We were unable to load Disqus. If you are a moderator please see our troubleshooting guide.
You can check out her new book: AI Engineering Building Application with Foundation Models, the book has more details, and you will further understand what is already mentioned, I am currently going through the book and I have learned a lot and would highly recommend it
Thanks for sharing! Means a lot.
This is such an informative article. Appreciate your efforts. Could you please explain the difference between Agentic AI and AI Agent?
Hey hey! Thanks for the great article!
Just wanted to let you know of a potential small mistake I observed, and maybe you need to edit it here or in your books
"Different models have different tool preferences. For example, GPT-4 seems to select a wider set of tools than ChatGPT. ChatGPT seems to favor image captioning, while GPT-4 seems to favor knowledge retrieval."
I am not sure if you meant to write ChatGPT here, as ChatGPT is not a model itself but rather the app that can have multiple different models behind, and you probably meant another OpenAI model.
Please correct me if I am wrong
Thank you so much for your effort!
A wonderful post on agents @Chip Huyen. Looking forward to reading your book on AI Engineering. We are working on the text2code problem and agentic AI systems look promising for us. Mathematical and image tools look interesting. Latency is of concern for us. Any suggestions on how to decrease the latency of an agentic system?
+1 on sharing your knowledge without restrictive paywall.
I spent few hours reading this article on a flight and this is an extremely well written, easy to understand, and more importantly of great help to anyone working with LLMs and agents. Highly recommend! I'll be getting my copy of 'AI Engineering Building Application with Foundation Models'.
Thanks, @Chip! I would like to clarify if any agent is expected to handle planning by itself (dynamically picking up the tools)? Or do we consider a system with rigid sequence of actions (e.g. we have a router (LLM-based) where each router label sends flow into a specific predefined sequence of actions/tools without ad-hoc planning by the system itself) also to be "an agent"?
Thanks for the great content ! A question I have is about the differences and definitions between : Agent, AI Agent and foundation model agent. Could you elaborate on those?
Loved the section on planning. I recently built the tools regression model and it was super similar to your approach. Also super excited for the book for so long now, when is it releasing in India???🤞
Great article. I'll be interested in reading the follow up that discusses frameworks and libraries.
It didn't discuss testing very much, and testability / evaluation / transparency is probably one one of the most important aspects of agent / workflow development and framework selection. Will be interested to get your take on this.
Thanks for sharing your knowledge without restrictive paywall. Your last article "Building GenAI platform" was really great to revise about RAGs. Explaining the concepts right from the simplest architecture to the complex, really took me on a journey on learning.
First of all, I just skimmed through the content and it seems to cover almost what i have been mean to read. I didn't mean to dive deep into agents, still wanted to understand how agents are being integrated into what we already know about LLMs. This would a good read over the weekend.
A suggestion, if you could include a source if someone wants to understand further than what is already mentioned.