ThursdAI - The top AI news from the past week

Claude Sonnet 3.5, the new LLM king, shakes up the AI world with groundbreaking advancements beating GPT-4o. Topics also cover DeepSeek V2, Meta Chameleon MMIO models, NousResearch LLama3, Big CO LLMs, & Ilya Sutskever's SSI.inc for safe super intelligence.

This podcast covers Apple's new generative AI wave, the growth of $APPL stock post-announcement, new eval frameworks like MixEvals, a video model from Luma, SD3, AI Engineer expo in SF, open source LLMs, Microsoft's Samba architecture, Sakana AI's LLM squared, Abacus and Yan Lecun's Live Bench.AI benchmark, and the debate on Apple's payment to OpenAI.

Qwen2 announces the new 72B model surpassing Llama-3, Jina and Nomic discuss multimodal embeddings, Sora introduces SORA text-to-video models, Suno and Udio present new audio features, and various AI advancements are highlighted including total recall tool by Microsoft

The podcast discusses advancements in AI models and embeddings, including Mistral and LLMs. It covers speed breakthroughs in NLP by SambaNova and Groq, as well as innovative architectures like Cartesia's state-space models. The episode explores benchmarks, model rankings, and efficient real-time intelligence with Cartija. Updates on OpenAI, partnerships, and computing costs are also highlighted.

This podcast covers AI news from OpenAI's troubles to Microsoft's Build announcements, NPUs in new AI native devices, and new models like Phi-3 small/large. Topics include HuggingFace's ZeroGPU commitment, Mistral 7B model, LMSys hard prompts, Cohere's Aya 23 release, and Microsoft's new features and functionalities like Copilot. The episode also delves into advancements in AI models, interpretability in language models, the impact of AI on digital property and privacy, on-device AI progress, and collaborations with Khan Academy.

This podcast covers the exciting launch of OpenAI's GPT-4o model, the host's hackathon experience with Meta LLama-3, and the highlights from Google IO. Dive into discussions on AI advancements, emotive customization capabilities, and the integration of AI models with wearable technology.

The podcast features an interview with the OpenUI creator, discussing the project's success and community feedback. Other topics include breaking SOTA on Swe-Bench with Open Devin, new GPT chatbots from OpenAI, AlphaFold 3 by Google, and the latest AI innovations such as XLSDM and DeepSeek V2.

Reka from Reka AI Lab discusses Evals, Vibe checks, and a new paper on overfitting. Idan Gazit from GitHub Next talks about Copilot Workspace. Topics include LLama-3 rankings, GPT2-chat, and data contamination issues in AI models.

This podcast covers the release of Microsoft's Phi-3 3.8B model, Snowflake's hybrid MOE transformer, and llama models ranking in the top 6. Discussions include fine-tuning techniques, big AI models like Devin worth $2BN, and advancements in AI agents and startups. The episode also explores hardware product analysis, UI design impacts, and Apple's on-device AI inference plans.

LLaMa 3 day celebration with new 8B and 70B models announced, Meta's collaboration with Google, model releases with improved benchmarks, AI integration in Premiere Pro, advancements in ML ops partnerships and fine-tuning libraries, discussions on Weights and Biases involvement, parameter usage in AI models, Reka's multi-modal capabilities, and weekly recap with gratitude to contributors and listeners.