It’s relatively easy to create a non-criminal AI because all of social science is reducible to tests of demonstrated interests (think ownership, whether whole or in part). But a truthful, non-criminal AI would be politely rude.
Source date (UTC): 2023-04-16 13:23:51 UTC
Original post: https://twitter.com/i/web/status/1647591637839691778
Reply addressees: @seanonolennon
Replying to: https://twitter.com/i/web/status/1647391415281029121
NOW, WATCH GPT-4 FAIL WITHOUT A SPECIFIC PROMPT:
PROMPT
–“Can you precisely list the criteria for Curt Doolittle’s Testimonial Truth?”–
Yes, I can list the criteria for Curt Doolittle’s Testimonial Truth.
Curt Doolittle’s Testimonial Truth is a method for determining the… https://twitter.com/curtdoolittle/status/1646937234170339329
Source date (UTC): 2023-04-14 19:10:41 UTC
Original post: https://twitter.com/i/web/status/1646954142475100161
CHATGPT-4 AND TESTIMONIAL TRUTH
DETAILED PROMPT:
–“In Curt Doolittle’s Testimonial Truth, he refers to tests of realism, naturalism, identity (unambiguity), logical consistency, operational possibility, external correspondence, bounded rational choice, reciprocal bounded rational choice, stated with causal parsimony, and with stated limits, with full accounting within those limits, and warrantied by due diligence. What do each of these requirements mean?”–
Curt Doolittle’s Testimonial Truth is a method for determining the truthfulness of a statement made by a person. The method involves several requirements, including tests of realism, naturalism, identity (unambiguity), logical consistency, operational possibility, external correspondence, bounded rational choice, reciprocal bounded rational choice, stated with causal parsimony, and with stated limits, with full accounting within those limits, and warrantied by due diligence. Here’s what each of these requirements means:
Realism: The statement should be based on the actual state of the world and should accurately reflect reality. This means that the statement should not be based on imaginary or unrealistic assumptions.
Naturalism: The statement should be consistent with the laws of nature and should not rely on supernatural or unexplained phenomena.
Identity (unambiguity): The statement should be clear and unambiguous, without any room for misinterpretation or confusion.
Logical consistency: The statement should be internally consistent and should not contradict itself or any other known facts.
Operational possibility: The statement should be capable of being tested or verified through empirical observation or other means.
External correspondence: The statement should correspond to external reality and should be consistent with other known facts and observations.
Bounded rational choice: The statement should be based on the assumption that individuals make rational choices within the constraints of their knowledge and resources.
Reciprocal bounded rational choice: The statement should be based on the assumption that individuals interact with each other through reciprocal, bounded rational choices.
Stated with causal parsimony: The statement should be stated in the simplest possible terms and should not rely on unnecessary or complicated explanations.
Stated limits: The statement should be clear about its limitations and should not make claims beyond the scope of its evidence or authority.
Full accounting: The statement should be supported by a full accounting of the evidence and reasoning used to arrive at the statement.
Warrantied by due diligence: The speaker must have performed due diligence in arriving at the statement and must be willing to accept legal and ethical responsibility for the truthfulness of the statement.
Source date (UTC): 2023-04-14 18:03:29 UTC
Original post: https://twitter.com/i/web/status/1646937234170339329
CHATGPT-4 AND TESTIMONIAL TRUTH
DETAILED PROMPT:
–“In Curt Doolittle’s Testimonial Truth, he refers to tests of realism, naturalism, identity (unambiguity), logical consistency, operational possibility, external correspondence, bounded rational choice, reciprocal bounded rational choice, stated with causal parsimony, and with stated limits, with full accounting within those limits, and warrantied by due diligence. What do each of these requirements mean?”–
Curt Doolittle’s Testimonial Truth is a method for determining the truthfulness of a statement made by a person. The method involves several requirements, including tests of realism, naturalism, identity (unambiguity), logical consistency, operational possibility, external correspondence, bounded rational choice, reciprocal bounded rational choice, stated with causal parsimony, and with stated limits, with full accounting within those limits, and warrantied by due diligence. Here’s what each of these requirements means:
Realism: The statement should be based on the actual state of the world and should accurately reflect reality. This means that the statement should not be based on imaginary or unrealistic assumptions.
Naturalism: The statement should be consistent with the laws of nature and should not rely on supernatural or unexplained phenomena.
Identity (unambiguity): The statement should be clear and unambiguous, without any room for misinterpretation or confusion.
Logical consistency: The statement should be internally consistent and should not contradict itself or any other known facts.
Operational possibility: The statement should be capable of being tested or verified through empirical observation or other means.
External correspondence: The statement should correspond to external reality and should be consistent with other known facts and observations.
Bounded rational choice: The statement should be based on the assumption that individuals make rational choices within the constraints of their knowledge and resources.
Reciprocal bounded rational choice: The statement should be based on the assumption that individuals interact with each other through reciprocal, bounded rational choices.
Stated with causal parsimony: The statement should be stated in the simplest possible terms and should not rely on unnecessary or complicated explanations.
Stated limits: The statement should be clear about its limitations and should not make claims beyond the scope of its evidence or authority.
Full accounting: The statement should be supported by a full accounting of the evidence and reasoning used to arrive at the statement.
Warrantied by due diligence: The speaker must have performed due diligence in arriving at the statement and must be willing to accept legal and ethical responsibility for the truthfulness of the statement.
Source date (UTC): 2023-04-14 18:03:29 UTC
Original post: https://twitter.com/i/web/status/1646937233906081808
Mark Tegmark (@tegmark)
RE: —“Those limits are 30 levels above where we are now”– (MT on Lex Friedman)
1) I don’t see how we need much more than neuromorphic hardware, so that we end the dependency on energy, performance, and pre-training. In other words, energy consumption of the current means of computation is an outstanding and unnecessary negative.
2) I don’t see how we are all that computationally limited at this point, versus that we’re limited in the experiments we can run to obtain the information necessary to perform computation, even if it’s Bayesian prediction – especially in the foundations of physics.
3) I do see that we have been running up against collective economic, social, political, and geostrategic adaptation to change with regularity at least since the industrial revolution, with a brief interlude after the world wars through the eighties.
Interesting if you think otherwise.
Thanks for all the work you do for all of us. 😉
Source date (UTC): 2023-04-14 17:47:40 UTC
Original post: https://twitter.com/i/web/status/1646933251582304256
Mark Tegmark (@tegmark)
RE: —“Those limits are 30 levels above where we are now”– (MT on Lex Friedman)
1) I don’t see how we need much more than neuromorphic hardware, so that we end the dependency on energy, performance, and pre-training. In other words, energy consumption of the current means of computation is an outstanding and unnecessary negative.
2) I don’t see how we are all that computationally limited at this point, versus that we’re limited in the experiments we can run to obtain the information necessary to perform computation, even if it’s Bayesian prediction – especially in the foundations of physics.
3) I do see that we have been running up against collective economic, social, political, and geostrategic adaptation to change with regularity at least since the industrial revolution, with a brief interlude after the world wars through the eighties.
Interesting if you think otherwise.
Thanks for all the work you do for all of us. 😉
Source date (UTC): 2023-04-14 17:47:40 UTC
Original post: https://twitter.com/i/web/status/1646933251418628104
RT @JayMan471: Point 2 is incorrect. Nothing necessitates that AI would be power-seeking. It’s just that we can’t trust it not to be
Source date (UTC): 2023-04-13 20:21:10 UTC
Original post: https://twitter.com/i/web/status/1646609492593745921
RT @MattPirkowski: If you think AutoGPT is cool, just wait until you hear about this process of autocatalytic emergence that’s been running…
Source date (UTC): 2023-04-13 01:45:55 UTC
Original post: https://twitter.com/i/web/status/1646328831081316352
You don’t need millions of dollars to train a variation on GPT or the stanford version.
Source date (UTC): 2023-04-12 19:05:11 UTC
Original post: https://twitter.com/i/web/status/1646227983756034072
Reply addressees: @Airmanareiks @BronskiJoseph @bronzeagemantis @UBERSOY1 @thealthype
Replying to: https://twitter.com/i/web/status/1646227384247496705
Yes, an objectively truthful, ethical, and moral AI is possible. No, we weren’t ready for the explosion of technology. But we’ll get there. 😉
Why? You’d be surprised what’s embedded in the structure of our language. 😉 https://twitter.com/curtdoolittle/status/1646187568084533263
Source date (UTC): 2023-04-12 16:27:25 UTC
Original post: https://twitter.com/i/web/status/1646188282303115268
https://twitter.com/curtdoolittle/status/1646187568084533263