Joe Cox
Region: US
Thursday 21 May 2026 01:34:07 GMT
2434
57
4
2
Music
Download
Comments
Jay :
Aren’t we all statistical text predictors, pretrained since born, RL’d in college?
2026-05-21 01:52:18
0
Andrei :
Yes it lands. On the moon! 😎
2026-05-21 06:06:56
0
Sam 🇺🇲🌎 :
I appreciate your take on this and your other videos. I've followed you for a while now because I like to see other perspectives. I'll try and do a video response if/when I have time because comments are constraining and this is an interesting topic. I think for me agents and LLM/AI gets muddy where the line is. But for the sake of discussion, let's draw a hard line that they are fundamentally different. Agents are what can be useful at some point. Just like any software tailored to a task. Machine learning has been around for a long, long, long time. Human assisted and otherwise. Take the IDL programming language. By today's standards a shell script on Linux would be a rudimentary agent, it can make decisions, open close, documents edit, etc. Where I've hit hard walls is stepping out of the software world into life. Trying to use an agent to do something as simple as programming an esp32 to use the low level channel state of the WiFi to detect motion is an exercise in frustration. You spend more time fixing the end product than it would take to learn the code and do it, it use a code scaffold/outline and modify it. Agents are something I think might have real world niche applications that do have some use, but so far I see it hitting hard walls in almost all applications above a certain complexity, especially if there's nuance involved. Long comment, and I hope I got the core essence of what I'm trying to say in a way that is very much said with respect for your views.
2026-05-21 04:01:04
1
To see more videos from user @joecoxtech, please go to the Tikwm
homepage.
