I’m a huge fan of LLM efforts that emphasise multilingualism because they make AI far more inclusive and accessible. Congratulations @cohere
@nandodf
-
AI Leaders Network at Exclusive Dinner Event
By
–
Thanks @osanseviero and @huggingface for inviting me to a wonderful AI dinner, where I had the pleasure of catching up with old friends, @sarahookr @neilzegh @laurentsifre @ylecun
, meet new amazing people, and do one of the things I absolutely love: brainstorm about AI with -
Blind Robot Learns World Representation Through Touch Prediction
By
–
A blind robot that predicts what it expects to sense next with its touch sensors, learns an internal representation of what the world looks like. In this case, predicting hand sensors enables to robot to understand the orientation of green cake slice.
— Nando de Freitas (@NandoDF) 29 mars 2024
I believe this accentuates… pic.twitter.com/ggOC1xLnGjA blind robot that predicts what it expects to sense next with its touch sensors, learns an internal representation of what the world looks like. In this case, predicting hand sensors enables to robot to understand the orientation of green cake slice. I believe this accentuates
-
Next Word Prediction: How Language Models Learn Knowledge
By
–
Predicting the next word "only" is sufficient for language models to learn a large body of knowledge that enables then to code, answer questions, understand many topics, chat, and so on.
— Nando de Freitas (@NandoDF) 29 mars 2024
This is clear to many researchers now, and there are nice tutorials on why this works by… pic.twitter.com/L95G6hWRNVPredicting the next word "only" is sufficient for language models to learn a large body of knowledge that enables then to code, answer questions, understand many topics, chat, and so on. This is clear to many researchers now, and there are nice tutorials on why this works by
-
Pioneering AI Dreams at Google DeepMind
By
–
Thank you for pioneering the AI dream, when few others believed in it. You created a unique opportunity for many scientists and engineers, including myself. You brought many of us together at @GoogleDeepMind in an inclusive atmosphere that allowed for many interesting ideas to be
-
Publishing Criteria vs Practical AI Effectiveness: MM1 Insights
By
–
There appears to be a mismatch between publishing criteria in AI conferences and "what actually works". It is easy to publish new mathematical constructs (e.g. new models, new layers, new modules, new losses), but as Apple's MM1 paper concludes: 1. Encoder Lesson: Image
-
Azure GPU Investment: How Academia Shaped AI Infrastructure
By
–
Only 10 years ago, @Azure had no GPUs. Lessons: 1. If you’re a company, invest in academia because one of the random professors might inform you of what is needed for you to become one of the main AI companies in the world. @Azure now powers @OpenAI and its #ChatGPT 2. AI is
-
Understanding Intelligence Symbols and Consciousness Beyond Internal Theatre
By
–
One of the most revealing ideas in the quest to understand intelligence, symbols, concepts and consciousness. There’s no internal theatre #atheaterism I first came across it when studying concepts 7 years ago. It took long to grasp it, but once you see it, the activations are on!
-
Gratitude for AI Education Work and Visit
By
–
I’m also grateful for your visit, and for your amazing work on AI education. Thank you
-
Hinton argues intelligent machines possess understanding and subjective experience
By
–
An important take on intelligent machines today, arguing that they already have understanding and subjective experience , by Geoffrey Hinton | https://
youtu.be/iHCeAotHZa4?si
=xOljp9ngspiela_v
… via @YouTube