Science, Technology & the Future | Philosophy of Artificial Intelligence - Graham Oppy @scfu | Uploaded September 2024 | Updated October 2024, 3 hours ago.
Professor Graham Oppy discusses the Turing Test, whether AI can understand, whether it can be more ethical than humans, moral realism, ai alignment, incoherence in human value, indirect normativity and much more.
Chapters:
0:00 The Turing test
6:06 Agentic LLMs
6:42 Concern about non-anthropocentric intelligence
7:57 Machine understanding & the Chinese Room argument
10:21 AI 'grokking' - seemingly understanding stuff
13:06 AI and fact checking
15:01 Alternative tests for assessing AI capability
17:35 Moral Turing Tests - Can AI be highly moral?
18:37 Philosophy's role in AI development
21:51 Can AI help progress philosophy?
23:48 Increasing percision in the language of philosophy via technoscience
24:54 Should philosophers be more involved in AI development?
26:59 Moral realism & fining universal principles
31:02 Empiricism & moral truth
32:09 Challenges to moral realism
33:09 Truth and facts
36:26 Are suffering and pleasure real?
37:54 Signatures of pain
39:25 AI leaning from morally relevant features of reality
41:22 AI self-improvement
42:36 AI mind reading
43:46 Can AI learn to care via moral realism?
45:42 Bias in AI training data
46:26 Metaontology
48:27 Is AI conscious?
49:45 Can AI help resolve moral disagreements?
51:07 'Little' philosophical progress
54:09 Does the human condition prevent or retard wide spread value convergence?
55:04 Difficulties in AI aligning to incoherent human values
56:30 Empirically informed alignment
58:41 Training AI to be humble
59:42 Paperclip maximizers
1:00:41 Indirect specification - avoiding AI totalizing narrow and poorly defined goals
1:02:35 Humility
1:03:55 Epistemic deference to 'jupiter-brain' AI
1:05:27 Indirect normativity - verifying jupiter-brain oracle AI's suggested actions
1:08:25 Ideal observer theory
1:10:45 Veil of ignorance
1:13:51 Divine psychology
1:16:21 The problem of evil - an indifferent god?
1:17:21 Ideal observer theory and moral realism
See Wikipedia article on Graham Oppy: en.wikipedia.org/wiki/Graham_Oppy
x: https://x.com/oppygraham
#AI #philosophy #aisafety #ethics
Many thanks for tuning in!
Please support SciFuture by subscribing and sharing!
Have any ideas about people to interview? Want to be notified about future events? Any comments about the STF series?
Please fill out this form: docs.google.com/forms/d/1mr9PIfq2ZYlQsXRIn5BcLH2onbiSI7g79mOH_AFCdIk
Kind regards,
Adam Ford
- Science, Technology & the Future - #SciFuture - scifuture.org
Professor Graham Oppy discusses the Turing Test, whether AI can understand, whether it can be more ethical than humans, moral realism, ai alignment, incoherence in human value, indirect normativity and much more.
Chapters:
0:00 The Turing test
6:06 Agentic LLMs
6:42 Concern about non-anthropocentric intelligence
7:57 Machine understanding & the Chinese Room argument
10:21 AI 'grokking' - seemingly understanding stuff
13:06 AI and fact checking
15:01 Alternative tests for assessing AI capability
17:35 Moral Turing Tests - Can AI be highly moral?
18:37 Philosophy's role in AI development
21:51 Can AI help progress philosophy?
23:48 Increasing percision in the language of philosophy via technoscience
24:54 Should philosophers be more involved in AI development?
26:59 Moral realism & fining universal principles
31:02 Empiricism & moral truth
32:09 Challenges to moral realism
33:09 Truth and facts
36:26 Are suffering and pleasure real?
37:54 Signatures of pain
39:25 AI leaning from morally relevant features of reality
41:22 AI self-improvement
42:36 AI mind reading
43:46 Can AI learn to care via moral realism?
45:42 Bias in AI training data
46:26 Metaontology
48:27 Is AI conscious?
49:45 Can AI help resolve moral disagreements?
51:07 'Little' philosophical progress
54:09 Does the human condition prevent or retard wide spread value convergence?
55:04 Difficulties in AI aligning to incoherent human values
56:30 Empirically informed alignment
58:41 Training AI to be humble
59:42 Paperclip maximizers
1:00:41 Indirect specification - avoiding AI totalizing narrow and poorly defined goals
1:02:35 Humility
1:03:55 Epistemic deference to 'jupiter-brain' AI
1:05:27 Indirect normativity - verifying jupiter-brain oracle AI's suggested actions
1:08:25 Ideal observer theory
1:10:45 Veil of ignorance
1:13:51 Divine psychology
1:16:21 The problem of evil - an indifferent god?
1:17:21 Ideal observer theory and moral realism
See Wikipedia article on Graham Oppy: en.wikipedia.org/wiki/Graham_Oppy
x: https://x.com/oppygraham
#AI #philosophy #aisafety #ethics
Many thanks for tuning in!
Please support SciFuture by subscribing and sharing!
Have any ideas about people to interview? Want to be notified about future events? Any comments about the STF series?
Please fill out this form: docs.google.com/forms/d/1mr9PIfq2ZYlQsXRIn5BcLH2onbiSI7g79mOH_AFCdIk
Kind regards,
Adam Ford
- Science, Technology & the Future - #SciFuture - scifuture.org
