Foundation models have enabled amazing human-in-the-loop systems (ChatGPT, Copilot, ++). How can we bring them to bear on important batch computing tasks (like information extraction) - where we need efficiency and reliability at scale?
Early thoughts at: hazyresearch.stanford.edu/blog/2023-04-1…
Brandon Yang
257 posts
San Francisco, CA
Joined August 2014
- I work at cartesia but unfortunately am bad at math contests. If you're bad at math, there's a home for you here too!14% of @cartesia is named Brandon, collectively winning the USAMO 4 times If you're a Brandon, come find a home here
- The independent human evals are coming out - Sonic is the highest quality TTS model with conversational latency.2.5 months ago @elevenlabs put up this comparison with our 10 day old Sonic model: elevenlabs.io/blog/elevenlab… The team took it as a challenge, here's our new scorecard. Higher quality, cheaper & the fastest voice model period. labelbox.com/guides/evaluat… Next 3 months will be fun.
- Thrilled to be sharing some of our early work at Cartesia - Sonic, a blazing fast generative voice model. New architectures will be key to the next generation AI - real time, interactive, and on-device. Grateful to be building with an amazing team!Today, we’re excited to release the first step in our mission to build real time multimodal intelligence for every device: Sonic, a blazing fast (🚀 135ms model latency), lifelike generative voice model and API. Read cartesia.ai/blog/sonic and try Sonic play.cartesia.ai
- New model is out! This has been endlessly fun to play with - and opens up a new way to create audio that sounds exactly the way you like. Try it on our Playground!We're releasing a new model called Voice Changer. Transform any input voice clip into an output voice from your voice library, and preserve key characteristics of the input voice like intonation, prosody, and emphasis. Try now at play.cartesia.ai
00:00 - It's been a amazing to see our work on SSMs go from academia to powering real-time voice in production across thousands of customers. And excited to share a sneak peak into our research on multi-stream models for multimodal data. Grateful for our early team and supporters :)We've raised $27M from Index Ventures, Lightspeed, Factory, Conviction, SVA, General Catalyst, A* and our wonderful angels. Cartesia's audio models power the next generation of voice agents, digital media, and assistants across startups and large enterprises. Our mission is to
- I joined Snorkel because I was inspired by the team + vision for a unique data-centric approach to transform every step of the AI lifecycle. Two years later and somehow I'm more excited than I was, and there's even more left to do. Come join us!🦄We are delighted to announce our $85 million Series C at a $1 billion valuation to accelerate #DataCentricAI, with funding by @BlackRock, Addition Capital, @lightspeedvp, @GreylockVC, @googleventures, and more. Read more in @Fortune ↓ snkl.ai/seriesc
- We launched Snorkel Flow today! We're building a data-first platform that leads the way towards iterative, end-to-end ML development. Grateful to get to work with folks I deeply admire @SnorkelML. More on our product and vision at: snorkel.ai/07-14-2020-sno….
- So excited to share our work using LLMs for large-scale information extraction, with asymptotically lower costs! My personal takeaway - LLMs enable fundamentally new system designs, with lots of fun, new trade-offs to explore!LMs can be expensive for document processing. E.g., inference over the 55M Wiki pages costs >$100K (>$0.002/1k toks)💰 We propose a strategy that reduces inference cost by 110x and can even improve quality vs. running inference over each doc directly! 💻 github.com/HazyResearch/e…
- Thrilled to share our work to bring AI to the edge with SSMs. The early AI applications of today run on the datacenter as APIs you can query - but I think the next generation of AI application will run on your device: always on, proactively helpful, fully private, and secure.Today, we’re unveiling a significant milestone in our journey toward ubiquitous artificial intelligence: AI On-Device. Our team pioneered a radically more efficient architecture for AI with state space models (SSMs). Now, we’ve optimized and deployed them at the edge. We believe
- Super excited about this and new possibilities around genome level design and understanding. I'm also a believer in the principle of getting down to the most fundamental level (raw nucleotides) and scaling up with long context models + compute + data. Also pretty pictures :)A new Science study presents “Evo”—a machine learning model capable of decoding and designing DNA, RNA, and protein sequences, from molecular to genome scale, with unparalleled accuracy. Evo’s ability to predict, generate, and engineer entire genomic sequences could change the
- We’re thrilled to lead $27M in new funding for @cartesia as they build the next generation of real-time AI. Their pioneering SSMs offer multimodal intelligence available on any device, powering new solutions for customer service, healthcare, transportation, robotics, and more.
00:00 - Excited to share results and weights for scaling up Mamba text models by @_albertgu @tri_dao with @cartesia @togethercompute! Strong performance, fast inference - and Apache 2.0 🚀Cartesia Chief Scientist @_albertgu teamed up with Together Chief Scientist @tri_dao to release a new 3B Mamba text model trained on the SlimPajama dataset, in a close collaboration with Cartesia & @togethercompute. Read more on our blog: cartesia.ai/mamba-3b-slimpj
- Thanks for having me @DrBahijjaRaimiA! Applications to healthcare are what first got me excited about AI, and it was great to discuss how the field is evolving, outstanding challenges, and how we're working to address some of them at @SnorkelAI!Tune into tomorrow's episode where Dr. Bahijja, our host, and machine learning engineer Brandon Yang (@bclyang) from Snorkel AI (@SnorkelAI) dive deep into the application of AI in healthcare! @anchor @spotifypodcasts #mondaysciencepodcast #DeepLearning #ArtificialIntelligence



















