Congratulations to all my Google friends for the cool announcements at I/O! I'm personally looking forward to Gemini with 2 million token input context window and better support for on-device AI — should open up new opportunities for application builders!
@andrewyng
-
GPT-4o Release: 2x Faster, 50% Cheaper Token Pricing
By
–
Congrats to OpenAI for the release of GPT-4o! 2x faster and 50% cheaper tokens will be great for everyone using agentic AI workflows. When an agentic job that used to take 10min now takes 5min just by switching APIs, that's great progress!
-
Multimodal Search and RAG with Contrastive Learning
By
–
New short course: Building Multimodal Search and RAG", by @weaviate_io's @sebawita.
— Andrew Ng (@AndrewYNg) 13 mai 2024
Contrastive learning is used to train models to map vectors into an embedding space by pulling similar concepts closer together and pushing dissimilar concepts away from each other. This… pic.twitter.com/pUl3SXdOQGNew short course: Building Multimodal Search and RAG", by @weaviate_io
's @sebawita
. Contrastive learning is used to train models to map vectors into an embedding space by pulling similar concepts closer together and pushing dissimilar concepts away from each other. This -
Open Source Fights Back Against AI Regulation at Capitol
By
–
Last week, I spoke about AI and regulations at an event at the U.S. Capitol attended by legislative and business leaders. I’m encouraged by the progress the open source community has made fending off regulations that would have stifled innovation. But opponents of open source are
-
Building Agentic RAG with LlamaIndex: First Short Course
By
–
I’m excited to kick off the first of our short courses focused on agents, starting with Building Agentic RAG with LlamaIndex, taught by @jerryjliu0, CEO of @llama_index.
— Andrew Ng (@AndrewYNg) 8 mai 2024
This covers an important shift in RAG (retrieval augmented generation), in which rather than having the… pic.twitter.com/KwRvDEhENSI’m excited to kick off the first of our short courses focused on agents, starting with Building Agentic RAG with LlamaIndex, taught by @jerryjliu0
, CEO of @llama_index
. This covers an important shift in RAG (retrieval augmented generation), in which rather than having the -
Quantization in Depth: Compressing ML Models Efficiently
By
–
Have you used quantization with an open source machine learning library, and wondered how quantization works? How can you preserve model accuracy as you compress from 32 bits to 16, 8, or even 2 bits? In our new short course, Quantization in Depth, taught by @huggingface's… pic.twitter.com/7sj1Ivv5a4
— Andrew Ng (@AndrewYNg) 6 mai 2024Have you used quantization with an open source machine learning library, and wondered how quantization works? How can you preserve model accuracy as you compress from 32 bits to 16, 8, or even 2 bits? In our new short course, Quantization in Depth, taught by @huggingface
's -
Scaling Deep Learning: The Controversial Path to AI Performance
By
–
Back then, the idea that scaling deep learning would lead to significant performance gains was controversial. Several senior academic colleagues were advising me not to waste time trying to scale deep learning, but to just focus on inventing new algorithms, which was where they
-
Google Brain’s Iconic Cat Image: A Defining Moment
By
–
Yup. By the way @quocleix
, I still remember the time we were both in the office, and you waved at me to come over to your desk to look at your new result. And right there on your monitor was the now-famous Google Cat image. That was a defining moment for Google Brain! -
DistBelief Pioneer Reflects on Distributed Machine Learning Legacy
By
–
You were a real pioneer @unJADded ! Looking back, honestly I feel a bit bad at how hard DistBelief was to use, and the complexity of the C++ interface we built…. Nonetheless, I'm glad it turned out to be a useful product of its time!
-
DistBelief Lessons: Scaling Deep Learning Architecture Evolution
By
–
I'm with @JeffDean on this. DistBelief taught us early important lessons about scaling up deep learning, and it was general enough for many algorithms including supervised backprop. Obviously, we got a lot of software and hardware architecture details "wrong" back in 2012 —