@alex_prompter Agree! Best if you use multiple companies’ LLMs b/c agents from the same company all have same data, biases and approach even if you assign them different roles. I developed Multi-Origin Divergence Adversarial Council (MODAC).
https://t.co/XixYs4NPna
I came up with AI concept MODAC** that is specifically modeled after Grand Rounds and what you both are describing. 👩🏻⚕️😌
Worth noting that AI tends to build AI agents and models for consensus, not divergence, so this is a novel approach (though seems normal to us).
Divergence is where things get most interesting and also forces human to synthesize and make the decision, therefore preserving unique human contribution and decision. This is not necessarily why I developed it. I think it’s most effective and efficient way to do work and get the most out of LLMs.
**Multi-Origin Divergence Adversarial Council
Authorized disclosure under Averitas Holdings LLC → Incepta Labs sublicense pending. Original inventor and owner: Dr. Melinda B. Chu
This paper is also available at:
https://t.co/DR5I1mtmCz
Agree; avoiding sycophancy and actively encouraging divergent analysis is essential for more trustworthy human–AI reasoning.
Just published: “An Antidote to Sycophancy: Toward Epistemic Divergence in Human–AI Reasoning”
It introduces MODAC identical prompts run across independent LLMs (different vendors, tabula-rasa conditions). Divergence is deliberately preserved as diagnostic signal, not noise. Human remains the final adjudicator. It is an LLM corollary to Medical Grand Rounds.
https://t.co/33HVff4QYP
https://t.co/ASUksMPgk2
This blew my mind. 🤯 I always thought based on media / hype that with new AI /Tech Bio companies were that you have idea, start the project, get list of compounds, synthesize them, ➡️ automated lab ➡️ Phase 1 human study, and then “cure ALL diseases in 5-10 years.” That’s what people say.
🚨 So I figured most TechBio would have to pivot to drug development b/c that’s wheee th real value in Biotech is and there’s not enough customers (other than Big Pharma) for all these SaaS.
🚨BUT looking into this, I learned that most companies only do 1 small step in drug design, so in theory you would need 6-10 of them to actually get to compound synthesis.
🚨This means 2 things:
1.). Means they can’t really pivot to developing drug assets b/c it’s only 1 small step of process.
2.) Reinforces my view that there aren’t enough customers for each SaaS
🌟 Check out this technical paper where I go into this in more detail and demonstrate my multi-model orchestration approach.
@jack Documentation and Validation of Human Provenance in in Human+AI era will be most important.
https://t.co/cJsIShrn0H
Human Conception Ledger (HCL)
https://t.co/JhmrJai3zp
Team Contribution Attribution Ledger (HCL)
https://t.co/L2RB03d5yF
This is also available at:
https://t.co/uZQQ93IbW8
Authorized disclosure under Averitas Holdings LLC → Incepta Labs sublicense pending. Original inventor and owner: Dr. Melinda B. Chu