AI Sentience, Agency and Catastrophic Risk with Yoshua Bengio - 654

Поділитися
Вставка
  • Опубліковано 14 тра 2024
  • Today we’re joined by Yoshua Bengio, professor at Université de Montréal. In our conversation with Yoshua, we discuss AI safety and the potentially catastrophic risks of its misuse. Yoshua highlights various risks and the dangers of AI being used to manipulate people, spread disinformation, cause harm, and further concentrate power in society. We dive deep into the risks associated with achieving human-level competence in enough areas with AI, and tackle the challenges of defining and understanding concepts like agency and sentience. Additionally, our conversation touches on solutions to AI safety, such as the need for robust safety guardrails, investments in national security protections and countermeasures, bans on systems with uncertain safety, and the development of governance-driven AI systems.
    🔔 Subscribe to our channel for more great content just like this: ua-cam.com/users/twimlai?sub_confi...
    🗣️ CONNECT WITH US!
    ===============================
    Subscribe to the TWIML AI Podcast: twimlai.com/podcast/twimlai/
    Join our Slack Community: twimlai.com/community/
    Subscribe to our newsletter: twimlai.com/newsletter/
    Want to get in touch? Send us a message: twimlai.com/contact/
    📖 CHAPTERS
    ===============================
    00:00 - Catching up with Yoshua
    04:26 - Bayesian learning and causal theories
    07:13 - AI risks
    13:34 - Human-level competence in AGI
    15:13 - Power concentration
    16:55 - Risks in LLMs
    21:17 - Dangers of AGI
    23:47 - Sentience and agency
    25:13 - Relating agency and goal conditioning in RL
    28:24 - Sentience
    30:30 - Difference between sentience and consciousness
    36:33 - Social impacts of AI
    40:30 - Approaching risks from a research perspective
    44:12 - AI safety solutions
    49:26 - Letter on moratorium on AI research
    50:35 - Knowledge gaps in AI
    54:35 - Potential risks of unintended AI behavior
    56:20 - Pointer and resources
    🔗 LINKS & RESOURCES
    ===============================
    Flow Network based Generative Models for Non-Iterative Diverse Candidate Generation - papers.nips.cc/paper_files/pa...
    FAQ on Catastrophic AI Risks (Superhuman AI) - yoshuabengio.org/2023/06/24/f...
    Pushing Back on AI Hype with Alex Hanna - 649 - twimlai.com/podcast/twimlai/p...
    The Montreal Declaration for a Responsible Development of Artificial Intelligence - recherche.umontreal.ca/englis...
    My testimony in front of the U.S. Senate - The urgency to act against AI threats to democracy, society and national security - July 2023 - yoshuabengio.org/2023/07/25/m...
    Pause Giant AI Experiments: An Open Letter (Letter on Moratorium on AI Research) - futureoflife.org/open-letter/...
    For a COMPLETE LIST of links and references, head over to twimlai.com/go/654.
    📸 Camera: amzn.to/3TQ3zsg
    🎙️Microphone: amzn.to/3t5zXeV
    🚦Lights: amzn.to/3TQlX49
    🎛️ Audio Interface: amzn.to/3TVFAIq
    🎚️ Stream Deck: amzn.to/3zzm7F5
  • Наука та технологія

КОМЕНТАРІ • 4

  • @user-yt4wf3bv3x
    @user-yt4wf3bv3x 6 місяців тому

    Interesting talk. But, i think that wrt to discussions of whether agents feel pain - it has to be kept in mind that the interactions we have with agents are purely cognitive - asking about something that is physical in nature is just not a viable question. These agents are embodiments of mind - not physical beings. I would like to see that aspects of such discussions diminished...

    • @danieljones5195
      @danieljones5195 5 місяців тому

      Cognitive pain is the worst pain I've experienced in my whole life...

  • @hehehehe1828
    @hehehehe1828 6 місяців тому

    Sometimes my head was hurting from the sudden cuts from 1-person view to the 2-person view and back. Most of the cuts were pointless, just to show the host nodding. My advice: show just the person speaking and try to minimize time you do split screen. Show either the host or the guest, but not both on splitscreen. Joe Rogan does it well

  • @gustafa2170
    @gustafa2170 5 місяців тому

    "That doesn't count for sentience... It has to be connected to emotion. There is the word motion there in the root which has to do with action. Acting to achieve or avoid some situation, which isn't really what's going on with these LLMs".
    He seems clueless about consciousness. Par for course for these AI researcher that are trying to study something they haven't spent the time to examine directly using meditation, introspection ,etc. Emotion is a felt, direct, experience.