00:00:00 Intro. Bogdan's journey through the French system towards PhD in AI. Inspiration by early DeepMind papers, research on LSTM and other recurrent architectures.

00:05:29 Oxford postdoc between ML and Neuroscience, theory of mind. Turn towards safety. Influence of Nick Bostrom's 2014 book Superintelligence https://en.wikipedia.org/wiki/Superintelligence:_Paths,_Dangers,_Strategies.

00:07:29 AI acceleration. AlphaGo, Startcraft, BART, GPT-2. Language. Learning people's preferences.

00:10:24 The curious path of an independent AI safety researcher. 80000 hours https://80000hours.org/, the Alignment Forum https://www.alignmentforum.org/, LessWrong https://www.lesswrong.com/, effective altruism.

00:18:15 GPT: in-context learning, math.

00:23:47 GPT: planning and thinking. Planning in the real world (reinforcement learning) vs planning in a math proof, planning as problem solving.

00:27:29 GPT: chain of thought. "Let's think about this step by step."

00:31:47 GPT: lying? HAL from 2001 Space Odyssey. Does GPT have the will to do something? Simulators, Bayesian inference, simulacra, autoregressivity. The surprising coherence of GPT-4. Playing personas.

00:43:38 GPT: reinforcement learning with human feedback. RLHF is like an anti-psychotic drug? Or cognitive behavioral therapy?

00:45:38 GPT: Vervaeke's dialog Turing test https://balazskegl.substack.com/p/gpt-4-in-conversation-with-itself.

00:52:36 AI Safety. The issue of timescale. The OpenAI initiative https://openai.com/blog/our-approach-to-ai-safety. Aligning by debating.

00:57:46 Direct alignment research; Bogdan's pessimism.

The 2-step approach: automate alignment research. Who will align the aligner AI?

01:04:11 Alignment by giving agency to AI. Embodiment. Let them confabulate but confront reality.

01:12:09 Max Tegmark's waterfall metaphor. Munk debate on AI https://www.youtube.com/watch?v=144uOfr4SYA, Yoshua Bengio's interview https://www.youtube.com/watch?v=0RknkWgd6Ck.

01:22:21 Open source AI. George Hotz interview https://www.youtube.com/watch?v=dNrTrx42DGQ. Bogdan's counterargument: engineering a pandemic. Some tools make a few people very powerful.

01:28:15 Adversarial examples.

01:31:32 Bogdan's dreams and fears, where are we heading?

Hosted on Acast. See acast.com/privacy for more information.

00:00:00 Intro.

00:01:34 Jonas' story. Math, physics towards computer science, AI, and robotics.

00:05:09 Embodiment in intelligence.

00:07:25 LLMs in robots. Should we put LLMs in robots? Can we teach them as kids if they can speak? Will they develop their personalities depending on their unique experience? Should we add episodic memory to them?

00:19:02 Relevance realization. How to filter important information from the immense incoming flow of signals? The top-down aspect of perception. Cultural learning and binding.

00:23:30 Limits of GI. Is there an inherent limit to how intelligent a being can be? What if too much intelligence makes the map take over the agent, leading to something like schizophrenia?

00:29:20 Continual learning. The brain is a little scientist. The scientific method: where do the hypotheses come from? Where does the value of a proposition come from? How do we decide what proposition to prove or what experiment to run? Why did I work on the Higgs boson?

00:37:37 Dog intelligence. Do dogs want to "go beyond" what is "visible", or is it a purely human drive?

00:40:00 Collective alignment. Higher level collective consciousness and its relationship to human and AI alignment.

00:47:21 AGI. How far are we from AGI?

00:50:38 Robots. Bodies are the bottleneck of robotics research.

00:57:09 Jonas' dream: connecting the dots, merging the cognitive modules and experiment in the real world, towards a dog intelligence in 5-10 years.

01:01:13 High-functioning zombies. Should we be afraid of the them: an agent smart enough to plan, but not smart enough to see the harm that some of the planned actions may cause?

I, scientist blog: https://balazskegl.substack.com

Twitter: https://twitter.com/balazskegl

Artwork: DALL-E

Music: Bea Palya https://www.youtube.com/channel/UCBDp3qcFZdU1yoWIRpMSaZw

Hosted on Acast. See acast.com/privacy for more information.

1:35 Screwdriver hands: how Giuseppe became a scientist.

6:07 Reinforcement learning (RL) explained to Giuseppe's grandma.

9:01 Model-based reinforcement learning: how to fry an egg.

10:45 The three components of model-based reinforcement learning: the actor, the model, and the planner.

16:29 Planning = thinking: ham & eggs and Google map

19:01 RL is responsible for collecting its own data

22:05 Vervaeke's first two Ps: propositional (GPT) and procedural (RL)

24:05 Fear from AI: the paperclip scenario, evil, indifference, and foolishness

28:19 Should we add agency into AI? Can we socialize AI as we do with kids?

32:22 Do we need to embody AI to align it? Can GPT bike? Is textual knowledge everything?

36:28 Why to aspire to the dream of creating AI? 1) Why not :)? 2) Curiosity. 3) To understand ourselves better.

41:52 How AI changed our lives. Recommendation engines, addiction, jujitsu, conspiracy theories, the attention economy.

49:12 Open source large language models. AI as a mirror. Individuation of LLMs.

51:16 Putting AI into things: they will individuate.

52:56 Vervaeke's 2nd person Turing test. Let GPTs talk to each other.

54:37 Could AI manipulate us?

56:11 Closing.

I, scientist blog: https://balazskegl.substack.com

Twitter: https://twitter.com/balazskegl

Artwork: DALL-E

Music: Bea Palya https://www.youtube.com/channel/UCBDp3qcFZdU1yoWIRpMSaZw

Hosted on Acast. See acast.com/privacy for more information.