Would you want people to make a digital ghost in your likeness after your death?
A new Google DeepMind paper says it will be commonplace for us to have digital ghosts.
Woh. My brain is trying to grasp this. And what about when we have life-like robots? 🤔
https://t.co/gDyHVQ7zDr
Here's some interesting research. I had no idea hallucinations were so high in LLMs using RAG (Retrieval Augmented Generation.) This explains why our Agents we are building in the Search Bar using VertexAI + RAG have so many issues with hallucinations!
This new research introduces techniques to reduce the rate of hallucinations in RAG systems by 2-10%. They use another LLM to judge whether the system was given sufficient context, and also they fine tune the AI to be more likely to say, "I don't know" when it's unsure of the answer.
https://t.co/kkrIXAd7jt
LOL, Sora. What the heck?
Prompt: A moving scene where a woman is working on multiple experiments and is exhausted and then eventually gets the one that works.
Introducing Helix, our newest AI that thinks like a human
To bring robots into homes, we need a step change in capabilities
Helix can generalize to any household item 🧵
Grok 3 was released this week, with X announcing they have a cluster of 200,000 AI GPUs. That is a massive amount of compute.
Elon told Lex Fridman that there's only a limited amount of training data on the web...yet X, Teslas and eventually Optimus robots will provide limitless training data from the real world.
Keep an eye on Grok 3. It will be improving daily and I expect it will be quite powerful soon.
It's worth watching this Demis Hassabis clip. He says we are entering a very interesting phase in the next few years on the web in the way we interact with websites.
DeepMind published a new paper with a method called SCoRe that uses reinforcement learning to train language models to correct themselves with iterative revision. This makes their output much more accurate and reliable!
https://t.co/xWkmPZztKD
Anthropic's Dario Amodei says we're only a few years away from a world where AI is better than humans at everything.
“I don’t know exactly when it will come. I don’t know if it will be 2027, I think it’s plausible it could be longer than that. I don’t think it will be a whole bunch longer than that when AI systems are better than humans at almost everything. Better than almost all humans at almost everything and then eventually better than all humans at everything, even robotics.”
Anthropic's Dario Amodei says we're only a few years away from a world where AI is better than humans at everything.
“I don’t know exactly when it will come. I don’t know if it will be 2027, I think it’s plausible it could be longer than that. I don’t think it will be a whole bunch longer than that when AI systems are better than humans at almost everything. Better than almost all humans at almost everything and then eventually better than all humans at everything, even robotics.”
Google's new Titans architecture is so cool. I wrote up my thoughts on the paper and also what it might mean for Google's Search algorithms:
https://t.co/4r8tb9y8lz
Woo. https://t.co/h9r4qOFtfC quoted in @sejournal🦾
Have you used Deep research? I used it 4 or 5 times today already. It's super helpful. But also, doesn't feel fair.
Who knows, perhaps the sites listed in it will get more clicks. We will see.
https://t.co/6MtagTFKdc