An important reminder to those investing in tech+bio that regardless of how much tech there is, you are still buying bio risk. And if you don’t understand the bio risk, it’ll bite you. …matter how large you foundation models is https://
science.org/content/blog-p
ost/ai-and-hard-stuff
…
RESEARCH
-
Tech-Bio Investments: Understanding Biological Risk Beyond AI
By
–
-
State of GPT: Karpathy Explains LLM Training Paradigms at Microsoft Build
By
–
State of GPT – Andrej Karpathy at Microsoft Build Such awesome talk on state of GPT and other LLMs. Karpathy in his usual best teaching style talks about the training paradigms of large language models(pretraining, supervised finetuning, reward modelling, RLHF), prompting
-
Combining LLMs with Text-to-Image Diffusion Models
By
–
New BAIR Blog post on combining LLMs like GPT-4 with text-to-image diffusion models. https://
bair.berkeley.edu/blog/2023/05/2
3/lmd/
… -
Gorilla AI Model Released into Wild by Berkeley
By
–
The Gorilla has been released into the wild… 🦍 https://t.co/4rs6X74821 https://t.co/xg9irsMlyM
— Berkeley AI Research (@berkeley_ai) 25 mai 2023The Gorilla has been released into the wild… https://
gorilla.cs.berkeley.edu -
Cerebras Model Lab: Compare LLMs, Estimate Training Costs
By
–
Introducing Model Lab: a new tool we built to make sense of training LLMs! The app lets you easily: – compare different models (LLaMA, Pythia etc)
– estimate training FLOPS, memory, inference cost
– visualize loss curves
Give it a try here https://
cerebras.net/model-lab/ -
Google Research innovations behind Google I/O 2023 announcements
By
–
Many of the exciting projects and products announced on-stage at #GoogleIO 2023 leveraged innovations by Google Research. Learn about some of the technologies and underlying research behind these exciting announcements at https://
goo.gle/3BXOHAK -
Generative AI Carbon Footprint: Environmental Impact of Chatbots
By
–
#GenerativeAI is tech's hottest trend, but is it cranking up Earth's thermostat? Check out #AI researcher @kate_saenko_
's carbon rundown and the possibility of "greener" #chatbots. #ChatGPT #BERT #carbonfootprint #environment http://
ow.ly/FCcw50OwWEF -
Are Language Models Training Us Instead of We Training Them?
By
–
Are we training the language models, or are the language models training us?
-
Language Model Performance: Emotional Intelligence vs Reasoning Strategies
By
–
This battle is quite revealing. Karina's prompting strategy was getting the language model to say it understands (emotional intelligence). My strategy was to get the model to reason more (IQ). Our strategies each worked on the respective language models that we are used to using.
-
Confirming GPT-4 Limitation with Simple Prompting Approach
By
–
First, I confirm that it doesn't work on GPT-4 with simple prompting, to prove the prompt matters, and it's not just because GPT-4 is smarter than Claude.