Eric Lin, MD

Clinical Assistant Professor (Affiliated) at Stanford Psychiatry ;
Formerly: Harvard Psych, Yale Psych, UCLA Med
Interests: AI/LLMs/ML for psychiatry

[Joel Jang @jang_yoel]

Excited to introduce 𝐋𝐀𝐏𝐀: the first unsupervised pretraining method for Vision-Language-Action models.

Outperforms SOTA models trained with ground-truth actions
30x more efficient than conventional VLA pretraining

📝: https://t.co/duKBjyJLDH
🧵 1/9

[Isaac Kohane @zakkohane]

When medicine is failing, the question for AI applications becomes: Compared To What?

[Isaac Kohane @zakkohane]

How well can AI (GPT4+RAG) identify patients for clinical trials relative to experts using human-defined queries? @NEJM_AI https://t.co/y2KzWrUxWb (spoiler: Very well)

CODA: Which one of the authors is featured in https://t.co/idupAIvwaS ?

[Thomas Wolf @Thom_Wolf]

The new interpretability paper from Anthropic is totally based. Feels like analyzing an alien life form.

If you only read one 90-min-read paper today, it has to be this one

https://t.co/9yecogVqJf

[Aaditya Ura @aadityaura]

Introducing OpenBioLLM-Llama3-70B & 8B: The most capable openly available Medical-domain LLMs to date! 🩺💊🧬

Outperforms industry giants like GPT-4, Gemini, Meditron-70B, Med-PaLM-1, and Med-PaLM-2 in the biomedical domain. 🏥📈 🌟

OpenBioLLM-70B delivers SOTA performance, setting a new state-of-the-art for models of its size.
OpenBioLLM-8B model and even surpasses GPT-3.5, Gemini, and Meditron-70B! 🚀

Today's release is just the beginning! In the coming months, we'll be introducing:

- Expanded medical domain coverage 🧠
- Longer context windows 📜🔍
- Better benchmarks 📈🏆
- Multimodal capabilities🖥️🩺📊🔬

Medical-LLM Leaderboard: https://t.co/rSM98TLidd

#gpt #gpt4 #gemini #medical #llm #chatgpt #opensource #llama3 #meta

[Eiko Fried @EikoFried]

Neuroimaging research in psychology has led to so many false positive, non replicable findings due to questionable research practices & publication bias that Science Advances now publishes null findings to clean up the literature (this is from 2022).

https://t.co/c7EuzQvqrJ

[Roy Perlis @royperlis]

We showed the problem with math in 2009 (https://t.co/Ws9QgFS1pB): the conditions under which personalized standard antidepressant selection is particularly useful are often unrealistic, in terms of differences in effectiveness (bc much of benefit placebo-like)

[Eric Lin, MD @emailELin]

Important to watch out for AI regulations that effectively bolster the position of big tech incumbents by shutting out open source researchers and developers.

More of the right overwatch is good. Less of the FTX crypto - esque regulation lobbying

https://t.co/ztZfPINFpv

[Roy Perlis @royperlis]

Re LLM in medicine: “We encourage efforts to understand how and when models report wrong results, with examples of such responses.” https://t.co/ayx9cqjyB0

[Jeremy Howard @jeremyphoward]

I just uploaded a 90 minute tutorial, which is designed to be the one place I point coders at when they ask "hey, tell me everything I need to know about LLMs!"

It starts at the basics: the 3-step pre-training / fine-tuning / classifier ULMFiT approach used in all modern LLMs.

[Misha Teplitskiy | Science of Science @MishaTeplitskiy]

The longer people are in academia, the more they realize that when reading papers it's best to ignore Intro, Discussion etc. and just look at Methods and Results
https://t.co/EzQL7WI71Q

[Philipp Schmid @_philschmid]

Code Llama with @huggingface🤗 Yesterday, @MetaAI released Code Llama, a family of open-access code LLMs!
Today, we release the integration in the Hugging Face ecosystem🔥

Models:
👉 https://t.co/CSnWDscpi6

blog post:
👉 https://t.co/5VQrONcBmt

Blog post covers how to use it!

[Yura Gorishniy @YuraFiveTwo]

In our current project (coming soon, stay tuned!), we (with @irubachev) have evaluated our tabular DL models on the datasets from this paper by @LeoGrint @GaelVaroquaux et al.

Let's discuss the results!

🧵 1/7

[Thomas Wolf @Thom_Wolf]

What was going on with the Open LLM Leaderboard?

Its numbers didn't match the ones reported in the LLaMA paper!

We've decided to dive in this rabbit hole with friends from the LLaMA & Falcon teams and got back with a blog post of learnings & surprises: https://t.co/bREo0oiQ01

[Jared Palmer @jaredpalmer]

Added more OSS language models to the Vercel AI Playground via @huggingface:

◆ OpenAssistant/oasst-sft-4-pythia-12b-epoch-3.5
◆ EleutherAI/gpt-neox-20b
◆ bigcode/santacoder
◆ bigscience/bloom

→ https://t.co/RbTRy65q2F

[AI Breakfast @AiBreakfast]

OpenAI's speech-to-text API "Whisper" just got supercharged:

This tool transcribes audio 70x faster than Whisper

A 2-hour podcast can now be transcribed in ~30 seconds using Whisper JAX: The Fastest Whisper API

Try it here using your mic: https://t.co/avz4i3vzwE

[Sebastian Raschka @rasbt]

Want to adopt the latest LLMs for a specific target domain (e.g., finance data) or target task like document classification?

Finetuning becomes more feasible as we see more and more pretrained LLMs become available under open-source licenses.

https://t.co/5XITvDF9B5

Here's an intro before I go into more depth on various parameter-efficient finetuning techniques, reinforcement learning with human feedback, and more in the upcoming weeks.

[Together AI @togethercompute]

First, an instruction-tuned large language model, fine-tuned for chat from EleutherAI’s GPT-NeoX-20B with over 43 million instructions on 100% carbon negative compute available under Apache-2.0 license on @huggingface.

https://t.co/sluFlx9Jsk

[Niels Rogge @NielsRogge]

Informer, from "Beyond Efficient Transformer for Long Sequence Time-Series Forecasting" (AAAI'21 Best Paper) is now available @huggingface Transformers! 🙌

We created a tutorial for multi-variate forecasting, check it out: https://t.co/ndMGFQeTbc

Colab with @ESimhayev

(1/2)

[Eric Lin, MD @emailELin]

Great discussion on GPT/LLMs and whether they can truly understand the world or are merely statistical parrots

https://t.co/mEzGoLoh6R