Concerning AI | Existential Risk From Artificial Intelligence

Concerning AI | Existential Risk From Artificial Intelligence

By Brandon Sanders & Ted SarvataSociety & CultureReligion & SpiritualityTechnologyPhilosophy
Download on the App Store

Concerning AI | Existential Risk From Artificial Intelligence episodes

  • 0040: If it were superintelligent, it would be hard to argue with

    Concerning AI is the podcast about two questions:

    • Is there a credible existential threat to humanity from a future superhuman AI?
    • If yes, what do we do about it?

      • We thought we’d start to catalog possible answers to that 2nd question. Here’s what we have so far:

      • As Fast As Possible (AFAP). We discussed this a few episodes ago: Concerning AI: Episode XXXIV – A New Hope.

        • Listener Feedback

        • Randy’s email
          • Thinking, Fast and Slow book by Daniel Kahneman
          • Eric’s voice memo (Yes, please do more of this everyone! Send them to [email protected])
            • Adrian Tchaikovsky’s novel Children of Time
            • Rico’s email about augmenting humans as a way to AI safety
            • Elon’s new company Neuralink
            • We barely touched on it on the show, but this whole thread is gold: intentional anthropomorphization thesis, inclding John’s quote

              • Subscribe in Overcast, iTunes, or through our feed. Please send your feedback for us to use on the show as a voice memo attachment in an email to [email protected]

              recorded April 2, 2017

              http://traffic.libsyn.com/friendlyai/ConcerningAI-episode-0040-2017-04-02.mp3

               

              Concerning AI | Existential Risk From Artificial Intelligence is made with Jetpack Podcast. Full show notes and every episode at https://concerning.ai/

              0 min
            • 0039: We Need More Sparrow Fables

              (0:00) The Unfinished Fable of the Sparrows, from Superintelligence by Bostrom

              Followup to 0038: We Don’t Want to Die

              • (3:12) fitness / utility function and “Is this all there is”
              • (5:33) Joy VanBuskirk‘s comment and efficiency

                • (9:19) Orthogonality Thesis: Orthogonality Thesis at Less Wrong

                (23:57) Instrumental Convergence Thesis

                (31:24) What is the connection?

                (32:57) Call to action

                (33:29) Would it make sense to explore the two dimensional space of types of goals vs levels of intelligence?

                (35:26) We need more fables. Send voice memos (preferred), images or text to [email protected]

                (39:00) After-show: post-hoc justification for this episode

                http://traffic.libsyn.com/friendlyai/ConcerningAI-episode-0039-2017-03-25.mp3

                Concerning AI | Existential Risk From Artificial Intelligence is made with Jetpack Podcast. Full show notes and every episode at https://concerning.ai/

                0 min
              • 0038: We Don’t Want to Die

                This show is about two questions:

                • Is there a credible existential threat to humanity from a future superhuman AI?
                • If yes, what do we do about it?

                  • We thought we’d start to catalog possible answers to that 2nd question. Here’s what we have so far:

                  • As Fast As Possible (AFAP). We discussed this a few episodes ago: Concerning AI: Episode XXXIV – A New Hope.
                  • The Locality Effect. We talked about this back on episodes 0026: The Locality Principle and 0027: Listener Feedback and the Locality Effect

                    • Folowup

                    • 0036: Baby You Can Drive My Car about self-driving cars
                      • email from Randy about intention lights
                      • Ed Orman‘s idea:
                        • 0037: Listeners Gone Wild about fitness functions
                          • John Stoner‘s comment and subsequent thread
                            • Facebook Group
                            • Don’t fear super intelligent AI TED talk and resulting Facebook group discussion
                              • Grady Booch and Neil deGrasse Tyson don’t think it’s possible to build machines as smart as us
                              • Book recommendation
                                • http://traffic.libsyn.com/friendlyai/ConcerningAI-episode-0038-2017-03-12.mp3

                                Concerning AI | Existential Risk From Artificial Intelligence is made with Jetpack Podcast. Full show notes and every episode at https://concerning.ai/

                                0 min
                              • 0037: Listeners Gone Wild

                                Listener Voicemail & Comments

                                • Eric’s voicemail
                                • Evan’s comment (Our interview with Evan: ep 0011: Evan Prodromou, AI practitioner (part 1), ep 0012: Evan Prodromou, AI practitioner (part 2)
                                • John’s comment

                                  • Ted got the author’s name wrong Predictably Irrational by Dan Ariely

                                  (25:12) Moving on from Feedback into what’s going to get us from here to there

                                  • Instantaneous Judgement (Stimulus-Response)
                                  • Reinforcement Learning
                                    • http://traffic.libsyn.com/friendlyai/ConcerningAI-episode-0037-2017-02-26.mp3

                                    Concerning AI | Existential Risk From Artificial Intelligence is made with Jetpack Podcast. Full show notes and every episode at https://concerning.ai/

                                    0 min
                                  • 0036: Baby You Can Drive My Car

                                    Main topic of this show: Unexpected Consequences of Self Driving Cars by Rodney Brooks

                                    Send us your feedback:

                                    1. Use your phone’s voice memo app to record a minute or less message and send to [email protected]
                                    2. Drop us a note at the Concerning AI Facebook group.

                                      3. Subscribe in Overcast, iTunes, or through our feed.

                                      http://traffic.libsyn.com/friendlyai/ConcerningAI-episode-0036-2017-02-19.mp3

                                      Concerning AI | Existential Risk From Artificial Intelligence is made with Jetpack Podcast. Full show notes and every episode at https://concerning.ai/

                                      0 min
                                    3. 0035: New Water Story

                                      What should our values be?

                                      Red Mars, Blue Mars, Green Mars

                                      Elon: How (and Why) SpaceX Will Colonize Mars (Wait But Why)

                                      Putting Time in Perspective (Wait But Why)

                                      Join in! What’s your candidate story as a replacement for the consumption story? Concerning AI Facebook group

                                      Article about the New Story. http://yesmagazine.org/pdf/kortennewstory.pdf

                                      Discuss this episode: https://www.facebook.com/groups/ConcerningAI/permalink/1364502150268703/

                                      Subscribe in Overcast, iTunes, or through our feed.

                                      http://traffic.libsyn.com/friendlyai/ConcerningAI-episode-0035-2017-02-12.mp3

                                      Concerning AI | Existential Risk From Artificial Intelligence is made with Jetpack Podcast. Full show notes and every episode at https://concerning.ai/

                                      0 min
                                    4. Concerning AI: Episode XXXIV – A New Hope

                                      (0:32) AI Panel Superintelligence: Science or Fiction? – (youtube) (discussion in Concerning AI Facebook group)

                                      (1:14) Going as fast as we can now and hardware overhang

                                      (6:50) Tell us why this is wrong

                                      (7:08) What are you excited about with AI? Philosophy on a deadline. Making us better humans

                                      (19:21) CPG Gray video Humans Need Not Apply

                                       

                                      http://traffic.libsyn.com/friendlyai/ConcerningAI-episode-0034-2017-02-05.mp3

                                      Concerning AI | Existential Risk From Artificial Intelligence is made with Jetpack Podcast. Full show notes and every episode at https://concerning.ai/

                                      0 min
                                    5. 0033: Mind Game and the Curse of Dimensionality

                                      Wind up your propeller hats! This one is a doozy. Hopefully someone can explain it to me (Ted).

                                      (0:00) Welcome

                                      (0:54) Followup (link)

                                      Rodney Brooks Article link

                                      What’s your prediction for superintelligence (in numbers)? (link)

                                      (6:04) Thought Exercise – Gradient

                                      (19:20) Curse of Dimensionality

                                      (30:28) Why do this thought experiment?

                                      What are your favorite visualizations?

                                      http://traffic.libsyn.com/friendlyai/ConcerningAI-episode-0033-2017-01-29.mp3

                                      Concerning AI | Existential Risk From Artificial Intelligence is made with Jetpack Podcast. Full show notes and every episode at https://concerning.ai/

                                      0 min
                                    6. 0032: Westworld

                                      If you’d like to send us links or better show notes, please feel free and we’ll add them here. Best way to get in touch is the Concerning AI facebook group.

                                      (0:20) Brandon derails us to propose a labeling system so we could talk about our various points of view. Brandon is a 5-30-70. Ted is a 15-30-50. Where are you?

                                      (7:59) Self-driving cars. Too bad we hadn’t read this before recording! http://rodneybrooks.com/unexpected-consequences-of-self-driving-cars/

                                      (8:35) Ted fails to give credit to Roy Amara.

                                      (9:27) Westworld – SPOILER ALERT

                                      Wait But Why ladder of intelligence (turns out it’s a staircase)

                                      (34:00) Teasing next episode

                                      http://traffic.libsyn.com/friendlyai/ConcerningAI-episode-0032-2017-01-22.mp3

                                      Concerning AI | Existential Risk From Artificial Intelligence is made with Jetpack Podcast. Full show notes and every episode at https://concerning.ai/

                                      0 min
                                    7. 0031: Listener Feedback

                                      Too time constrained for show notes this time. If you want to send us notes to be added here, please do it! Best place to reach is is the Concerning AI group on Facebook. All of the listener feedback in this episode comes from that group. Thank you all!

                                      Subscribe in Overcast, iTunes, or through our feed.

                                       

                                      http://traffic.libsyn.com/friendlyai/ConcerningAI-episode-0031-2017-01-15.mp3

                                      Concerning AI | Existential Risk From Artificial Intelligence is made with Jetpack Podcast. Full show notes and every episode at https://concerning.ai/

                                      0 min

                                    About Concerning AI | Existential Risk From Artificial Intelligence

                                    From the publisher's feed

                                    Is there an existential risk from Human-level (and beyond) Artificial Intelligence? If so, what can we do about it?