the CEO of GitHub disputes "Report: GitHub Copilot Loses an Average of $20 Per User Per Month" Timestamps + Links: (00:00)  Intro / Banter Tools & Apps (03:00) Reka launches Yasa-1, a multimodal AI assistant to take on ChatGPT (06:48) Arc browser’s new AI-powered features combine OpenAI and Anthropic’s models (11:00) ElevenLabs Launches Voice Translation Tool to Break Down Language Barriers for Content (13:11) Canva’s new AI tools automate boring, labor-intensive design tasks (16:32) Adobe previews AI upscaling to make old, fuzzy videos and GIFs look fresh (20:45) Google Bard is gaining a new 'Memory' toggle to remember key details (22:22) Assistant with Bard: A step toward a more personal assistant (24:14) Android 14’s AI-generated wallpapers might be its coolest new feature Applications & Business (24:50) Waymo’s robotaxi service is now available to tens of thousands of people in San Francisco (28:54) Report: GitHub Copilot Loses an Average of $20 Per User Per Month (32:34) TSMC Sales Fell Less Than Feared as AI Demand Offsets Slump (34:54) Microsoft could debut its AI chip next month: Report (37:06) Exclusive: ChatGPT-owner OpenAI is exploring making its own AI chips (39:08) Google announces new generative AI search capabilities for doctors Projects & Open Source (40:05) Replit’s new AI Model now available on Hugging Face (43:46) Introducing Stable LM 3B: Bringing Sustainable, High-Performance Language Models to Smart Devices (46:30)  Protesters Decry Meta’s “Irreversible Proliferation” of AI Research & Advancements (52:10) Scaling up learning across many different robot types (59:35) Decomposing Language Models Into Understandable Components (01:06:54) Promptbreeder: Self-Referential Self-Improvement Via Prompt Evolution (01:11:45) China's First 28nm Lithography Tool to Be Delivered This Year (01:15:03) LLMs can’t self-correct in reasoning tasks, DeepMind study finds (01:18:07) Fine-tuning Aligned Language Models Compromises Safety, Even When Users Do Not Intend To! Policy & Safety (01:22:08) RISC-V technology emerges as battleground in US-China tech war (01:26:48) Meta and X questioned by lawmakers over lack of rules against AI-generated political deepfakes (01:33:44) Five Takeaways From Bellwether AI Copyright Case (01:36:19) A.I. Could Soon Need as Much Electricity as an Entire Country (01:38:12) Governments race to regulate AI tools (01:40:09) US curbs on chip tools to China nearly finalized, government posting shows Synthetic Media & Art (01:45:35) WGA Ratifies Three-Year Deal With Studios, Officially Ending Hollywood Strike (01:47:50)  AI Watermarks Are No Match for Attackers (01:50:00) Stable Signature: A new method for watermarking images created by open source generative AI (01:51:36) How an AI deepfake ad of MrBeast ended up on TikTok (01:53:53) Disney’s Loki faces backlash over reported use of generative AI See Privacy Policy at https://art19.com/privacy and California Privacy Notice at https://art19.com/privacy#do-not-sell-my-info.

17 Okt 20231h 57min

Our 139th episode with a summary and discussion of last week's big AI news, now back with the usual hosts! Read out our text newsletter and comment on the podcast at https://lastweekin.ai/ Email us your questions and feedback at contact@lastweekin.ai Check out our sponsor, the SuperDataScience podcast. You can listen to SDS across all major podcasting platforms (e.g., Spotify, Apple Podcasts, Google Podcasts) plus there’s a video version on YouTube. Timestamps + Links: (00:00) Intro / Banter (01:30) SuperDataScience podcast ad (03:07) Response to listener comments / corrections Tools & Apps (05:59) ChatGPT goes multimodal: now supports voice, image uploads (10:44) ChatGPT can now search the web in real time (13:30) Meet the A.I. Jane Austen: Meta Weaves A.I. Throughout Its Apps (18:26) Meta's new AI chatbot trained on public Facebook and Instagram posts (19:32) Windows 11’s next big update is now available with Copilot, AI-powered Paint, and more (20:21) Adobe launches Photoshop’s web version with Firefly-powered AI tools (21:32) YouTube Unveils Suite of AI-Powered Tools For Video Creators (23:20) Google is opening up its generative AI search experience to teenagers Applications & Business (25:00) AI startup Lamini bets future on AMD's Instinct GPUs (29:28) Amazon to Invest Up to $4 Billion in AI Startup Anthropic  (36:24) China to Challenge ASML with a better technology than EUV (40:00) Intel says newest laptop chips, software will handle generative AI (43:03) Getty made an AI generator that only trained on its licensed images (46:02) AI Startup Writer Pens $100M Round (46:40) AI chip startup Kneron secures $49M investment (48:45) AI startup AlphaSense valued at $2.5 billion after latest funding round (51:01) A Silicon Valley Supergroup Is Coming Together to Create an A.I. Device (52:42) Microsoft Cloud hiring to "implement global small modular reactor and microreactor" strategy to power data centers (55:24) Google adds a switch for publishers to opt out of becoming AI training data Projects & Open Source (56:11) Mistral AI makes its first large language model free for everyone (01:01:12) Adept AI Labs Open-Sources Persimmon-8B: A Powerful Fully Permissively-Licensed Language Model with <10 Billion Parameters (01:03:20) BTLM-3B-8K: 7B Parameter Performance in a 3B Parameter Model Research & Advancements (01:07:10) AnyMAL: An Efficient and Scalable Any-Modality Augmented Language Model (01:09:35) DreamGaussian: Generative Gaussian Splatting for Efficient 3D Content Creation / Text-to-3D using Gaussian Splatting (01:12:40)  Google DeepMind AI tool assesses DNA mutations for harm potential (01:15:38) Prompt engineering for Claude's long context window (01:17:26) John Carmack and Rich Sutton partner to accelerate development of Artificial General Intelligence Policy & Safety (01:19:46) Top GOP senator teams up with key Dem on ‘light-touch’ AI bill (01:24:03) AI facial recognition tech leads to wave of lawsuits from Black plaintiffs after mistaken identities end in arrests (01:26:20) California governor vetoes bill banning robotrucks without safety drivers (01:27:47) Fake naked pictures of young girls created with AI spark fury in a small Spanish town (01:29:00) AI Security Center to Open at National Security Agency (01:30:36) We Can Prevent AI Disaster Like We Prevented Nuclear Catastrophe   Synthetic Media & Art (01:33:19) Hollywood’s Deal With Screenwriters Just Rewrote the Rules Around A.I. (01:35:10) Indian actor Anil Kapoor wins court battle over AI use of his likeness (01:36:24) Tom Hanks warns dental plan ad image is AI fake (01:37:36) Amazon restricts authors from self-publishing more than three books a day after AI concerns (01:39:59) Project Gutenberg has implemented one of the worst AI fears of striking actors (01:42:17) Meta’s AI stickers are here and already causing controversy (01:45:22) Franzen, Grisham and Other Prominent Authors Sue OpenAI (01:48:55) OpenAI and ChatGPT Lawsuit List (01:50:16) Outro See Privacy Policy at https://art19.com/privacy and California Privacy Notice at https://art19.com/privacy#do-not-sell-my-info.

10 Okt 20231h 51min

Our 138th episode with a summary and discussion of last week's big AI news, with guest host Jon Krohn of the the SuperDataScience podcast!! Note: this one is coming out a week late, but we'll be back on schedule going forward! Read out our text newsletter and comment on the podcast at https://lastweekin.ai/ Email us your questions and feedback at contact@lastweekin.ai Timestamps + Links: (00:00) Intro / Banter (04:30) Preview of news Tools & Apps (05:40) DALLE-3 (12:47) YouAgent (Richard Socher thread) Applications & Business (20:03) Google's Gemini AI surpasses Chat GPT-4 fivefold: Report (25:16) Google begins external testing of its GPT-4 competitor "Gemini" (26:30) From sanctions to silicon: China’s semiconductor breakthrough (35:41) AI app Character.ai is catching up to ChatGPT in the U.S. (41:19) SoftBank planning OpenAI investment: Financial Times (44:30) OpenAI Hustles to Beat Google to Launch ‘Multimodal’ LLM (46:40) The AI Detection Arms Race Is On (51:30) Nvidia's dominance in AI chips deters funding for startups (55:15) AI chip startup Enfabrica raises $125 mln, with backing from Nvidia (56:41) Databricks raises over $500 mln at $43 bln valuation Projects & Open Source (01:01:15) Meta Is Developing a New, More Powerful AI System as Technology Race Escalates Research & Advancements (01:10:19) NExT-GPT: Any-to-Any Multimodal LLM (01:15:35) On measuring situational awareness in LLMs (01:22:50) Language Modeling Is Compression (01:26:22) How AI can Revolutionize Science  (Jeremie’s gift link) Policy & Safety (01:33:33) Inside the Senate’s Private AI Meeting With Tech’s Billionaire Elites (01:36:08) Anthropic's Responsible Scaling Policy (01:41:45) Rishi Sunak considers banning Chinese officials from half of AI summit (01:43:48) EU to let ‘responsible’ AI startups train models on its supercomputers (01:44:26) Biden-⁠Harris Administration Secures Voluntary Commitments from Eight Additional Artificial Intelligence Companies to Manage the Risks Posed by AI     (01:45:23) Microsoft president and Nvidia chief scientist to testify in Senate AI hearings  (01:46:40) AI Chatbots Are Invading Your Local Government—and Making Everyone Nervous (01:47:27) Inside Elon Musk's Struggle for the Future of AI (01:50:18) Gary Gensler confirms SEC’s use of AI for financial surveillance Synthetic Media & Art (01:50:55) Self-publishers must declare if content sold on Amazon’s site is AI-generated  See Privacy Policy at https://art19.com/privacy and California Privacy Notice at https://art19.com/privacy#do-not-sell-my-info.

3 Okt 20231h 55min

Our 137th episode with a summary and discussion of last week's big AI news! With guest host Jessica Dai. Check out her Reboot publication! Read out our text newsletter and comment on the podcast at https://lastweekin.ai/ Email us your questions and feedback at contact@lastweekin.ai Check out our sponsor, the SuperDataScience podcast. You can listen to SDS across all major podcasting platforms (e.g., Spotify, Apple Podcasts, Google Podcasts) plus there’s a video version on YouTube. Timestamps + Links: (00:00) Intro  (01:37) SuperDataScience podcast ad Tools & Apps (02:32) Salesforce launches AI assistant across its apps including Slack and Tableau (06:06) Roblox’s new AI chatbot will help you build virtual worlds (09:22) Bybit debuts AI-powered ‘TradeGPT’ for market analysis and data driven Q&A (11:14) China’s Ant Group unveils finance AI model as race heats up Applications & Business (13:24) TSMC warns AI chip crunch will last another 18 months (16:50) Even AI Hasn’t Helped Microsoft’s Bing Chip Away at Google’s Search Dominance (20:05) OpenAI will host its first developer conference on November 6 (22:54) AI reading coach startup Ello raises $15M to bolster child literacy (25:71) Nasdaq gets SEC nod for first exchange AI-driven order type (27:17) NVIDIA Adds New Software That Can Double H100 Inference Performance Projects & Open Source (30:10) Meta Platforms reportedly building open-source generative AI system to rival OpenAI’s ChatGPT (34:13) Indian Developers Top Hugging Face Leaderboard with GenZ 70B (38:31) Supporting the Open Source AI Community Research & Advancements (40:31) Artificial intelligence allows paralysis patient to speak for first time in 18 years (45:42) Efficient Benchmarking (of Language Models) (49:31) Online AI-based test for Parkinson’s disease severity shows promising results (52:52) Scientists used machine learning to perform quantum error correction Policy & Safety (56:18) Tech leaders including Musk, Zuckerberg call for government action on AI (01:01:32) US court rules that artificial intelligence generated artwork cannot be copyrighted (01:04:00) 2 Senators Propose Bipartisan Framework for A.I. Laws (01:05:47) Transcript: US Senate Judiciary Hearing on Oversight of A.I. (01:06:02) U.S. Copyright Office Invites Public To Comment On AI Synthetic Media & Art (01:09:10) Venice Film Festival 2023 Is Cinematic AI’s Coming Out Party (01:13:52) Revolution Software is using their own AI technology to remake Broken Sword (01:16:24) As his Kickstarter passes $1.3M, publisher defends Terraforming Mars’ generative AI art: ‘It’s too powerful a technology’ (01:17:23) Outro See Privacy Policy at https://art19.com/privacy and California Privacy Notice at https://art19.com/privacy#do-not-sell-my-info.

20 Sep 20231h 18min

Our 136th episode with a summary and discussion of last week's big AI news! With guest host Daniel Bashir. Check out his AI interview podcast! Read out our text newsletter and comment on the podcast at https://lastweekin.ai/ Email us your questions and feedback at contact@lastweekin.ai Check out our sponsor, the SuperDataScience podcast. You can listen to SDS across all major podcasting platforms (e.g., Spotify, Apple Podcasts, Google Podcasts) plus there’s a video version on YouTube. Timestamps + links: (00:00) Intro  (01:15) SuperDataScience Ad (01:51) Response to listeners Tools & Apps (02:47) Anthropic’s Claude AI chatbot gets a paid plan for heavy users (04:36) Watch out, Midjourney! Ideogram launches AI image generator with impressive typography (06:37) Intuit launches generative AI–powered digital assistant for small businesses and consumers (07:17) Zoom Is Jumping on the AI Chatbot Bandwagon Applications & Business (08:47) China lets Baidu, others launch ChatGPT-like bots to public, tech shares jump (11:23) Tencent releases AI model for businesses as competition in China heats up (12:00) Microsoft says it will take the heat if Copilot AI users get sued (14:52) How We Chose the TIME100 Most Influential People in AI (17:30) ChatGPT creator OpenAI is reportedly earning $80M a month (19:16) AI chip startup d-Matrix raises $110 mln with backing from Microsoft (21:00) ThetaRay nabs $57M for AI tools to ID and fight money laundering (22:18) Sapeon raises $46m for AI chips (23:20) Imbue raises $200M to build AI models that can ‘robustly reason’ Projects & Open Source (25:48) Announcing the commercial relicensing and expansion of DINOv2, plus the introduction of FACET (27:02) UAE launches Arabic large language model in Gulf push into generative AI (29:23) New Open-Source ‘Falcon’ AI Language Model Overtakes Meta and Google Research & Advancements (31:01) Qwen-VL: A Frontier Large Vision-Language Model with Versatile Abilities (33:57) RLAIF: Scaling Reinforcement Learning from Human Feedback with AI Feedback (36:22) Perception, performance, and detectability of conversational artificial intelligence across 32 university courses (39:28) SyncDreamer: Generating Multiview-consistent Images from a Single-view Image Policy & Safety (41:00) US curbs AI chip exports from Nvidia and AMD to some Middle East countries (42:55) China suspected of using AI on social media to sway US voters, Microsoft says (46:37) Trusting A.I.-written mushroom hunting guides sold on Amazon could get you killed. But like deadly fungi, identifying them is tricky (48:25) Ads for AI sex workers are flooding Instagram and TikTok (50:30) The UK releases key ambitions for global AI summit Synthetic Media & Art (51:55) AI Took the Stage at the World's Largest Arts Festival. Here's What Happened (54:12) Ghostwriter Returns With an A.I. Travis Scott Song, and Industry Allies (56:17) Artists sign open letter saying generative AI is good, actually (58:57) The latest canvas for Refik Anadol’s AI-generated art? The new Sphere in Las Vegas (01:01:20) Outro See Privacy Policy at https://art19.com/privacy and California Privacy Notice at https://art19.com/privacy#do-not-sell-my-info.

10 Sep 20231h 2min

Our 135th episode with a summary and discussion of last week's big AI news! Apologies for pod being late again... -Andrey Read out our text newsletter and comment on the podcast at https://lastweekin.ai/ Email us your questions and feedback at contact@lastweekin.ai Check out our sponsor, the SuperDataScience podcast. You can listen to SDS across all major podcasting platforms (e.g., Spotify, Apple Podcasts, Google Podcasts) plus there’s a video version on YouTube. Timestamps + links: (00:00) Intro / Banter (02:07) SuperDataScience podcast ad (04:00) Shout out to listener comments & reviews (05:43) News Preview Tools & Apps (06:32) Google’s Duet AI is now available in Gmail, Docs, and more for $30 a month (10:51) Poe’s new desktop app lets you use all the AI chatbots in one place (13:47) Developers are now using AI for text-to-music apps (15:50) Yahoo Mail Debuts AI Enhancements for a Smarter Inbox (18:05) South Korea's Naver launches generative AI services Applications & Business (21:19) Introducing ChatGPT Enterprise (25:25) We Analyzed Millions of ChatGPT User Sessions: Visits are Down 29% since May, Programming Assistance is 30% of Use (29:50) Survey finds relatively few Americans actually use (or fear) ChatGPT (31:32) Nvidia’s Q2 earnings prove it’s the big winner in the generative AI boom (33:30) How Nvidia Built a Competitive Moat Around A.I. Chips (37:03) Nvidia's $25 billion buyback 'a head-scratcher' for some shareholders (40:25) Chinese AI Company Claims Huawei AI GPUs Are On Par With NVIDIA A100, Will Compete With GPT-4 LLM In 2024 (44:08) AI’s Share Of US Startup Funding Doubled In 2023 (45:42) Hugging Face raises $235M from investors including Salesforce and Nvidia (48:13) AI21 Labs Raises $155M in Series C Funding; Valued at $1.4 Billion (50:32) DP Technology Raises 100 million US dollars, Achieves AI Milestones in Science Projects & Open Source (52:15) Meta releases an AI model that can transcribe and translate close to 100 languages (56:01) Line Open-Sources ‘japanese-large-lm’: A Japanese Language Model With 3.6 Billion Parameters Research & Advancements (58:55) Consciousness in Artificial Intelligence: Insights from the Science of Consciousness (01:05:18) Reinforced Self-Training (ReST) for Language Modeling (01:08:27) Hoodwinked: Deception and Cooperation in a Text-Based Game for Language Models (01:11:47) Head-to-Tail: How Knowledgeable are Large Language Models (LLM)? A.K.A. Will LLMs Replace Knowledge Graphs? (01:14:14) Nvidia AI Image Generator Fits on a Floppy Disk and Takes 4 Minutes to Train (01:16:18) MagicEdit: High-Fidelity and Temporally Coherent Video Editing (01:17:44) A Survey on Large Language Model based Autonomous Agents Policy & Safety (01:19:00) UK to spend £100m in global race to produce AI chips (01:22:03) Meta confirms AI ‘off-switch’ incoming to Facebook, Instagram in Europe (01:25:44) Beijing to restrict use of generative AI in online healthcare activities (01:28:04) Schumer to host AI forum with CEOs including Elon Musk and Mark Zuckerberg (01:32:52) Bletchley Park to host AI safety talks in November (01:34:47) Spain launches AI regulation agency in bid to become industry leader Synthetic Media & Art (01:36:11) Google made a watermark for AI images that you can’t edit out (01:39:20) Outro See Privacy Policy at https://art19.com/privacy and California Privacy Notice at https://art19.com/privacy#do-not-sell-my-info.

6 Sep 20231h 40min

A special non-news episode in which Andrey and Jeremie discussion AI X-Risk! Please let us know if you'd like use to record more of this sort of thing by emailing contact@lastweekin.ai or commenting whether you listen. Outline: (00:00) Intro (03:55) Topic overview (10:22) Definitions of terms (35:25) AI X-Risk scenarios (41:00) Pathways to Extinction (52:48) Relevant assumptions (58:45) Our positions on AI X-Risk (01:08:10) General Debate (01:31:25) Positive/Negative transfer (01:37:40) X-Risk within 5 years  (01:46:50) Can we control an AGI (01:55:22) AI Safety Aesthetics (02:00:53) Recap (02:02:20) Outer vs inner alignment (02:06:45) AI safety and policy today (02:15:35) Outro Links Taxonomy of Pathways to Dangerous AI Clarifying AI X-risk Existential Risks and Global Governance Issues Around AI and Robotics Current and Near-Term AI as a Potential Existential Risk Factor AI x-risk, approximately ordered by embarrassment Classification of Global Catastrophic Risks Connected with Artificial Intelligence X-Risk Analysis for AI Research The Alignment Problem from a Deep Learning PerspectiveSee Privacy Policy at https://art19.com/privacy and California Privacy Notice at https://art19.com/privacy#do-not-sell-my-info.

30 Aug 20232h 16min