Video Series

IFoRE Sigma Xi Conference 2023: Dr. Craig Kaplan provides an overview of AGI and explains the safety challenges and current AI approaches .

A short window exists during which humans can influence the development of AGI and SuperInteligence.

Narrow AI is rapidly developing into AGI which will then inevitably lead to SuperIntelligence much smarter and more powerful than humans.

Constitutional AI is a tuning approach whereby researchers first train one AI on what is right or wrong based on a “constitution” written b

Customize AAAIs by interacting with them, teaching them both our knowledge and our values.

Customizable AI agents (AAAIs) and the universal problem solving framework are 2 key technical elements of the fastest/safest path to AGI.

Dr. Kaplan argues that the fastest and safest path to AGI involves building a collective intelligence network of humans and AI agents.

AGI is potentially a “winner-take-all” scenario where the first AGI will likely dominate slower approaches.

Nearly half of AI experts surveyed say there is a 10% or greater chance of extinction by AI.

How will Artificial Intelligence determine what is right and what is wrong?

Could we lock up AI like we secure plutonium or other dangerous technology? Can we prevent it from falling into the hands of bad actors?

Regulation is a standard answer for dangerous technologies, but there are problems with this approach. Dr. Kaplan discusses three major chal

A presentation by Dr. Craig A. Kaplan at the IFoRE / Sigma Xi Conference on 11/10/23.

There is a relatively short window during which humans can influence the development of AGI and SuperIntelligence.

A recap of the main points of the Surviving AI series.

An overview of multiple pending patents describing in detail how to build safe and ethical Artificial General Intelligence (AGI).

Three ways are explained to increase the safety of AGI.

There is more than one way to teach an AI.

iQ Company research shows that it is possible to change LLM behavior.