Cornell research on AI storytelling and training seeds

← Back to Blog

A great
article from Cornell University on AI storytelling that opens up questions
about the foundations of AI LLMs and the challenges that some of us see with
AI on a daily basis. Not all AI models were trained on the Cornell Wildcat
dataset, but it was quite influential in training some AI models.

The article demonstrates that the Seeds used to train models could be
influential in the responses that AI models generate, in this case, to the
questions Tell me a story………….and the tracing of story traits to
training data.

While it’s a bit tangential, what does this mean for AI outputs where the
parameter models used do not contain all or enough data to provide a
reasonable, trusted response for topics such as cybersecurity, risk, and
assurance?

Can current AI models be relied upon, solely, to provide repeatable,
accurate evaluations of risk, compliance, and assurance over control design
and control effectiveness alone, or do they require guidance to close gaps in
existing parameter models and training data? Humans have a propensity for
chain of thought that is based on many years of knowledge, experience, and
skill, which AI does not yet have.

Aegis Labs Faisal Khan Liz Sandwith CFIIA Ryan Buckner

#ai #artificalintelligence #chainofthough

https://lnkd.in/eYx7agkj

https://lnkd.in/eb242EXc

View Original Post https://www.linkedin.com/feed/update/urn%3Ali%3Ashare%3A7518616951905058816

Categories

Artificial Intelligence, Cyber Risk Management, Governance & ESG

Tags

Verified by MonsterInsights