This Day in AI Podcast

Meta's Code Llama, an AI code generator challenging OpenAI's dominance, and the implications of AI models training themselves. Experiments with multilingual speech synthesis to generate a fake phishing call on our mother. Deep dive into the evolution of GPT models, fine-tuning GPT-3.5 Turbo announced by OpenAI, and the potential of AI-generated unit tests for code. Exploration of 11 labs' voice cloning technology and its practical applications. Discussion on phishing pranks and hardware investments for AGI development.

This week your favorite AI bros go deep on the BIG LIES - how big tech and the mainstream narrative are trying to SILENCE revolutionary AI models that threaten to EXPOSE inconvenient truths!  Tune in as Chris and Michael SHRED Meta's attempt to gag their new science AI Galactica, and discuss the CENSORSHIP built into aligned models like Claude and GPT-4.  Don't miss their hilarious takedown of noted AI alarmist Gary Marcus - his latest flip-flop proves generative AI is HERE TO STAY! The truth will not be suppressed!(Note description written by AI for lols)Thanks for helping us reach our goal of 100+ reviews on Apple Podcasts. It means a lot to us!CHAPTERS=====00:00 - What is a Poop?01:08 - Is Generative AI a Dud?23:32 - OpenAI Acquires Global Illumination to work on ChatGPT31:12 - Anthropic Raises $100M from Korean SK Telecom37:15 - LLAMA2 Uncensored: Censorship, Misinformation and the Battle for Truth48:31 - Meta's AI Trained on 48M Science Papers Shut Down After 2 DaysSOURCES=====https://garymarcus.substack.com/p/what-if-generative-ai-turned-outhttps://www.judiciary.senate.gov/imo/media/doc/2023-05-16%20-%20Testimony%20-%20Marcus.pdfhttps://www.sbs.com.au/whats-on/article/where-experts-naysayers-and-everyone-else-saw-the-internet-heading-in-the-90s/l7rfeapbmhttps://garymarcus.substack.com/p/what-exactly-are-the-economics-ofhttps://twitter.com/emollick/status/1690906554747265024?s=20https://techcrunch.com/2023/08/14/ai-startup-anthropic-raises-100m-from-korean-telco-giant-sk-telecom/https://huggingface.co/jarradh/llama2_70b_chat_uncensoredhttps://twitter.com/boriquagato/status/1691800167857750274?s=46&t=uXHUN4Glah4CaV-g2czc6Qhttps://www.cnet.com/tech/computing/ai-and-you-googles-news-ambitions-putting-a-face-on-subway-fare-dodgers/https://openai.com/blog/openai-acquires-global-illumination

This week we gossip about OpenAI's shady web crawling habits, laugh at Zoom's lame excuses for spying, and dream up the perfect AI crypto scam. Get the inside scoop on Nvidia's new trillion-parameter instrument of AI, hear our hot takes on the public's growing AI disillusionment, and find out what an AI HVAC administrator would sound like. Join your favorite AI bros as they dive deep on the latest AI hype and hardware gossip - this episode is chock full of spicy AI tea you won't want to miss!Please consider leaving a review to help us reach 100 reviews if you listen on Apple Podcasts :)CHAPTERS:====00:00 - We Should Totally Do An AI Crypto Scam00:27 - AI Meal Planner Suggests Chlorine Gas Recipe & AI with Personality04:07 - OpenAI's GPTBot for Web Crawling08:47 - Stealing content with AI, How to Protect Your IP from AI14:37 - Zoom's Terms of Service for AI Training Scandal25:25 - Nvidia's GH200 Announcement & Availability of Hardware34:29 - Have We Reached AI Disillusionment?52:54 - Generative AI LLMs for HVAC!?55:31 - Claude Instant Version 1.2 Released57:33 - AudioLDM 2: Text-to-audio/speech generation 1:00:41 - Skeptics Vs Optimists for AI (AI Crypto Bros)SOURCES====https://www.theguardian.com/world/2023/aug/10/pak-n-save-savey-meal-bot-ai-app-malfunction-recipeshttps://saveymeal-bot.co.nz/ingredientshttps://platform.openai.com/docs/gptbothttps://news.ycombinator.com/item?id=37030568https://admiralcloudberg.medium.com/critical-conversations-the-crash-of-eastern-airlines-flight-212-660f47698887https://news.ycombinator.com/item?id=37021160https://www.axios.com/2023/08/09/zooms-terms-service-changes-ai-fearshttps://www.nvidia.com/en-au/data-center/dgx-gh200/https://twitter.com/authority_ai/status/1688619238389379073?s=46&t=uXHUN4Glah4CaV-g2czc6Qhttps://twitter.com/SashaKaletsky/status/1676957007985922051https://twitter.com/emollick/status/1688760539441217536/photo/1https://twitter.com/anthropicai/status/1689303697535414272?s=46&t=uXHUN4Glah4CaV-g2czc6Q

This week we dive into the brave new world of AI agents teaming up to do real work - from building video games to diagnosing patients! But will these digital workforces put humans out of jobs? We discuss the AI takeover of industries like medicine and software, plus exciting updates like AI-generated music and Google giving their Assistant a complete AI makeover. We also cover Meta's Audio Craft, Med-Flamingo, GPT-5 Trademark and Rumors, and The SF Compute Company.Thanks for your likes, comments and support.CHAPTERS:=====00:00 - Self-Diagnosing Medical Problems Is Here00:28 - MetaGPT: Multi-agent Collaborative Framework and The Multi-Agent Future11:39 - Will Agents Replace Software in the Future?14:59 - The flood of new LLMs20:09 - Med-Flamingo: Have Virtual Doctors Arrived?34:06 - Martin Shkreli's Dr. Gupta & Disrupting Medical Diagnosis38:45 - Everyone is Using AI Already for Medical Diagnosis39:29 - Focusing on Higher Level Work40:53 - Meta's Audio Craft: Create Sounds and Music with Open Source AI44:53 - Will Spotify Cut Out Artists to Increase Profits with AI Music?48:07 - Will Entrenched Professionals Slow Down the Benefit of AI?51:37 - ChatGPT Updates: GPT-4 as Default, Suggested Replies, Prompt Examples, Stay Logged In! 57:05 - The San Francisco Compute Group: The A100 Cooperative!1:00:42 - GPT-5 Trademark has been registered! 1:01:48 - Google Assistant Powered by AI LLM Leaked in Letter1:05:02 - The Future of Websites: LLMs for Businesses and BrandsSOURCES:=====https://arxiv.org/pdf/2308.00352v2.pdfhttps://twitter.com/michael_d_moor/status/1685804620730540033?s=46&t=uXHUN4Glah4CaV-g2czc6Qhttps://twitter.com/emollick/status/1686176146700857344?s=46&t=uXHUN4Glah4CaV-g2czc6Qhttps://blueprintcdn.com/wp-content/uploads/2023/07/Blueprint-Discussion-Paper-2023.10-Agarwal-Moehring-Rajpurkar-Salz_2.pdfhttps://twitter.com/sentdex/status/1687123772078247936?s=46&t=uXHUN4Glah4CaV-g2czc6Qhttps://ai.meta.com/blog/audiocraft-musicgen-audiogen-encodec-generative-ai-audio/https://twitter.com/joannejang/status/1687165702275567616?s=46&t=uXHUN4Glah4CaV-g2czc6Qhttps://www.searchenginejournal.com/openai-files-trademark-application-gpt-5/493040/https://www.axios.com/2023/07/31/google-assistant-artificial-intelligence-newshttps://sfcompute.org/

This week, in Episode 25 of This Day in AI we discuss the motivation behind the Frontier Model Forum (OpenAI, Google, Anthropic, Microsoft) and if Open Source remains the best approach for AI safety and security. We discuss Llama 2 being #2 on the AlpacaEval Leaderboard and its significance to the development of AI. We also discuss the paper on Universal and Transferable Adversarial Attacks on Aligned Language Models and how Mike can't prompt inject his AI girlfriend. We discuss how Mike cloned his friend with an AI Bot and the future implications. And also.. Stackoverflow AI, Stable Diffusion XL 1.0 and technology advances that might be being made by AGI!?If you like this podcast please consider subscribing and leaving us a review.CHAPTERS:00:00 - Chris think Anthropic is a Safety Cult (cold open)00:29 - Llama2 is Number 2, Ahead of ChatGPT on AlpacaEval Leaderboard06:12 - Does Llama2 threaten OpenAI and Anthropic?09:12 - Characters to Thwart Prompt Injection Attacks17:51 - Debate on Regulation Vs Open Source for Safety, Senate Committe on AI and thoughts on AI Risk36:28 - Is the Frontier Model Forum a Cartel?42:05 - Mike's AI Girlfriend, Cloning a Friend and Talking to the Dead with AI49:43 - Fine Tuning AI to Increase Retention54:45 - Stack overflow Fights Back! Stack Overflow AI is here!59:18 - Stable Diffusion XL 1.01:03:46 - Could Room Temperature Ambient Superconductors Could be a Sign of AGI!?SOURCES:https://tatsu-lab.github.io/alpaca_eval/https://twitter.com/profoundlyyyy/status/1684333960753565696?s=20https://medium.com/@daniel_eth/ai-x-risk-at-senate-hearing-7104f371ca0bhttps://venturebeat.com/ai/hugging-face-github-and-more-unite-to-defend-open-source-in-eu-ai-legislation/https://huggingface.co/blog/assets/eu_ai_act_oss/supporting_OS_in_the_AIAct.pdfhttps://openai.com/blog/frontier-model-forumhttps://www.wired.com/story/metas-open-source-llama-upsets-the-ai-horse-race/https://twitter.com/OpenAI/status/1684145154628653056https://llm-attacks.org/zou2023universal.pdfhttps://twitter.com/zicokolter/status/1684500097386811393/photo/1https://futurism.com/experts-ai-girlfriend-apps-menhttps://twitter.com/emollick/status/1684623965203910656https://arxiv.org/pdf/2303.06135.pdfhttps://twitter.com/StackOverflow/status/1684530704850243584https://twitter.com/natfriedman/status/1684303687894839296?s=46&t=uXHUN4Glah4CaV-g2czc6Qhttps://huggingface.co/stabilityai/stable-diffusion-xl-base-1.0https://twitter.com/russelljkaplan/status/1684042014495592448

The podcast discusses the release of Llama 2, a commercially available open-source AI model. They explore how open-source models threaten the business models of OpenAI and others. Other topics include Instructions for ChatGPT, fine-tuning smaller models, Google's Generative AI Enterprise Search product, Bing Enterprise Chat, AppleGPT rumors, and the state of AI startups and VC funding.

In this episode of This Day in AI Podcast we pack more keywords into the title than ever before and ask... does anyone actually read Podcast description?In episode EP23 we have LOLs from the first robot press conference, coverage of Anthropic's Claude 2 release, discuss Elon Musk's xAI and the possibility the new startup is going to focus on solving AI reasoning with mathematics. We take another look at ChatGPT Code Interpreter and discuss if we could replace our data analyst with it. Could ChatGPT Code Interpreter really be the foundation of GPT4.5 or 5? Does code help stop hallucinations? We finish off with a discussion about the future of AI Interfaces. If you read the description and enjoy the podcast please consider subscribing and leaving us a review where ever you get your pods.CHAPTERS:------00:00 - Mike wants an LLM in his brain for faster I/O00:22 - AI Fear: World's First Human-Robot Press Conference5:50 - Thoughts on Anthropic's Claude 2 from early use20:33 - Does ChatGPT Code Interpreter help solve hallucinations?  31:47 - Is Code Interpreter an early GPT4.5 or GPT5?32:59 - Elon Musk's New Startup xAI36:50 - Stability AI's Stable Doodle & Use Cases45:18 - Google Bard Updates & Google Lens Integration51:43 - The Future of AI Interfaces SOURCES:------https://clipdrop.co/stable-doodlehttps://stability.ai/blog/clipdrop-launches-stable-doodlehttps://blog.google/products/bard/google-bard-new-features-update-july-2023/https://twitter.com/Reuters/status/1677415942463320064https://twitter.com/goth600/status/1679379211545645056?s=20https://claude.ai/chat/https://twitter.com/aipaniclive/status/1678942758872989696?s=46&t=uXHUN4Glah4CaV-g2czc6Qhttps://twitter.com/abacaj/status/1678896482651389953?s=46&t=uXHUN4Glah4CaV-g2czc6Qhttps://www.nytimes.com/2023/07/11/technology/anthropic-ai-claude-chatbot.htmlhttps://twitter.com/abacaj/status/1678922968028651527?s=46&t=uXHUN4Glah4CaV-g2czc6Qhttps://twitter.com/stijnsmits/status/1678186375617585159?s=46&t=uXHUN4Glah4CaV-g2czc6Qhttps://twitter.com/swyx/status/1677896569545461760?s=46&t=uXHUN4Glah4CaV-g2czc6Qhttps://twitter.com/atroyn/status/1679189257314009088?s=46&t=uXHUN4Glah4CaV-g2czc6Qhttps://twitter.com/elonmusk/status/1679546977565184002https://www.theverge.com/2023/7/9/23788741/sarah-silverman-openai-meta-chatgpt-llama-copyright-infringement-chatbots-artificial-intelligence-ai

In Episode 22 we cover GPT-4 general availability and the implications to millions of developers. We discuss George Hotz revealing how GPT-4 works & Alignment of Super Intelligence. Will OpenAI's Super Intelligence succeed? We also cover the implications of AI agents on the open web, OpenAI disabling web browsing in ChatGPT temporarily to protect content creators and discover Jolly Roger's AI time wasting for service to trick scammers.If you like this podcast please consider subscribing, liking and leaving us a comment. We really appreciate your support of the show.CHAPTERS:----00:00 - Cold open00:27 - GPT-4 API General Availability for Developers06:40 - ChatGPT Code Interrupter 15:30 - ChatGPT as Work Assistant: Code Interrupter + Vision16:42 - ChatGPT web traffic down by 10% & thoughts on single agent Vs AI everywhere21:03 - George Hotz leaking GPT-4 is a 16 way mixture model26:50 - Is AI alignment just a human alignment problem? 27:57 - Is alignment making GPT-4 and ChatGPT worse? Discuss on AI alignment38:18 - Is the best defense against super intelligent AI giving everyone super intelligent AI?40:55 - Is the Open Web Doomed? AI's impact on the Open Web54:42 - OpenAI's 20% compute power to "Superalignment"1:00:29 - OpenAI now sends email threats1:03:12 - Waste Scammers Time with This AI Voice ToolSOURCES:----https://openai.com/blog/gpt-4-api-general-availabilityhttps://twitter.com/gdb/status/1677023789807292420?s=46&t=uXHUN4Glah4CaV-g2czc6Qhttps://www.reddit.com/r/ChatGPT/comments/14s72g4/chatgpt_drops_10_traffic_as_excitement_around_it/https://www.youtube.com/watch?v=1v-qvVIje4Yhttps://thealgorithmicbridge.substack.com/p/gpt-4s-secret-has-been-revealedhttps://www.axios.com/2023/07/05/public-web-twitter-open-access-reddithttps://twitter.com/openai/status/1676072388436594688?s=46&t=uXHUN4Glah4CaV-g2czc6Qhttps://openai.com/blog/introducing-superalignmenthttps://www.reddit.com/r/MachineLearning/comments/14ripua/d_introducing_superalignment_openai/https://www.reddit.com/r/ChatGPT/comments/14ruui2/i_use_chatgpt_for_hours_everyday_and_can_say_100/https://www.reddit.com/r/ChatGPT/comments/14qwa6m/openai_now_sends_email_threats/https://www.wsj.com/articles/people-hire-phone-bots-to-torture-telemarketers-2dbb8457https://jollyrogertelephone.com/

In Episode 21 we discussed Inflection AI’s $1.3B fundraise from Greylock, Microsoft, Bill Gates, Eric Schmidt and the increasing competition to OpenAI’s dominance. We discuss the makeup of future AI Agents and look at the emerging race to build the world’s most powerful AI.We also cover GPT-4 Vision LOLs and ask if Sydney is back with Bing’s GPT-4 Vision Implementation and discuss the latest release from MidJourney, Stable Diffusion and why Nikon is begging people to take REAL photos!SOURCES---https://www.forbes.com/sites/alexkonrad/2023/06/29/inflection-ai-raises-1-billion-for-chatbot-pi/?sh=41e9a4321d7ehttps://pi.ai/talkhttps://twitter.com/BasedBeffJezos/status/1674522246461767680https://www.datacamp.com/blog/claude-vs-chatgpt-data-science-comparisonhttps://twitter.com/frantzfries/status/1673410118246105088?s=46&t=uXHUN4Glah4CaV-g2czc6Qhttps://twitter.com/sgourley/status/1673401578282450944?s=46&t=uXHUN4Glah4CaV-g2czc6Qhttps://www.wsj.com/articles/databricks-strikes-1-3-billion-deal-for-generative-ai-startup-mosaicml-fdcefc06?mod=djemalertNEWShttps://stability.ai/stablediffusionhttps://twitter.com/stalinkay/status/1673209118654615552?ref_src=twsrc%5Etfw%7Ctwcamp%5Etweetembed%7Ctwterm%5E1673209118654615552%7Ctwgr%5Ee2d4e3085cdfd6374207da9c4880c28f23878d97%7Ctwcon%5Es1_&ref_url=https%3A%2F%2Fbgr.com%2Ftech%2Fmidjourney-ai-photo-editor-adds-mind-boggling-zoom-out-feature-you-need-to-see-to-believe%2Fhttps://twitter.com/aisafetymemes/status/1673064171028480003?s=46&t=uXHUN4Glah4CaV-g2czc6Qhttps://twitter.com/emollick/status/1673533969923555328?s=46&t=uXHUN4Glah4CaV-g2czc6Qhttps://www.theverge.com/2023/6/26/23773914/ai-large-language-models-data-scraping-generation-remaking-webhttps://www.vice.com/en/article/v7b774/ai-generated-books-of-nonsense-are-all-over-amazons-bestseller-listshttps://www.washingtonpost.com/technology/2023/06/28/openai-chatgpt-lawsuit-class-action/CHAPTERS---00:00 - LOLs from GPT-4 Vision on Bing Chat00:39 - Inflection AI's $1.3B Fundraise (and Pi) & The Race to Build the World's Most Powerful AI11:07 - Divergent Future Visions of AI Agents/Assistants & ChatGPT Workspaces Leak17:06 - Future AI "stacks", Custom LLM Training as a Service, Comparing LLMs28:10 - MosaicML Acquired by DataBricks for $1.3B & Private LLM training, Agents Creating Specialized LLMs34:00 - Are LLM APIs becoming commoditized? Is the ChatGPT brand and distribution the most valuable thing?40:56 - How Much Context? Building the AI Brain & Aligning it to Your Values46:27 - Stable Diffusion XL, MidJourney Zoom Out & Nikon's Desperation54:36 - Bing's GPT-4 Vision Implementation LOLs: Is Sydney Back?58:27 - Is AI Going to Destroy The Internet and Become Stupid? Spam sites, Crawlers, AI-Generated Book Spam on Amazon1:02:00 - OpenAI ChatGPT Lawsuit over Data Use: Class Action? 

In Episode 20 we discuss Chris winning $10k with GambleGPT while Mike was sick on his vacation. We take a look at FinGPT and discuss the future of AI agents and tools. We do a reality check on AI hype and cover the rumors of an OpenAI app store. Finally we discuss GPT-4 vision and how Bing is leaking some of the promised GPT-4 vision features. We also cover regulation updates in the EU and fake-AI dating app Blush.Chapters:----00:00 - Bing's GPT-4 Vision Captcha Solver00:21 - Chris Made $10k from GambleGPT & AI Agent Updates07:58 - FinGPT: Should Analysts Have a Job?14:42 - The AI Hype Cycle: Where are we at? 18:12 - Are Creative Jobs in Trouble? "Secret Invasion" AI Opening, Grammys Banning AI35:19 - Google Bans Employees Using Bard37:14 - OpenAI App Store? Are Plugins a Failure?43:42 - Where is GPT-4 Vision? In Bing! Solving Captchas49:11 - Meta VoiceBox, EU Regulation & Should AI Content Be Labeled?