I have been spending a lot more time learning about AI lately and realised that I still get lost sometimes, the field is just moving so fast…
Thought I might use this account to share things that I am learning. From interesting AI news I come across or just stuff I’m building
OpenAI’s Chief Scientist described modern AI as being “grown more than designed.”
Sounds very sci-fi, but the idea is pretty simple:
we don’t manually program all of an AI’s behaviour. We design the training process, scale it massively, and capabilities emerge.
OpenAI also says some of the ways they’ve been monitoring how models reason are becoming less reliable as the models get better and I do think that this is getting a little uncomfortable.
Kinda wild that making AI smarter and understanding AI are not necessarily progressing at the same speed.
https://t.co/C6YeIebpWG
@alexanderrX_ Yup agreed, I have been speaking to peers and friends in the same field as I am / fields related. The vibe that I’m getting from them is that they just feel that AI is just a chatbot. They don’t use MCPs or any plugins in Gemini or Claude or ChatGPT too.
So what is an AI agent, and how does it differ from a chatbot?
For me, the useful distinction is that chatbots mostly answer you, while agents can actually do stuff.
Agents can figure out steps, use tools and keep working towards a goal. There’s obviously more nuance to it, but this mental model has been useful whenever my friends ask me, “what is an AI agent?”
I have been spending a lot more time learning about AI lately and realised that I still get lost sometimes, the field is just moving so fast…
Thought I might use this account to share things that I am learning. From interesting AI news I come across or just stuff I’m building