- AI Minds Newsletter
- Posts
- Altman Questioned by Senator John Fetterman, Karpathy on New LLM Paradigm, and Google Chief Scientist talks AI Coders
Altman Questioned by Senator John Fetterman, Karpathy on New LLM Paradigm, and Google Chief Scientist talks AI Coders
This week, we find that Sam Altman was questioned by Sen. John Fetterman (D-PA) about the Singularity at the AI Hearing. Also, Andrej Karpathy discusses the new LLM-fueled paradigm on X. And Google's Chief Scientist explains why we should expect AI systems to code like Junior-Engineers next year. All this and more in this week's edition of AI Minds!
Jose Nicholas Francisco
May 13, 2025
Welcome (back) to AI Minds, a newsletter about the brainy and sometimes zany world of AI, brought to you by the Deepgram editorial team.
In this edition:
π₯ Senator John Fetterman Asks Sam Altman About The Singularity during Hearing
π§ͺ GenAI is on its Way to Predicting Emergent Chemical Phenomena
ποΈ New Diffusion Transformer Improves Video Generation of Humans
β‘ Hear the Mona Lisa Speak: Lemon Slice Live and Deepgram (Featured Last Week)
π¦ Social Media Buzz: Karpathy talks new LLM Paradigm
π€ Google Chief Scientist says AI systems will code at a Junior-Engineer level in a year
π² Three new, trending AI Apps for you!
ποΈ The AI Minds Podcast!
Thanks for letting us crash your inbox; letβs party. π
Looking for a cutting-edge AI medical transcription model? Click here. π₯³
π₯ John Fetterman Asks OpenAI CEO Sam Altman Straight-Up About The Singularity During Hearing On AI
Video Description: βAt Thursday's Senate Commerce Committee hearing, Sen. John Fetterman (D-PA) questioned OpenAI CEO Sam Altman about the Singularity.β
π GenAI on its Way to Predicting Emergent Chemical Phenomena and New Diffusion Transformer Improves Video Generation of Humans
Generative artificial intelligence for computational chemistry: a roadmap to predicting emergent phenomena - Generative AI methods have made significant progress in sampling molecular structures across chemical species, developing force fields, and speeding up simulations. This paper offers a structured overview, beginning with the fundamental theoretical concepts in both Generative AI and computational chemistry.
HumanDiT: Pose-Guided Diffusion Transformer for Long-form Human Motion Video Generation- Human motion video generation has advanced significantly, while existing methods still struggle with accurately rendering detailed body parts like hands and faces, especially in long sequences and intricate motions. To address such limitations, this paper proposes HumanDiT, a pose-guided Diffusion Transformer (DiT)-based framework trained on a large and wild dataset containing 14,000 hours of high-quality video to produce high-fidelity videos with fine-grained body rendering.
β‘ Hear the Mona Lisa Speak: Lemon Slice Live and Deepgram (Featured Last Week)
Bringing Characters to Life in Real Time: Lemon Slice Live Launches with Deepgram - Lemon Slice is pioneering the next generation of expressive AI with a video foundation model that lets anyone create and interact with animated, talking characters. Their newest release, Lemon Slice Live, takes this vision even further. It allows users to video chat in real time with characters generated from any photo or illustration. Check it out here!
Google Chief Scientist, Jeff Dean:
We will have AI systems operating at the level of junior engineers within a year.
β Haider. (@slow_developer)
10:50 PM β’ May 12, 2025
Andrej Karpathy calls large language models the new computing paradigm:
CPU -> LLM
bytes -> tokens
RAM -> context windowthis is the large language model OS (LMOS)
β βΞ΅sam (@Hesamation)
8:49 PM β’ May 9, 2025
Can LLMs help us discover the right molecules for new medicines?
An MIT-led approach combines an LLM w/graph-based models to create new molecules w/specific properties. Users can request molecules in plain language & get a description of how to make it: bit.ly/3RerIZY
β MIT CSAIL (@MIT_CSAIL)
4:03 PM β’ Apr 10, 2025
π² Three new, trending AI Apps for you!
Competera is a cutting-edge pricing solution designed to empower retailers to optimize their pricing strategies dynamically. With its advanced AI and deep learning algorithms, Competera analyzes billions of price combinations and market factors, offering enterprise clients the ability to boost their gross margin by an average of 6%.
RapidCanvas stands out by drastically reducing the time to value for AI projects by over 90%. This no-code platform is intuitively designed for business users, requiring no coding skills or data science experience. Click the link to learn how it simplifies the AI solution development process.
Aerospike is a pioneering force in the realm of databases, especially designed for businesses seeking high performance, scalability, and cost-effectiveness. It's not just a database; it's a comprehensive real-time data platform that includes the Aerospike Database, Aerospike Graph, and Aerospike SQL. This multi-model NoSQL cloud managed database service is engineered to support different data models effectively, making it versatile for various business applications.
ποΈ The AI Minds Podcast!
Shraey Bhatia, Co-Founder & Chief Scientist at Leaping AI, a company revolutionizing enterprise businesses, call centers, and sales with self-improving Voice AI. Founded in Germany in August 2023, Leaping AI is part of Y Combinator's prestigious W25 batch, driving the future of AI-powered communication.
π Social Media Buzz: Karpathy talks new LLM Paradigm and More!