TAILOR Conference Lisbon ‘24

Notes from Notebook

• 
• 
• 
• 
• 
• 
• 
• 
• 
• 
• 
• 
• 

OCR Notes - processed with ChatGPT

• Wendy Ju
  • Robots should have good manners.
  • Engagement.
  • People treat robots like children.
• Carles Sierra
  • People are nervous about AI.
  • Psychopaths
  • How is this related to AI?
  • Adapt technology as a community.
  • Sustainable collective.
  • Governing the commons.
  • L’Horta watering communities.
  • Groups talking about technology.
  • Ostrom’s criteria.
  • Community arms.
• “I must be…”
  • Valve
  • Goals
  • Tailor thing
• Turns out Isabelle is part of this, but they ran out of funds so she couldn’t come. XD
  • Logistics
  • Breakout Room 5
  • “Why!”
  • “I think it’s cool to assign teams & make people uncomfortable.”
  • “I am a leader!! (ugh.)”
  • Short presentation
  • Neuro
  • Uncertainty in humans:
    • State of mind
    • Semantic
  • Uncertainty in life is short.
  • Reasoning shortcuts in NS.
    • Deterministic vs probabilistic
    • Concept rehearsal
  • Feeling weird as a fellow lab member.
  • Internal vs External symbolic neuro
  • Symbolic AI
    • Pointing → Saliency
    • Wording → RLHF
    • Approximate retrieval (VSRAGIL)
    • He just destroyed LIM prompts.
    • Using the internet as training
  • LIMs → Good at common sense.
  • Approximate omniscience
  • LIMs are masters of style.
• She hasn’t stopped for one second.
  • KL also now available.
  • WTF did she actually use, and how is she keeping track of outages?
• Glioblastoma → Dead.
  • Low data
  • Genetic expressions
  • Multimodal classification
  • Patching, how?
  • Pankaj Pandey → FNFIELD
  • Started in Norway
  • Adaptive, green, trustworthy
  • Dislike (hate) stable diffusion
• Explainable Malware Detection
  • Ontologies as explanation?
  • Concept learning, but what model?
  • They published a lot though.
• Trustworthy robotic teammates
  • Humans seem to trust robots less initially but more over time compared to other humans.
  • What forms the expectation? Set low expectations.
• Collaborative Human-AI
  • Trusting black-box models
  • Started with the theory of mind, now game theory
  • AI is the new steam engine?
  • Transpilation 2. (Quantum)
• Federalized NSGA-II
  • Soft unification
  • Class distribution bias
  • Spectrogram → Original biological component.
  • Perturbation → Robustness distribution.
  • What kind of perturbations?
  • How distribution?
  • Safety requirements.
  • Operation
• Standardization & legal means.
  • Fundamental research → End user?
  • B2B vs B2C
  • Companies don’t care about research.
  • Academia doesn’t care about the product.
  • AI: No time to wait for results, collaborative research.
  • Transfer Lab = Industry + Academia.
  • AI combines the chain, but customers own the data, not us.
  • Trust the platform.
  • European data space.
  • Business model → Safety
  • Standardization & funding
  • B2B
• People are skeptical of their own abilities.
  • The more you know, the less you trust.
  • Replica AI → Social AI?
  • Tailor individual collaboration.
  • Certification for AI?
  • Politics
  • Methods for trusted AI?
  • Evaluation tools.
  • Neural explicit models.
  • Valves → Code: Schwartz theory
  • Goals are proxies of values.
  • Optimize for fairness.
  • Privacy vs Trust.
  • Choices
  • Differences in values across cultures.