Summary of #82 - Dr. JOSCHA BACH - Digital Physics, DL and Consciousness [UNPLUGGED]

This is an AI generated summary. There may be inaccuracies.
Summarize another video · Purchase summarize.tech Premium

00:00:00 - 01:00:00

In this video, Dr. JOSCHA BACH discusses digital physics and consciousness. He argues that, even if the universe is limited in some ways, the human mind can still understand truths that are not provable within formal systems. He concludes by asking the audience if they can think of an example of something that is infinite in spatial extent, and the audience is unable to come up with an answer.

00:00:00 Dr. Joshua Bach is a cognitive scientist who focusses on cognitive architectures models of mental representation, emotion, motivation and sociality. He is also the host of the popular MLST podcast. In this episode, he discusses the relevance of computation for artificial general intelligence (AGI), Go Dell and consciousness. He explains that computation is easier than most people think and that it refers to the ability to decompose complex phenomena into individual steps. He also discusses computational models and how they can be used to understand the world. Joshua talks about the Turing machine and how it is a powerful tool for understanding computation. He concludes the interview by discussing how computational models are important for teaching computer science, and how they can be used to help us understand complex phenomena.
00:05:00 In this video, Dr. JOSCHA BACH discusses digital physics and consciousness. He argues that, even if the universe is limited in some ways, the human mind can still understand truths that are not provable within formal systems. If this is true, then our brains might be what Turing referred to as "oracles," computers with access to non-computable oracles that can be utilized for hyper computation. He asks his audience are they open to the possibility that our brains are oracles, and if not, what is their response to penrose's arguments. In his opinion, Guru was misunderstood by philosophers, and his conclusion was that there is something fundamentally going on with mathematics that doesn't change when you call it a call the function again. This, in turn, suggests that classical mathematics doesn't work. However, computers still cannot build any kind of mathematics, as it doesn't allow for the construction of a machine.
00:10:00 Dr. Joscha Bach discusses the concept of geometry, and how it is an approximation at various levels of description in reality. He discusses the implications of this for foundational physics, and how machines that can learn at multiple levels will be necessary for AGI.
00:15:00 Dr. JOSCHA BACH discusses the possibility that the universe is infinite in spatial extent, and explains how this could be impossible to conceive of without using a language that has bugs in it. If we use a language that has no contradictions, however, then the stuff that we express in that language is not meaningful. He concludes by asking the audience if they can think of an example of something that is infinite in spatial extent, and the audience is unable to come up with an answer.
00:20:00 Infinities and continuums play a role in standard models of physics, but these models may be flawed because they cannot account for Infinities. Joshua Bach discusses this paradox with a focus on deep learning and large language models.
00:25:00 Dr. JOSCHA BACH discusses the limitations of deep learning, and argues that symbolic methods are still ahead of deep learning in certain cases. Marcus responds by arguing that deep learning is hitting a "wall," and that we need to replace it with more scripts. Bach argues back that deep learning is not limited, and that other approaches are receiving more money.
00:30:00 The author discusses the concerns that are raised about the current state of deep learning, specifically its ability to generalize and its reliance on deep neural networks. He argues that other algorithms, such as evolutionary algorithms, can be used to complement deep learning and that the state of the field is not necessarily in a crisis.
00:35:00 According to the speaker, deep learning models can overfit and get very good at the training data but may not perform as well in the real world. One way to avoid overfitting is to use more training data, and to avoid spurious correlations in the training data.
00:40:00 In this video, Dr. JOSCHA BACH discusses his research into digital physics and consciousness. He notes that because direct adjacency between objects is not well suited for understanding complex structures, he developed ordered pair statistics instead. This led him to realize that grammar and semantics can be extracted from a sentence's structure. He also discusses how large language models could be used for information retrieval.
00:45:00 Dr. Joscha Bach discusses the relationship between digital physics, DL, and consciousness. He points out that while these models may improve our productivity, they may also spoil society. He believes that this is inevitable, but that the creative process itself is magical.
00:50:00 Dr. JOSCHA BACH discusses the possibility of digital physics, DL, and consciousnes
00:55:00 Dr. JOSCHA BACH discusses the mysterious phenomenon of consciousness, and compares it to the theory of information processing over cells. He believes that consciousness can be measured in terms of an integrated information theory, and that it can be understood in terms of a story that the physical system tells itself.

01:00:00 - 01:15:00

In this video, Dr. Joscha Bach discusses his controversial theory that self-consciousness arises from the ability to make deductions about the consequences of one's own actions. He believes this requires a counter factual depth in one's observations, which grounds inference. Some argue that this theory requires consciousness in its own right, but Bach disagrees.