Video

DVIDSVideoPlayer

Video by Kevin D Schmidt

Dr. Daniel Dennett

Air Force Research Laboratory

March 10, 2023 | 01:12:58

In this edition of QuEST, leading thinker and writer Dr. Daniel Dennett discusses “The concept of cherry-picking in AI”. Dr. Dennett has been writing about AI for 50 years and is a senior advisor on the Canadian Institute of Advanced Research, Mind, Brain, and Consciousness Project.
Key Moments and Questions in the video include:
- First main point Dennett makes is about resisting the move to denying ChatGPT, GPT3, etc will never get to ‘real understanding’
- Second main point Dennett makes is “we shouldn’t try to make counterfeit people”
- Ensure the penalties for making counterfeit consciousness/people are so high as to be prohibitive.
- Question: Can I ask a point of clarification on what you were just saying. When you’re talking about counterfeit people, are you limiting that to trying to pass off systems like GPT3 as being people when they are not? Or would you consider any sort of a…non-meat human to be a counterfeit human or would you be open to that possibility?
- Notes that none of our existing technologies that we have are people; key here would be that nobody would be fooled by the AI (into thinking it was human)
- Concept of the zombie with zimbos - ‘there’s no special sauce that a human conscious being has and a zimbo doesn’t’
- The idea of an “off-switch” and the human desire to “kill the off-switch”
- Question: Is your advice kind of saying we shouldn’t try to make humanoids with social attention mechanisms because it could easily take us into the direction of the things that you’re concerned about?
- COG and the readiness of very intelligent people to be fooled into thinking COG was conscious
- Discussion of the paperclip apocalypse and Bostrom
- Questions of the future of humans, including references to Douglas Hofstadter
- Moth with ants painted on its wings, which birds are scared of - example of natural selection
- Caution in creating and employing AI systems