John Licato
johnlicato.bsky.social
John Licato
@johnlicato.bsky.social
AI Professor and Startup Owner | Associate Professor, USF Computer Science and Engineering | Director, Advancing Machine and Human Reasoning (AMHR) Lab | CEO and Founder, Actualization.AI
Reposted by John Licato
Agent Control: Steer LLM Beliefs with 'Belief Box' Prompts

Getting LLM agents to behave predictably and adhere to a consistent persona.

- from "The Light Bulb" Newsletter (Dec 16, 2025 Issue)
Track emerging directions in AI

@johnlicato.bsky.social

(see links in comment)
December 16, 2025 at 11:04 PM
“AI can only reproduce what it’s trained on.” People repeat this with total confidence. They’re wrong.

I’ve been working in AI for almost 20 years, and this is one of the most common claims I still hear. Modern AI is capable of creativity, by any reasonable definition of the term.
November 29, 2025 at 5:30 AM
Despite #DeepSeek getting heavily criticized for its apparent built-in censorship, looks like #Grok does so as well.

Seems like there are instructions in its system prompt to deny that Musk, Trump, or X are spreaders of misinformation.
February 24, 2025 at 12:38 AM
Finally created an account! Follow me here or on LinkedIn for my updates.
February 23, 2025 at 6:26 PM