Anubhav Shrimal

Anthropic has just open-sourced their interpretability tool that allows researchers to trace the internal "thoughts" of large language models (LLMs) in detail.

In my previous video, I provided an in-depth look at their research blog on LLM Interpretability.

Now, Anthropic has made this powerful interpretability technology openly available. Researchers can use it to generate "attribution graphs" that visualize and explain the key reasoning steps involved in language generation.

This is a crucial development for transparency and accountability as we work towards advanced AI systems.

read more about the release here: www.anthropic.com/research/open-source-circuit-tra…

1 year ago | [YT] | 1

Anubhav Shrimal

Andrej Karpathy recently tweeted about Moshi Chat praising its GPT4o like audio conversations.

It is really fast in talking which at times feels like it is reading your mind to reply that fast and also interrupts you at times.

I had talked about Moshi Chat and also tried it first hand in one of my previous videos. You can check it out here: https://youtu.be/nDla7jJBh80

Skip the Carousel

1 2

1 year ago (edited) | [YT] | 2

Anubhav Shrimal

OpenAI released a new model O1, codenamed strawberry 🍓.

openai.com/index/introducing-openai-o1-preview/

The model has been fine-tuned to have reasoning for solving hard problems using chain of thought, where it thinks for longer duration to come up with a details answer.

Let me know if you’d like me to make a detailed video on the model, it’s performance, and how it might have been trained.

Or comment below to let me know the topics you’d like to hear more about. :)

1 year ago | [YT] | 1

Anubhav Shrimal

Would you like to see a tutorial on creating images using AI with your face included in it using Flux model?

1 year ago | [YT] | 0