title

CHATGPT + WOLFRAM - THE FUTURE OF AI!

description

Episode #110
HUGE ANNOUNCEMENT, CHATGPT+WOLFRAM! You saw it HERE first!
Pod version: https://podcasters.spotify.com/pod/show/machinelearningstreettalk/episodes/110-Dr--STEPHEN-WOLFRAM---HUGE-ChatGPTWolfram-announcement-e210qda
Support us! https://www.patreon.com/mlst
MLST Discord: https://discord.gg/aNPkGUQtc5
Stephen's announcement post: https://writings.stephenwolfram.com/2023/03/chatgpt-gets-its-wolfram-superpowers/
OpenAI's announcement post: https://openai.com/blog/chatgpt-plugins
In an era of technology and innovation, few individuals have left as indelible a mark on the fabric of modern science as our esteemed guest, Dr. Steven Wolfram.
Dr. Wolfram is a renowned polymath who has made significant contributions to the fields of physics, computer science, and mathematics. A prodigious young man too, Wolfram earned a Ph.D. in theoretical physics from the California Institute of Technology by the age of 20. He became the youngest recipient of the prestigious MacArthur Fellowship at the age of 21.
Wolfram's groundbreaking computational tool, Mathematica, was launched in 1988 and has become a cornerstone for researchers and innovators worldwide. In 2002, he published "A New Kind of Science," a paradigm-shifting work that explores the foundations of science through the lens of computational systems.
In 2009, Wolfram created Wolfram Alpha, a computational knowledge engine utilized by millions of users worldwide. His current focus is on the Wolfram Language, a powerful programming language designed to democratize access to cutting-edge technology.
Wolfram's numerous accolades include honorary doctorates and fellowships from prestigious institutions. As an influential thinker, Dr. Wolfram has dedicated his life to unraveling the mysteries of the universe and making computation accessible to all.
First of all... we have an announcement to make, you heard it FIRST here on MLST! ....
[00:00] Intro
[02:57] Big announcement! Wolfram + ChatGPT!
[05:33] What does it mean to understand?
[13:48] Feeding information back into the model
[20:09] Semantics and cognitive categories
[23:50] Navigating the ruliad
[31:39] Computational irreducibility
[38:43] Conceivability and interestingness
[43:43] Human intelligible sciences

detail

{'title': 'CHATGPT + WOLFRAM - THE FUTURE OF AI!', 'heatmap': [{'end': 419.74, 'start': 339.661, 'weight': 0.891}, {'end': 1483.835, 'start': 1441.447, 'weight': 0.742}], 'summary': "Dr. stephen wolfram's computational tools like mathematica, wolfram alpha, and wolfram language merge statistical and symbolic approaches in ai, providing computational superpowers to chatgpt, while the video also discusses the deep understanding of language models, chatgpt's language modeling capabilities, and the potential role of ai in advancing science and developing human-intelligible sciences.", 'chapters': [{'end': 339.661, 'segs': [{'end': 26.247, 'src': 'embed', 'start': 0.329, 'weight': 2, 'content': [{'end': 6.013, 'text': 'I just want to tell you something up front, which is I started using Mathematica 2.2 in 1994.', 'start': 0.329, 'duration': 5.684}, {'end': 7.654, 'text': 'And I was blown away.', 'start': 6.013, 'duration': 1.641}, {'end': 8.735, 'text': 'I thought it was awesome.', 'start': 7.754, 'duration': 0.981}, {'end': 10.416, 'text': 'I was using it as an undergraduate.', 'start': 8.775, 'duration': 1.641}, {'end': 19.422, 'text': "And by about 2000 or so, it had gotten so good at doing integrals that I had this rule of thumb that was, if Mathematica couldn't do it, I can't.", 'start': 10.916, 'duration': 8.506}, {'end': 22.964, 'text': 'Unless I had a library and probably hours and a lot of luck.', 'start': 19.842, 'duration': 3.122}, {'end': 26.247, 'text': "Yes No, I think it's past humans.", 'start': 23.585, 'duration': 2.662}], 'summary': "Mathematica 2.2 impressed me in 1994. by 2000, if it couldn't do integrals, i couldn't either.", 'duration': 25.918, 'max_score': 0.329, 'thumbnail': 'https://coursnap.oss-ap-southeast-1.aliyuncs.com/video-capture/z5WZhCBRDpU/pics/z5WZhCBRDpU329.jpg'}, {'end': 109.665, 'src': 'embed', 'start': 51.278, 'weight': 0, 'content': [{'end': 52.94, 'text': 'A prodigious young man too.', 'start': 51.278, 'duration': 1.662}, {'end': 59.967, 'text': 'Wolfram earned a PhD in theoretical physics from the California Institute of Technology by the age of 20,', 'start': 52.94, 'duration': 7.027}, {'end': 66.073, 'text': 'and he became the youngest recipient of the prestigious MacArthur Fellowship at the age of 21..', 'start': 59.967, 'duration': 6.106}, {'end': 75.442, 'text': "Wolfram's groundbreaking computational tool, Mathematica, was launched in 1988 and has become a cornerstone for researchers and innovators worldwide.", 'start': 66.073, 'duration': 9.369}, {'end': 79.524, 'text': 'and in 2002 he published A New Kind of Science,', 'start': 76.042, 'duration': 3.482}, {'end': 86.909, 'text': 'a paradigm-shifting work which explored the foundations of science through the lens of computational systems.', 'start': 79.524, 'duration': 7.385}, {'end': 94.314, 'text': 'In 2009, Wolfram created Wolfram Alpha, a computational knowledge engine utilized by millions of users worldwide,', 'start': 87.189, 'duration': 7.125}, {'end': 101.699, 'text': 'and his current focus is on the Wolfram language, which is a powerful programming language designed to democratize access to cutting-edge technology.', 'start': 94.314, 'duration': 7.385}, {'end': 107.623, 'text': "Wolfram's numerous accolades include honorary doctorates and fellowships from prestigious institutions.", 'start': 102.359, 'duration': 5.264}, {'end': 109.665, 'text': 'And, as an influential thinker,', 'start': 108.103, 'duration': 1.562}], 'summary': 'Wolfram earned a phd at 20, launched mathematica in 1988, and created wolfram alpha in 2009, becoming an influential thinker.', 'duration': 58.387, 'max_score': 51.278, 'thumbnail': 'https://coursnap.oss-ap-southeast-1.aliyuncs.com/video-capture/z5WZhCBRDpU/pics/z5WZhCBRDpU51278.jpg'}, {'end': 205.384, 'src': 'embed', 'start': 178.029, 'weight': 1, 'content': [{'end': 181.832, 'text': "Why don't we start with the most exciting bit, which is this big announcement.", 'start': 178.029, 'duration': 3.803}, {'end': 183.213, 'text': "Why don't you tell us about that first?", 'start': 181.852, 'duration': 1.361}, {'end': 186.407, 'text': 'Yeah, well so, as of today,', 'start': 184.124, 'duration': 2.283}, {'end': 194.897, 'text': 'ChatGPT can sort of get computational superpowers by calling on Wolfram Alpha and Wolfram Language through its plug-in mechanism.', 'start': 186.407, 'duration': 8.49}, {'end': 197.56, 'text': 'And ChatGPT is..', 'start': 195.557, 'duration': 2.003}, {'end': 205.384, 'text': 'a large language model which has kind of gone through the web and lots of books and so on.', 'start': 199.942, 'duration': 5.442}], 'summary': 'Chatgpt now gains computational superpowers by integrating wolfram alpha and wolfram language.', 'duration': 27.355, 'max_score': 178.029, 'thumbnail': 'https://coursnap.oss-ap-southeast-1.aliyuncs.com/video-capture/z5WZhCBRDpU/pics/z5WZhCBRDpU178029.jpg'}, {'end': 265.358, 'src': 'embed', 'start': 242.588, 'weight': 3, 'content': [{'end': 250.99, 'text': "well Wolfram language we'll talk about that probably a bunch later our sort of computational language to represent the world in a sort of precise computational way.", 'start': 242.588, 'duration': 8.402}, {'end': 257.791, 'text': 'And Wolfram Alpha, the kind of consumer version of that that takes natural language input, converts that to Wolfram language,', 'start': 251.37, 'duration': 6.421}, {'end': 260.755, 'text': 'to precise computational language, and then does computations.', 'start': 257.791, 'duration': 2.964}, {'end': 265.358, 'text': 'And very conveniently, ChatGPT, in a sense, talks natural language too.', 'start': 261.154, 'duration': 4.204}], 'summary': 'Wolfram alpha converts natural language to precise computational language for computations.', 'duration': 22.77, 'max_score': 242.588, 'thumbnail': 'https://coursnap.oss-ap-southeast-1.aliyuncs.com/video-capture/z5WZhCBRDpU/pics/z5WZhCBRDpU242588.jpg'}], 'start': 0.329, 'title': "Dr. stephen wolfram's computational power", 'summary': 'Introduces dr. stephen wolfram, a polymath and influential thinker, who revolutionized the field of science with his computational tools, including mathematica, wolfram alpha, and wolfram language. it discusses the merger of statistical and symbolic approaches in ai, and the integration with chatgpt, providing computational superpowers to chatgpt.', 'chapters': [{'end': 339.661, 'start': 0.329, 'title': 'Dr. stephen wolfram and the power of computation', 'summary': 'Introduces dr. stephen wolfram, a polymath and influential thinker, who revolutionized the field of science with his computational tools, including mathematica, wolfram alpha, and wolfram language, while also announcing an integration with chatgpt, providing computational superpowers to chatgpt. it also discusses the merger of statistical and symbolic approaches in ai.', 'duration': 339.332, 'highlights': ['Dr. Stephen Wolfram, a polymath and influential thinker, revolutionized the field of science with his computational tools, including Mathematica, Wolfram Alpha, and Wolfram Language.', 'Dr. Wolfram announced an integration with ChatGPT, providing computational superpowers to ChatGPT through Wolfram Alpha and Wolfram Language.', "Wolfram's groundbreaking computational tool, Mathematica, launched in 1988, has become a cornerstone for researchers and innovators worldwide.", 'In 2009, Wolfram created Wolfram Alpha, a computational knowledge engine utilized by millions of users worldwide.', "Wolfram's current focus is on the Wolfram language, a powerful programming language designed to democratize access to cutting-edge technology.", 'Dr. Wolfram has dedicated his life to unraveling the mysteries of the universe and making computation accessible to all.', 'Dr. Wolfram became the youngest recipient of the prestigious MacArthur Fellowship at the age of 21.', 'Wolfram earned a PhD in theoretical physics from the California Institute of Technology by the age of 20.']}], 'duration': 339.332, 'thumbnail': 'https://coursnap.oss-ap-southeast-1.aliyuncs.com/video-capture/z5WZhCBRDpU/pics/z5WZhCBRDpU329.jpg', 'highlights': ['Dr. Stephen Wolfram revolutionized science with Mathematica, Wolfram Alpha, and Wolfram Language.', 'Wolfram announced integration with ChatGPT, providing computational superpowers.', 'Mathematica, launched in 1988, is a cornerstone for researchers and innovators worldwide.', 'Wolfram Alpha, a computational knowledge engine, is utilized by millions of users worldwide.', "Wolfram's focus is on the Wolfram language, a powerful programming language.", 'Dr. Wolfram became the youngest recipient of the MacArthur Fellowship at age 21.', 'Wolfram earned a PhD in theoretical physics from Caltech by the age of 20.']}, {'end': 532.972, 'segs': [{'end': 419.74, 'src': 'heatmap', 'start': 339.661, 'weight': 0.891, 'content': [{'end': 347.964, 'text': 'that they might be used as a tool to take a natural language intelligible input and transform it into a representation that can be used on a downstream system.', 'start': 339.661, 'duration': 8.303}, {'end': 348.944, 'text': "It's absolutely amazing.", 'start': 348.004, 'duration': 0.94}, {'end': 350.944, 'text': "But why don't we go on a bit of an intellectual journey?", 'start': 348.964, 'duration': 1.98}, {'end': 355.326, 'text': 'because I was very interested reading your book and I wondered whether you could.', 'start': 350.944, 'duration': 4.382}, {'end': 358.827, 'text': "we'll just start with how the language models work and kind of build from there.", 'start': 355.326, 'duration': 3.501}, {'end': 364.55, 'text': 'Philosophically, and I mean philosophically and deeply because our audience are actually experts on deep learning.', 'start': 359.607, 'duration': 4.943}, {'end': 372.095, 'text': 'I wondered if you could talk about how language models work from your kind of abstract deep understanding, what the reasons are for their success.', 'start': 364.931, 'duration': 7.164}, {'end': 381.921, 'text': 'And you said in the book that computing the probability of even 10 tuples of language would require more data than there are atoms in the universe,', 'start': 372.736, 'duration': 9.185}, {'end': 382.802, 'text': 'which is very interesting.', 'start': 381.921, 'duration': 0.881}, {'end': 385.804, 'text': 'So language models are clearly doing some kind of clever interpolation.', 'start': 382.842, 'duration': 2.962}, {'end': 390.988, 'text': "And, most importantly, the thing that I'm very interested in is what does it mean to understand?", 'start': 386.384, 'duration': 4.604}, {'end': 402.416, 'text': 'Do you think that our language encodes enough representational fidelity to afford a useful semantic mapping to grounded reality or to what most people would understand in the world?', 'start': 391.888, 'duration': 10.528}, {'end': 409.201, 'text': "And I'm sure you're aware of two major strands of thought in the philosophy of meaning, which is to say meaning as reference and meaning as use.", 'start': 402.836, 'duration': 6.365}, {'end': 411.983, 'text': 'So perhaps you could bring that into your answer as well.', 'start': 409.641, 'duration': 2.342}, {'end': 416.217, 'text': 'Okay, so what does it mean to understand something?', 'start': 412.754, 'duration': 3.463}, {'end': 419.74, 'text': 'One of the things when we built WolfMalfur in 2009,.', 'start': 417.138, 'duration': 2.602}], 'summary': "Discussion on language models' ability and understanding, with mention of atoms in the universe and philosophy of meaning.", 'duration': 80.079, 'max_score': 339.661, 'thumbnail': 'https://coursnap.oss-ap-southeast-1.aliyuncs.com/video-capture/z5WZhCBRDpU/pics/z5WZhCBRDpU339661.jpg'}, {'end': 411.983, 'src': 'embed', 'start': 355.326, 'weight': 0, 'content': [{'end': 358.827, 'text': "we'll just start with how the language models work and kind of build from there.", 'start': 355.326, 'duration': 3.501}, {'end': 364.55, 'text': 'Philosophically, and I mean philosophically and deeply because our audience are actually experts on deep learning.', 'start': 359.607, 'duration': 4.943}, {'end': 372.095, 'text': 'I wondered if you could talk about how language models work from your kind of abstract deep understanding, what the reasons are for their success.', 'start': 364.931, 'duration': 7.164}, {'end': 381.921, 'text': 'And you said in the book that computing the probability of even 10 tuples of language would require more data than there are atoms in the universe,', 'start': 372.736, 'duration': 9.185}, {'end': 382.802, 'text': 'which is very interesting.', 'start': 381.921, 'duration': 0.881}, {'end': 385.804, 'text': 'So language models are clearly doing some kind of clever interpolation.', 'start': 382.842, 'duration': 2.962}, {'end': 390.988, 'text': "And, most importantly, the thing that I'm very interested in is what does it mean to understand?", 'start': 386.384, 'duration': 4.604}, {'end': 402.416, 'text': 'Do you think that our language encodes enough representational fidelity to afford a useful semantic mapping to grounded reality or to what most people would understand in the world?', 'start': 391.888, 'duration': 10.528}, {'end': 409.201, 'text': "And I'm sure you're aware of two major strands of thought in the philosophy of meaning, which is to say meaning as reference and meaning as use.", 'start': 402.836, 'duration': 6.365}, {'end': 411.983, 'text': 'So perhaps you could bring that into your answer as well.', 'start': 409.641, 'duration': 2.342}], 'summary': 'Language models require vast data, use clever interpolation, and involve philosophical aspects such as understanding and meaning.', 'duration': 56.657, 'max_score': 355.326, 'thumbnail': 'https://coursnap.oss-ap-southeast-1.aliyuncs.com/video-capture/z5WZhCBRDpU/pics/z5WZhCBRDpU355326.jpg'}, {'end': 456.85, 'src': 'embed', 'start': 433.353, 'weight': 3, 'content': [{'end': 441.06, 'text': "You have this computer, it's doing what it's doing, but What does it mean to understand? Well, we had a very definite definition of understanding.", 'start': 433.353, 'duration': 7.707}, {'end': 448.004, 'text': 'It was take the natural language and convert it to our computational language from which a computation can be done.', 'start': 441.18, 'duration': 6.824}, {'end': 449.946, 'text': 'We had a target for understanding.', 'start': 448.064, 'duration': 1.882}, {'end': 456.85, 'text': 'We had something which was kind of the ultimate representation from which we could do all these things with computation and so on.', 'start': 449.986, 'duration': 6.864}], 'summary': 'Defining understanding as converting natural language to computational language for computation.', 'duration': 23.497, 'max_score': 433.353, 'thumbnail': 'https://coursnap.oss-ap-southeast-1.aliyuncs.com/video-capture/z5WZhCBRDpU/pics/z5WZhCBRDpU433353.jpg'}, {'end': 519.08, 'src': 'embed', 'start': 491.739, 'weight': 4, 'content': [{'end': 499.141, 'text': "so I've been interested for a long time in kind of how do you represent things in the world in a sort of formal computational way?", 'start': 491.739, 'duration': 7.402}, {'end': 503.182, 'text': "That's been the whole story of building our Wolfram language computational language.", 'start': 499.681, 'duration': 3.501}, {'end': 507.108, 'text': "I've also been sort of interested.", 'start': 505.806, 'duration': 1.302}, {'end': 508.85, 'text': "I've been curious for a long time.", 'start': 507.108, 'duration': 1.742}, {'end': 513.554, 'text': 'how do you take kind of everyday discourse and convert it also to something computational?', 'start': 508.85, 'duration': 4.704}, {'end': 519.08, 'text': "In our in Wolfram language, for example, if well, let's say I say I want to eat a piece of chocolate.", 'start': 513.794, 'duration': 5.286}], 'summary': 'Developing a computational language to represent everyday discourse.', 'duration': 27.341, 'max_score': 491.739, 'thumbnail': 'https://coursnap.oss-ap-southeast-1.aliyuncs.com/video-capture/z5WZhCBRDpU/pics/z5WZhCBRDpU491739.jpg'}], 'start': 339.661, 'title': 'Language models and understanding in computational language', 'summary': 'Delves into the deep understanding of language models, their success reasons, and the computational complexity involved, highlighting their ability to map semantic meaning to grounded reality. it also discusses the definition of understanding in computational language, focusing on the conversion of natural language to computational language for computation, and the representation of everyday discourse in a formal computational way.', 'chapters': [{'end': 411.983, 'start': 339.661, 'title': 'Understanding language models', 'summary': 'Explores the deep understanding of language models, their success reasons, and the computational complexity involved, highlighting the philosophical aspect and their ability to map semantic meaning to grounded reality.', 'duration': 72.322, 'highlights': ['Language models compute complex probabilities, with even 10 tuples of language requiring more data than there are atoms in the universe, showcasing their computational complexity and scale.', 'Philosophical discussion on whether language encodes enough representational fidelity to afford a useful semantic mapping to grounded reality, integrating the two major strands of thought in the philosophy of meaning, reference, and use.', 'Exploration of the abstract and deep understanding of how language models work, delving into their reasons for success and the clever interpolation they employ.']}, {'end': 532.972, 'start': 412.754, 'title': 'Defining understanding in computational language', 'summary': 'Discusses the definition of understanding in computational language, focusing on the conversion of natural language to computational language for computation, and the representation of everyday discourse in a formal computational way.', 'duration': 120.218, 'highlights': ['The operational definition of understanding was to convert natural language to computational language for computation. They had a definite definition of understanding that involved converting natural language to a computational language, which served as the operational representation for computation.', 'The focus was on representing everyday discourse in a formal computational way. The interest lay in converting everyday discourse into a formal computational representation, as demonstrated in the example of representing the desire to eat a piece of chocolate in Wolfram language.', 'The goal was to convert everyday discourse to something computational. The ultimate aim was to convert everyday discourse into a computational form, indicating an interest in formalizing everyday language for computational analysis and processing.']}], 'duration': 193.311, 'thumbnail': 'https://coursnap.oss-ap-southeast-1.aliyuncs.com/video-capture/z5WZhCBRDpU/pics/z5WZhCBRDpU339661.jpg', 'highlights': ['Language models compute complex probabilities, with even 10 tuples of language requiring more data than there are atoms in the universe, showcasing their computational complexity and scale.', 'Exploration of the abstract and deep understanding of how language models work, delving into their reasons for success and the clever interpolation they employ.', 'Philosophical discussion on whether language encodes enough representational fidelity to afford a useful semantic mapping to grounded reality, integrating the two major strands of thought in the philosophy of meaning, reference, and use.', 'The operational definition of understanding was to convert natural language to computational language for computation. They had a definite definition of understanding that involved converting natural language to a computational language, which served as the operational representation for computation.', 'The focus was on representing everyday discourse in a formal computational way. The interest lay in converting everyday discourse into a formal computational representation, as demonstrated in the example of representing the desire to eat a piece of chocolate in Wolfram language.', 'The goal was to convert everyday discourse to something computational. The ultimate aim was to convert everyday discourse into a computational form, indicating an interest in formalizing everyday language for computational analysis and processing.']}, {'end': 1328.71, 'segs': [{'end': 563.868, 'src': 'embed', 'start': 533.792, 'weight': 0, 'content': [{'end': 538.614, 'text': "But it is something I've long thought should be possible to represent in a kind of formal computational way.", 'start': 533.792, 'duration': 4.822}, {'end': 542.076, 'text': 'I think what ChatGPT is showing us is that, yes,', 'start': 539.135, 'duration': 2.941}, {'end': 549.079, 'text': 'it really is possible to think about language and the things we represent by language in a quite structured way.', 'start': 542.076, 'duration': 7.003}, {'end': 556.443, 'text': 'So, I mean, the way I like to see it is, ChatGPT has discovered regularities in language.', 'start': 549.539, 'duration': 6.904}, {'end': 563.868, 'text': 'So, first thing, just to comment on this question of when it wants to make a continuation of a piece of text.', 'start': 556.583, 'duration': 7.285}], 'summary': 'Chatgpt demonstrates the possibility of representing language in a structured, computable way, uncovering regularities in language.', 'duration': 30.076, 'max_score': 533.792, 'thumbnail': 'https://coursnap.oss-ap-southeast-1.aliyuncs.com/video-capture/z5WZhCBRDpU/pics/z5WZhCBRDpU533792.jpg'}, {'end': 630.792, 'src': 'embed', 'start': 601.481, 'weight': 4, 'content': [{'end': 607.147, 'text': "So the thing that's really interesting about neural nets is they do this modeling, they do this continuation,", 'start': 601.481, 'duration': 5.666}, {'end': 610.591, 'text': 'they do this extension beyond the merely statistical, so to speak.', 'start': 607.147, 'duration': 3.444}, {'end': 615.339, 'text': 'by in a way that corresponds to what we humans expect.', 'start': 611.195, 'duration': 4.144}, {'end': 624.887, 'text': "perhaps that isn't surprising, because neural nets are architecturally quite similar to brains, and you know, neural nets, as they exist in chat GPT,", 'start': 615.339, 'duration': 9.548}, {'end': 630.792, 'text': 'are remarkably similar to the kind of original idealization invented by McCulloch and Pitts in the 1940s.', 'start': 624.887, 'duration': 5.905}], 'summary': 'Neural nets model beyond statistics, similar to human expectations, resembling original mcculloch and pitts idealization.', 'duration': 29.311, 'max_score': 601.481, 'thumbnail': 'https://coursnap.oss-ap-southeast-1.aliyuncs.com/video-capture/z5WZhCBRDpU/pics/z5WZhCBRDpU601481.jpg'}, {'end': 909.125, 'src': 'embed', 'start': 881.262, 'weight': 1, 'content': [{'end': 887.343, 'text': 'It may just always be kind of a black box to us, but it can add its little bit of natural language magic and complete the circle.', 'start': 881.262, 'duration': 6.081}, {'end': 892.204, 'text': 'We now have almost this massive Turing machine that has these hybrid components to it.', 'start': 887.783, 'duration': 4.421}, {'end': 899.686, 'text': 'You know, is this potentially the first step towards artificial general intelligence, or are there still some missing ingredients??', 'start': 892.664, 'duration': 7.022}, {'end': 909.125, 'text': "Well, I think that the picture you paint, one of the points that I think you're making, is, in a sense, ChatGPT on its own,", 'start': 900.919, 'duration': 8.206}], 'summary': 'Chatgpt has hybrid components, potentially leading to artificial general intelligence.', 'duration': 27.863, 'max_score': 881.262, 'thumbnail': 'https://coursnap.oss-ap-southeast-1.aliyuncs.com/video-capture/z5WZhCBRDpU/pics/z5WZhCBRDpU881262.jpg'}, {'end': 1216.671, 'src': 'embed', 'start': 1192.803, 'weight': 3, 'content': [{'end': 1201.75, 'text': 'But until we get to the point where we have kind of a human narrative that kind of encompasses the kinds of things that are going on there,', 'start': 1192.803, 'duration': 8.947}, {'end': 1208.475, 'text': "until we have a reason to care, it's just out there in the computational universe, not something pulled into kind of the human sphere.", 'start': 1201.75, 'duration': 6.725}, {'end': 1212.748, 'text': 'I would love to explore this concept of creativity a little bit.', 'start': 1209.845, 'duration': 2.903}, {'end': 1216.671, 'text': 'I mean, first of all, reading your book was, it was incredible.', 'start': 1212.768, 'duration': 3.903}], 'summary': 'Discussion on the need for a human narrative and exploring creativity.', 'duration': 23.868, 'max_score': 1192.803, 'thumbnail': 'https://coursnap.oss-ap-southeast-1.aliyuncs.com/video-capture/z5WZhCBRDpU/pics/z5WZhCBRDpU1192803.jpg'}, {'end': 1290.828, 'src': 'embed', 'start': 1264.472, 'weight': 2, 'content': [{'end': 1270.137, 'text': "Because when we're talking about cognitive categories, having it fixed isn't necessarily what we want.", 'start': 1264.472, 'duration': 5.665}, {'end': 1272.539, 'text': 'Right, interesting point.', 'start': 1271.318, 'duration': 1.221}, {'end': 1280.743, 'text': "So you know, one of the things that, in a sense I find you know I'm just sort of coming to terms with it is, in a sense you know,", 'start': 1272.719, 'duration': 8.024}, {'end': 1288.226, 'text': 'ChatGPT and large language models in general are, as you say, taking a snapshot of the world, as we humans have created it,', 'start': 1280.743, 'duration': 7.483}, {'end': 1290.828, 'text': "and they're just saying okay, this is now the standard.", 'start': 1288.226, 'duration': 2.602}], 'summary': 'Cognitive categories should not be fixed; language models like chatgpt capture the world as a standard.', 'duration': 26.356, 'max_score': 1264.472, 'thumbnail': 'https://coursnap.oss-ap-southeast-1.aliyuncs.com/video-capture/z5WZhCBRDpU/pics/z5WZhCBRDpU1264472.jpg'}], 'start': 533.792, 'title': 'Language modeling', 'summary': "Discusses chatgpt's discovery of language regularities, its capabilities in language modeling, computational creativity impact, and challenges in continual learning and semantic grammar establishment within large language models.", 'chapters': [{'end': 577.718, 'start': 533.792, 'title': "Chatgpt's discovery of language regularities", 'summary': "Highlights the possibility of representing language in a formal computational way and chatgpt's discovery of regularities in language, demonstrating its ability to make continuations based on the probability of previous words.", 'duration': 43.926, 'highlights': ['ChatGPT has discovered regularities in language.', 'The possibility of representing language in a formal computational way.', 'Demonstrating its ability to make continuations based on the probability of previous words.']}, {'end': 1328.71, 'start': 578.018, 'title': 'Neural nets and language modeling', 'summary': 'Discusses the capabilities of neural nets, particularly chatgpt, in understanding and modeling language, its potential impact on computational creativity, and the emergence of a separate civilization of ais. it also examines the challenges of continual learning and the establishment of semantic grammar within large language models.', 'duration': 750.692, 'highlights': ['Neural nets extend beyond statistical modeling to correspond to human expectations, resembling the architecture of brains and offering a model that aligns with human understanding. Neural nets, such as ChatGPT, go beyond statistical modeling to emulate human expectations, reflecting the similarity in architecture to brains and providing a model that corresponds to human understanding.', 'ChatGPT has potentially discovered a semantic grammar of language, allowing for the generation of semantically reasonable sentences, essays, and more, which could lead to computational creativity. ChatGPT may have discovered a semantic grammar of language, enabling the creation of semantically reasonable content, potentially fostering computational creativity.', 'The integration of Wolfram Alpha with ChatGPT could lead to the development of a hybrid Turing machine, potentially marking a step toward artificial general intelligence. The combination of Wolfram Alpha and ChatGPT could result in a hybrid Turing machine, representing a potential advancement toward artificial general intelligence.', 'Large language models like ChatGPT capture a static snapshot of the world, establishing a standard for knowledge work based on existing human-created content. Large language models, including ChatGPT, capture a static snapshot of the world, setting a standard for knowledge work using existing human-created content.', 'The challenge lies in determining which computationally creative outputs are relevant and meaningful to humans, highlighting the need for a human narrative to encompass computational creativity. The challenge is to identify computationally creative outputs that are significant to humans, emphasizing the necessity of a human narrative to encompass computational creativity.']}], 'duration': 794.918, 'thumbnail': 'https://coursnap.oss-ap-southeast-1.aliyuncs.com/video-capture/z5WZhCBRDpU/pics/z5WZhCBRDpU533792.jpg', 'highlights': ['ChatGPT has potentially discovered a semantic grammar of language, enabling the creation of semantically reasonable content, potentially fostering computational creativity.', 'The combination of Wolfram Alpha and ChatGPT could result in a hybrid Turing machine, representing a potential advancement toward artificial general intelligence.', 'Large language models, including ChatGPT, capture a static snapshot of the world, setting a standard for knowledge work using existing human-created content.', 'The challenge is to identify computationally creative outputs that are significant to humans, emphasizing the necessity of a human narrative to encompass computational creativity.', 'Neural nets, such as ChatGPT, go beyond statistical modeling to emulate human expectations, reflecting the similarity in architecture to brains and providing a model that corresponds to human understanding.']}, {'end': 1759.141, 'segs': [{'end': 1353.804, 'src': 'embed', 'start': 1329.35, 'weight': 1, 'content': [{'end': 1335.672, 'text': "And the question of sort of how we evolve, how we still keep doing new things, that's an interesting question.", 'start': 1329.35, 'duration': 6.322}, {'end': 1343.874, 'text': 'And I think that, you know, computation is the giant escape valve, so to speak, because it allows us to do huge numbers of new things.', 'start': 1336.232, 'duration': 7.642}, {'end': 1353.804, 'text': 'The issue is, does our understanding kind of follow along with the actual sort of ability of computation to kind of create the new, so to speak.', 'start': 1344.334, 'duration': 9.47}], 'summary': 'Computation enables new possibilities, but does our understanding keep pace?', 'duration': 24.454, 'max_score': 1329.35, 'thumbnail': 'https://coursnap.oss-ap-southeast-1.aliyuncs.com/video-capture/z5WZhCBRDpU/pics/z5WZhCBRDpU1329350.jpg'}, {'end': 1394.795, 'src': 'embed', 'start': 1370.514, 'weight': 0, 'content': [{'end': 1376.4, 'text': "but I'm really happy that they have with thinking about fundamental physics and computational terms,", 'start': 1370.514, 'duration': 5.886}, {'end': 1384.527, 'text': "and that's led to this kind of big idea of this thing we call the ruley ad and which is kind of the entangled limit of all possible computations.", 'start': 1376.4, 'duration': 8.127}, {'end': 1388.99, 'text': 'Imagine you start all possible Turing machines with all possible initial tapes.', 'start': 1385.027, 'duration': 3.963}, {'end': 1394.795, 'text': 'You start them all off together and you sort of simply look at what states those things produce.', 'start': 1389.511, 'duration': 5.284}], 'summary': 'Research on fundamental physics and computational terms has led to the concept of the ruley ad, representing the entangled limit of all possible computations.', 'duration': 24.281, 'max_score': 1370.514, 'thumbnail': 'https://coursnap.oss-ap-southeast-1.aliyuncs.com/video-capture/z5WZhCBRDpU/pics/z5WZhCBRDpU1370514.jpg'}, {'end': 1483.835, 'src': 'heatmap', 'start': 1441.447, 'weight': 0.742, 'content': [{'end': 1449.554, 'text': "is that if you put only a small set of constraints on the way that we observe what's happening in the rouillade, actually really two constraints.", 'start': 1441.447, 'duration': 8.107}, {'end': 1452.336, 'text': 'One is that we are computationally bounded.', 'start': 1450.214, 'duration': 2.122}, {'end': 1454.697, 'text': "We're limited in the computations we can do.", 'start': 1452.736, 'duration': 1.961}, {'end': 1458.881, 'text': 'The other constraint is we believe that we are persistent in time.', 'start': 1455.238, 'duration': 3.643}, {'end': 1466.686, 'text': 'Even though at every moment we get made from different atoms of space, we have the belief and the experience that we persist through time.', 'start': 1459.481, 'duration': 7.205}, {'end': 1477.652, 'text': 'Those two constraints alone are enough to basically tell you that the slice of the Rulliad that we observe will satisfy physics as we know it.', 'start': 1469.028, 'duration': 8.624}, {'end': 1483.835, 'text': 'In fact, specifically the three big theories of 20th century physics, general relativity, the theory of gravity,', 'start': 1477.692, 'duration': 6.143}], 'summary': 'Constraints on observation in rouillade satisfy 20th century physics theories.', 'duration': 42.388, 'max_score': 1441.447, 'thumbnail': 'https://coursnap.oss-ap-southeast-1.aliyuncs.com/video-capture/z5WZhCBRDpU/pics/z5WZhCBRDpU1441447.jpg'}, {'end': 1536.978, 'src': 'embed', 'start': 1511.378, 'weight': 2, 'content': [{'end': 1519.765, 'text': 'But that leads one to this view, that sort of minds experience the Rulliad, experience, this sort of entangled limit of all possible computations.', 'start': 1511.378, 'duration': 8.387}, {'end': 1525.509, 'text': "It's like we are at a particular place in the Rulliad, experiencing sort of the world in this particular way,", 'start': 1520.125, 'duration': 5.384}, {'end': 1531.074, 'text': "just as we're at a particular place in physical space, experiencing the world with respect to that place in physical space.", 'start': 1525.509, 'duration': 5.565}, {'end': 1536.978, 'text': 'And so you think about sort of different minds as being at different points in the Rulliad.', 'start': 1531.574, 'duration': 5.404}], 'summary': 'Different minds experience the rulliad at different points.', 'duration': 25.6, 'max_score': 1511.378, 'thumbnail': 'https://coursnap.oss-ap-southeast-1.aliyuncs.com/video-capture/z5WZhCBRDpU/pics/z5WZhCBRDpU1511378.jpg'}, {'end': 1594.498, 'src': 'embed', 'start': 1567.004, 'weight': 3, 'content': [{'end': 1575.811, 'text': "and we're gradually expanding in real space, able to encompass more kinds of ways of looking at this ultimate computational structure.", 'start': 1567.004, 'duration': 8.807}, {'end': 1583.214, 'text': 'So I think that this kind of the thing that we see as we you know,', 'start': 1576.131, 'duration': 7.083}, {'end': 1590.597, 'text': 'the sort of the growth of our kind of well sort of the intellectual history of our civilization,', 'start': 1583.214, 'duration': 7.383}, {'end': 1594.498, 'text': 'can be thought of as this kind of gradual colonization of rural space.', 'start': 1590.597, 'duration': 3.901}], 'summary': "Our civilization's intellectual history involves gradual expansion in real space and computational structure.", 'duration': 27.494, 'max_score': 1567.004, 'thumbnail': 'https://coursnap.oss-ap-southeast-1.aliyuncs.com/video-capture/z5WZhCBRDpU/pics/z5WZhCBRDpU1567004.jpg'}], 'start': 1329.35, 'title': "Computation's role in evolving understanding", 'summary': "Explores the role of computation in driving new possibilities and the emergence of 'ruley ad', with significant breakthroughs. it also discusses how rulliad serves as the computational structure underlying physics and mathematics, leading to the observation of physics and the expansion of paradigms in rullial space.", 'chapters': [{'end': 1404.822, 'start': 1329.35, 'title': "Computation's role in evolving understanding", 'summary': "Discusses the role of computation in driving new possibilities and the emergence of the 'ruley ad', a concept involving the entangled limit of all possible computations, which has seen significant breakthroughs in recent years.", 'duration': 75.472, 'highlights': ["The 'ruley ad' concept, involving the entangled limit of all possible computations, has seen significant breakthroughs in recent years, contributing to our understanding of fundamental physics and computational terms.", 'Computation serves as a giant escape valve, enabling the exploration of a vast number of new possibilities.', 'The chapter raises thought-provoking questions about whether our understanding can keep pace with the capabilities of computation in creating the new.']}, {'end': 1759.141, 'start': 1404.822, 'title': 'Computational structure underlying physics and mathematics', 'summary': 'Discusses the concept of rulliad as the computational structure underlying physics and mathematics, with constraints on computation and persistence in time resulting in the observation of physics as we know it, and the expansion of paradigms as the gradual colonization of rullial space.', 'duration': 354.319, 'highlights': ['The concept of Rulliad as the computational structure underlying physics and mathematics, with constraints on computation and persistence in time resulting in the observation of physics as we know it.', "The expansion of paradigms as the gradual colonization of Rullial space, with the growth of civilization's intellectual history being thought of as such.", 'The idea of expanding in real space to encompass more ways of looking at the ultimate computational structure, leading to the growth of intellectual history.', 'The exploration of the concept of nested patterns as an example of something that could be thought about in the world but is not being considered, highlighting the potential for undiscovered ideas and concepts.']}], 'duration': 429.791, 'thumbnail': 'https://coursnap.oss-ap-southeast-1.aliyuncs.com/video-capture/z5WZhCBRDpU/pics/z5WZhCBRDpU1329350.jpg', 'highlights': ["The 'ruley ad' concept has seen significant breakthroughs, contributing to our understanding of fundamental physics and computational terms.", 'Computation serves as a giant escape valve, enabling the exploration of a vast number of new possibilities.', 'The concept of Rulliad as the computational structure underlying physics and mathematics, with constraints on computation and persistence in time resulting in the observation of physics as we know it.', "The expansion of paradigms as the gradual colonization of Rullial space, with the growth of civilization's intellectual history being thought of as such."]}, {'end': 2107.802, 'segs': [{'end': 1816.417, 'src': 'embed', 'start': 1759.141, 'weight': 0, 'content': [{'end': 1762.346, 'text': "in recent times, everybody who would see this pattern say oh, that's a fractal pattern.", 'start': 1759.141, 'duration': 3.205}, {'end': 1769.575, 'text': 'Um, but uh, you know, this is the kind of example of how sort of, when you have a new paradigm for thinking about things,', 'start': 1762.866, 'duration': 6.709}, {'end': 1772.178, 'text': "you kind of notice things you didn't notice before.", 'start': 1769.997, 'duration': 2.181}, {'end': 1776.899, 'text': "And a very interesting question is what things are there in the world that we haven't noticed before?", 'start': 1772.558, 'duration': 4.341}, {'end': 1780.119, 'text': 'In my own life, sort of the biggest one,', 'start': 1776.979, 'duration': 3.14}, {'end': 1786.661, 'text': 'that was this phenomenon that I already mentioned about how simple programs can lead to extremely complicated behavior.', 'start': 1780.119, 'duration': 6.542}, {'end': 1790.442, 'text': "That's a very fundamental scientific result that has all kinds of implications.", 'start': 1787.021, 'duration': 3.421}, {'end': 1797.466, 'text': "But that's a result which, as I say, after I had first done computer experiments that plainly showed that result.", 'start': 1790.862, 'duration': 6.604}, {'end': 1801.849, 'text': "it still took me several years to kind of realize oh, that's a real thing.", 'start': 1797.466, 'duration': 4.383}, {'end': 1810.255, 'text': 'And when I look back, I realize there were zillions of examples of this that existed for hundreds of years where people just ignored it.', 'start': 1802.25, 'duration': 8.005}, {'end': 1811.516, 'text': 'I mean, you know, you look at..', 'start': 1810.695, 'duration': 0.821}, {'end': 1816.417, 'text': "Yeah, I was wondering if I could jump in here, because what you're saying is, it's truly fascinating.", 'start': 1812.316, 'duration': 4.101}], 'summary': 'New paradigm leads to noticing unnoticed patterns; simple programs create complex behavior.', 'duration': 57.276, 'max_score': 1759.141, 'thumbnail': 'https://coursnap.oss-ap-southeast-1.aliyuncs.com/video-capture/z5WZhCBRDpU/pics/z5WZhCBRDpU1759141.jpg'}, {'end': 1920.148, 'src': 'embed', 'start': 1897.679, 'weight': 1, 'content': [{'end': 1905.722, 'text': 'Well so, first point to make is this notion of computational irreducibility, which I think is really a pretty core notion.', 'start': 1897.679, 'duration': 8.043}, {'end': 1911.284, 'text': "that sort of we're starting to see kind of pop up in a lot of different places.", 'start': 1905.722, 'duration': 5.562}, {'end': 1913.905, 'text': "It's an important piece of kind of intuition about science.", 'start': 1911.324, 'duration': 2.581}, {'end': 1918.587, 'text': 'The question is this, if you have a program You have a set of rules.', 'start': 1914.465, 'duration': 4.122}, {'end': 1920.148, 'text': 'You know the set of rules.', 'start': 1919.107, 'duration': 1.041}], 'summary': 'Computational irreducibility is a core notion in science, popping up in many places.', 'duration': 22.469, 'max_score': 1897.679, 'thumbnail': 'https://coursnap.oss-ap-southeast-1.aliyuncs.com/video-capture/z5WZhCBRDpU/pics/z5WZhCBRDpU1897679.jpg'}, {'end': 1970.929, 'src': 'embed', 'start': 1941.642, 'weight': 5, 'content': [{'end': 1943.643, 'text': "this is what's going to happen sometime in the future.", 'start': 1941.642, 'duration': 2.001}, {'end': 1946.946, 'text': "It's just once you know the rules, you can predict the future.", 'start': 1944.164, 'duration': 2.782}, {'end': 1953.573, 'text': "That's been the kind of paradigmatic idea that people have about science for the last 300 years or so.", 'start': 1947.566, 'duration': 6.007}, {'end': 1961.982, 'text': 'What we realize from thinking about computational universe, computational thinking about science based on programs rather than on equations,', 'start': 1954.113, 'duration': 7.869}, {'end': 1963.363, 'text': "is that isn't always true.", 'start': 1961.982, 'duration': 1.381}, {'end': 1964.765, 'text': "In fact, it's often not true.", 'start': 1963.543, 'duration': 1.222}, {'end': 1970.929, 'text': "Instead, what can happen is you've got the rules, You run the rules, you can explicitly run them, you run them for a billion steps.", 'start': 1965.265, 'duration': 5.664}], 'summary': 'Computational thinking challenges predictability in science based on programs, not equations.', 'duration': 29.287, 'max_score': 1941.642, 'thumbnail': 'https://coursnap.oss-ap-southeast-1.aliyuncs.com/video-capture/z5WZhCBRDpU/pics/z5WZhCBRDpU1941642.jpg'}, {'end': 2021.873, 'src': 'embed', 'start': 1995.539, 'weight': 3, 'content': [{'end': 2003.103, 'text': 'The principle of computational equivalence says that as soon as you see a system which is not obviously simple in its behavior,', 'start': 1995.539, 'duration': 7.564}, {'end': 2007.75, 'text': 'it will tend to be equivalent in the sophistication of computations it can do so.', 'start': 2003.103, 'duration': 4.647}, {'end': 2012.92, 'text': "you know, we've known since the 1920s, 1930s that there are universal computers.", 'start': 2007.75, 'duration': 5.17}, {'end': 2021.873, 'text': 'there are systems that are capable of doing that can be programmed to do any computation.', 'start': 2012.92, 'duration': 8.953}], 'summary': 'Principle of computational equivalence states non-simple systems are capable of universal computation since the 1920s, 1930s.', 'duration': 26.334, 'max_score': 1995.539, 'thumbnail': 'https://coursnap.oss-ap-southeast-1.aliyuncs.com/video-capture/z5WZhCBRDpU/pics/z5WZhCBRDpU1995539.jpg'}], 'start': 1759.141, 'title': 'New paradigms and unnoticed phenomena', 'summary': 'Discusses the emergence of new paradigms in thinking, simple programs leading to complex behavior, and the importance of recognizing previously unnoticed occurrences. it also explores computational irreducibility, the challenges of predicting complex behaviors, and the principle of computational equivalence, emphasizing the limitations and the importance of understanding complex systems step by step.', 'chapters': [{'end': 1816.417, 'start': 1759.141, 'title': 'New paradigms and unnoticed phenomena', 'summary': 'Discusses the emergence of new paradigms in thinking and the phenomenon of simple programs leading to complex behavior, highlighting the importance of recognizing previously unnoticed occurrences.', 'duration': 57.276, 'highlights': ['The phenomenon of simple programs leading to extremely complicated behavior is a fundamental scientific result with wide-ranging implications, exemplifying the significance of recognizing previously unnoticed occurrences.', 'The emergence of new paradigms in thinking allows for the discovery of previously unnoticed phenomena, prompting the consideration of what other things in the world have gone unnoticed.', 'The delay in realizing the significance of certain scientific results, despite clear experimental evidence, underscores the prevalence of unnoticed occurrences that have existed for centuries.']}, {'end': 2107.802, 'start': 1816.457, 'title': 'Navigating computational universe', 'summary': 'Discusses the concept of computational irreducibility, the challenges of predicting complex behaviors in computational universe, and the principle of computational equivalence, emphasizing the limitations and the importance of following through each step in understanding complex systems.', 'duration': 291.345, 'highlights': ['The concept of computational irreducibility and its implications in predicting complex behaviors in computational universe. It explains the limitations in predicting complex behaviors and the importance of following through each step in understanding complex systems.', 'The principle of computational equivalence and its ubiquity in systems exhibiting non-obviously simple behavior. It highlights the ubiquitous nature of computational equivalence in systems and its implications for understanding computational sophistication.', 'The challenges of predicting future outcomes in computational universe due to computational irreducibility. It discusses the challenges in predicting future outcomes and the limitations posed by computational irreducibility.']}], 'duration': 348.661, 'thumbnail': 'https://coursnap.oss-ap-southeast-1.aliyuncs.com/video-capture/z5WZhCBRDpU/pics/z5WZhCBRDpU1759141.jpg', 'highlights': ['The phenomenon of simple programs leading to extremely complicated behavior is a fundamental scientific result with wide-ranging implications, exemplifying the significance of recognizing previously unnoticed occurrences.', 'The concept of computational irreducibility and its implications in predicting complex behaviors in computational universe. It explains the limitations in predicting complex behaviors and the importance of following through each step in understanding complex systems.', 'The emergence of new paradigms in thinking allows for the discovery of previously unnoticed phenomena, prompting the consideration of what other things in the world have gone unnoticed.', 'The principle of computational equivalence and its ubiquity in systems exhibiting non-obviously simple behavior. It highlights the ubiquitous nature of computational equivalence in systems and its implications for understanding computational sophistication.', 'The delay in realizing the significance of certain scientific results, despite clear experimental evidence, underscores the prevalence of unnoticed occurrences that have existed for centuries.', 'The challenges of predicting future outcomes in computational universe due to computational irreducibility. It discusses the challenges in predicting future outcomes and the limitations posed by computational irreducibility.']}, {'end': 2652.97, 'segs': [{'end': 2189.759, 'src': 'embed', 'start': 2163.137, 'weight': 0, 'content': [{'end': 2170.282, 'text': 'Well, essentially, computational irreducibility is the enemy of neural net training, because as soon as there is computational irreducibility,', 'start': 2163.137, 'duration': 7.145}, {'end': 2177.747, 'text': "as soon as there's actually something you have to explicitly compute, you just can't do this approach of saying well,", 'start': 2170.282, 'duration': 7.465}, {'end': 2179.509, 'text': 'training is about sort of knowing what will happen.', 'start': 2177.747, 'duration': 1.762}, {'end': 2184.313, 'text': 'you know, being able to sort of explicitly burn in a prediction of what will happen.', 'start': 2180.59, 'duration': 3.723}, {'end': 2189.759, 'text': "And that is explicitly kind of, that's sort of the opposite direction from computational irreducibility.", 'start': 2184.674, 'duration': 5.085}], 'summary': 'Computational irreducibility hinders neural net training by requiring explicit computation, contrary to the approach of predicting outcomes.', 'duration': 26.622, 'max_score': 2163.137, 'thumbnail': 'https://coursnap.oss-ap-southeast-1.aliyuncs.com/video-capture/z5WZhCBRDpU/pics/z5WZhCBRDpU2163137.jpg'}, {'end': 2235.029, 'src': 'embed', 'start': 2206.966, 'weight': 1, 'content': [{'end': 2214.03, 'text': 'I mean, I kind of view sort of what ChatGPT is doing and a lot of its potential uses as being kind of a new form of user interface.', 'start': 2206.966, 'duration': 7.064}, {'end': 2219.919, 'text': "It's kind of a linguistic user interface in the way that we've had In the past, we had graphical user interfaces, et cetera.", 'start': 2214.07, 'duration': 5.849}, {'end': 2226.583, 'text': 'I mean, in a sense, Wolfram Alpha was, I think, the first serious example of a linguistic user interface on the input side.', 'start': 2219.939, 'duration': 6.644}, {'end': 2227.924, 'text': 'We were not dealing with the output side.', 'start': 2226.603, 'duration': 1.321}, {'end': 2228.845, 'text': "We're not writing essays.", 'start': 2227.944, 'duration': 0.901}, {'end': 2235.029, 'text': "We're just generating essentially more like a GUI kind of output linguistic input, so to speak.", 'start': 2228.905, 'duration': 6.124}], 'summary': "Chatgpt serves as a linguistic user interface, marking a new form of ui with potential uses, similar to wolfram alpha's language-based input.", 'duration': 28.063, 'max_score': 2206.966, 'thumbnail': 'https://coursnap.oss-ap-southeast-1.aliyuncs.com/video-capture/z5WZhCBRDpU/pics/z5WZhCBRDpU2206966.jpg'}, {'end': 2297.605, 'src': 'embed', 'start': 2268.76, 'weight': 2, 'content': [{'end': 2272.943, 'text': 'because they just were trained on the things we humans produce.', 'start': 2268.76, 'duration': 4.183}, {'end': 2280.907, 'text': 'They, in a sense, have a very finely tuned sense of sort of what the current state of how we humans think about things is.', 'start': 2273.003, 'duration': 7.904}, {'end': 2285.731, 'text': 'And so if you say, well, what did we miss? that becomes a thing that it can potentially answer.', 'start': 2281.388, 'duration': 4.343}, {'end': 2293.52, 'text': "You know, the question of and this is something I've been curious about and I did some experiments, but they haven't been very successful yet.", 'start': 2286.192, 'duration': 7.328}, {'end': 2297.605, 'text': "It's kind of of all possible, let's say mathematical theorems.", 'start': 2294.081, 'duration': 3.524}], 'summary': 'Ai models trained on human data can answer questions about the current state of human thinking, but may struggle with certain complex problems.', 'duration': 28.845, 'max_score': 2268.76, 'thumbnail': 'https://coursnap.oss-ap-southeast-1.aliyuncs.com/video-capture/z5WZhCBRDpU/pics/z5WZhCBRDpU2268760.jpg'}, {'end': 2342.798, 'src': 'embed', 'start': 2305.094, 'weight': 3, 'content': [{'end': 2312.503, 'text': 'but to be able to say, OK, large language model, you just learned all these things from what humans have cared about and put on the web.', 'start': 2305.094, 'duration': 7.409}, {'end': 2320.312, 'text': "Now tell us which things exist in the computational universe, but we haven't yet given names to and so on.", 'start': 2313.003, 'duration': 7.309}, {'end': 2321.694, 'text': 'Which ones should we give names to??', 'start': 2320.332, 'duration': 1.362}, {'end': 2331.368, 'text': 'Yes, I wholeheartedly agree that the gradient of interestingness is very important, but I also want to bring in the gradient of conceivability.', 'start': 2322.677, 'duration': 8.691}, {'end': 2337.195, 'text': "Just before I do that, I wanted to comment on you referring to it as a linguistic interface, and I'm glad that you did,", 'start': 2331.748, 'duration': 5.447}, {'end': 2342.798, 'text': "because It's an interface that can be used by many people in many different situations.", 'start': 2337.195, 'duration': 5.603}], 'summary': 'Using large language models to identify unnamed concepts in the computational universe, considering interestingness and conceivability, as a versatile linguistic interface.', 'duration': 37.704, 'max_score': 2305.094, 'thumbnail': 'https://coursnap.oss-ap-southeast-1.aliyuncs.com/video-capture/z5WZhCBRDpU/pics/z5WZhCBRDpU2305094.jpg'}, {'end': 2407.298, 'src': 'embed', 'start': 2378.475, 'weight': 5, 'content': [{'end': 2384.839, 'text': "And it's almost like we're cheating by creating these computational models which can, subject to irreducibility,", 'start': 2378.475, 'duration': 6.364}, {'end': 2386.76, 'text': 'can compute all of these different things.', 'start': 2384.839, 'duration': 1.921}, {'end': 2394.846, 'text': 'But the key question is whether or not we would recognize something as being interesting or whether it would be conceivable even if it were present.', 'start': 2386.88, 'duration': 7.966}, {'end': 2397.802, 'text': 'Right, so I think this is the thing.', 'start': 2395.979, 'duration': 1.823}, {'end': 2401.348, 'text': 'Progress is gradual and slow.', 'start': 2398.824, 'duration': 2.524}, {'end': 2407.298, 'text': 'In other words, if you jump to some random place in the Rulliad, you pick a random program, you look at it.', 'start': 2401.989, 'duration': 5.309}], 'summary': 'Computational models can compute various things, progress is gradual and slow.', 'duration': 28.823, 'max_score': 2378.475, 'thumbnail': 'https://coursnap.oss-ap-southeast-1.aliyuncs.com/video-capture/z5WZhCBRDpU/pics/z5WZhCBRDpU2378475.jpg'}, {'end': 2569.275, 'src': 'embed', 'start': 2539.905, 'weight': 6, 'content': [{'end': 2541.546, 'text': "Well, you're trying to achieve certain goals.", 'start': 2539.905, 'duration': 1.641}, {'end': 2551.653, 'text': "You're trying to and this is where you know my sort of lifelong story of kind of the programming languages versus the full-scale computational language idea.", 'start': 2541.987, 'duration': 9.666}, {'end': 2558.638, 'text': 'you know the programming languages which are really talking to computers in terms that computers intrinsically understand.', 'start': 2551.653, 'duration': 6.985}, {'end': 2559.999, 'text': 'you know, set this.', 'start': 2558.638, 'duration': 1.361}, {'end': 2569.275, 'text': 'set up this array and move this, do this loop, change this value from this to this to this.', 'start': 2561.57, 'duration': 7.705}], 'summary': 'The goal is to achieve certain computational language ideas through programming languages.', 'duration': 29.37, 'max_score': 2539.905, 'thumbnail': 'https://coursnap.oss-ap-southeast-1.aliyuncs.com/video-capture/z5WZhCBRDpU/pics/z5WZhCBRDpU2539905.jpg'}, {'end': 2656.911, 'src': 'embed', 'start': 2631.096, 'weight': 7, 'content': [{'end': 2639.2, 'text': 'Because you made a point in that article from February that Right now, to be honest, the construction of these architectures is an art.', 'start': 2631.096, 'duration': 8.104}, {'end': 2641.782, 'text': "It's really just people do some trial and error.", 'start': 2639.38, 'duration': 2.402}, {'end': 2647.326, 'text': 'They come up with some horseshoes or whatever shape of architecture happens to work.', 'start': 2641.822, 'duration': 5.504}, {'end': 2649.988, 'text': 'GPT is in that same kind of category.', 'start': 2647.526, 'duration': 2.462}, {'end': 2652.97, 'text': "You point out there's no real science behind that.", 'start': 2650.368, 'duration': 2.602}, {'end': 2656.911, 'text': 'And on the one hand, we could defer all that science to AIs,', 'start': 2653.47, 'duration': 3.441}], 'summary': 'Constructing architectures is an art with no science; ais can handle it.', 'duration': 25.815, 'max_score': 2631.096, 'thumbnail': 'https://coursnap.oss-ap-southeast-1.aliyuncs.com/video-capture/z5WZhCBRDpU/pics/z5WZhCBRDpU2631096.jpg'}], 'start': 2108.182, 'title': 'Computational modeling and cognitive bridge', 'summary': 'Explores computational irreducibility, the impact on neural net training, and the potential of language models in humanizing communication. it also delves into the importance of conceivability in computational modeling and the need for a cognitive bridge between human thinking and computational possibilities.', 'chapters': [{'end': 2321.694, 'start': 2108.182, 'title': 'Computational irreducibility and neural nets', 'summary': 'Discusses computational irreducibility, its impact on neural net training, and the potential of language models like chatgpt in humanizing communication with the computational world, as well as the use of neural nets to identify interesting human concepts. it also touches on the potential of large language models in identifying mathematical theorems considered interesting by humans.', 'duration': 213.512, 'highlights': ['The chapter discusses computational irreducibility and its impact on neural net training, highlighting that computational irreducibility hinders neural net training as it requires explicit computation, which goes against the approach of predicting outcomes. Computational irreducibility hinders neural net training as it requires explicit computation, which goes against the approach of predicting outcomes.', 'It mentions the potential of language models like ChatGPT in humanizing communication with the computational world, serving as a new form of linguistic user interface, and enhancing communication with the computational world. ChatGPT serves as a new form of linguistic user interface, enhancing communication with the computational world.', 'The chapter explores the use of neural nets to identify interesting human concepts, suggesting that they are trained on the things humans produce and may have a finely tuned sense of how humans think about things. Neural nets are trained on the things humans produce and may have a finely tuned sense of how humans think about things.', 'It briefly touches on the potential of large language models to identify mathematical theorems considered interesting by humans, and the possibility of using them to find unnamed concepts in the computational universe. Large language models may be used to identify mathematical theorems considered interesting by humans and to find unnamed concepts in the computational universe.']}, {'end': 2652.97, 'start': 2322.677, 'title': 'Importance of conceivability in computational modeling', 'summary': 'Discusses the importance of conceivability and intelligibility in computational modeling, emphasizing the gradual progress and the need for a cognitive bridge between human thinking and computational possibilities.', 'duration': 330.293, 'highlights': ['The importance of conceivability and intelligibility in computational modeling, emphasizing the need for a cognitive bridge between human thinking and computational possibilities, is discussed.', 'The gradual progress and the necessity for an intellectual path to recognize intriguing and conceivable concepts in computational models are emphasized.', 'The challenges of transitioning from traditional programming languages to a computational language that serves as a bridge between human thinking and computational possibilities are explored.', 'The current state of architecture construction in computational models is described as primarily artistic and reliant on trial and error, lacking a scientific foundation.']}], 'duration': 544.788, 'thumbnail': 'https://coursnap.oss-ap-southeast-1.aliyuncs.com/video-capture/z5WZhCBRDpU/pics/z5WZhCBRDpU2108182.jpg', 'highlights': ['Computational irreducibility hinders neural net training as it requires explicit computation, which goes against the approach of predicting outcomes.', 'ChatGPT serves as a new form of linguistic user interface, enhancing communication with the computational world.', 'Neural nets are trained on the things humans produce and may have a finely tuned sense of how humans think about things.', 'Large language models may be used to identify mathematical theorems considered interesting by humans and to find unnamed concepts in the computational universe.', 'The importance of conceivability and intelligibility in computational modeling, emphasizing the need for a cognitive bridge between human thinking and computational possibilities, is discussed.', 'The gradual progress and the necessity for an intellectual path to recognize intriguing and conceivable concepts in computational models are emphasized.', 'The challenges of transitioning from traditional programming languages to a computational language that serves as a bridge between human thinking and computational possibilities are explored.', 'The current state of architecture construction in computational models is described as primarily artistic and reliant on trial and error, lacking a scientific foundation.']}, {'end': 3449.589, 'segs': [{'end': 2680.402, 'src': 'embed', 'start': 2653.47, 'weight': 0, 'content': [{'end': 2656.911, 'text': 'And on the one hand, we could defer all that science to AIs,', 'start': 2653.47, 'duration': 3.441}, {'end': 2662.654, 'text': "potentially and maybe we'll never understand the science that they do to build the next better versions of themselves.", 'start': 2656.911, 'duration': 5.743}, {'end': 2671.297, 'text': "But I'm wondering if you also have any space in your vision for us developing human intelligible sciences that can do a better job of this.", 'start': 2663.114, 'duration': 8.183}, {'end': 2677.46, 'text': "And if so, like what might they look like and what other sciences might they resemble? Zoology? You know, I don't know.", 'start': 2671.377, 'duration': 6.083}, {'end': 2680.402, 'text': 'Right. so I think the thing okay.', 'start': 2678.36, 'duration': 2.042}], 'summary': "Exploring ai's potential to advance science and the need for human-intelligible sciences.", 'duration': 26.932, 'max_score': 2653.47, 'thumbnail': 'https://coursnap.oss-ap-southeast-1.aliyuncs.com/video-capture/z5WZhCBRDpU/pics/z5WZhCBRDpU2653470.jpg'}, {'end': 2750.309, 'src': 'embed', 'start': 2711.248, 'weight': 1, 'content': [{'end': 2716.971, 'text': 'We know something about how the microscopic brain works and neuron firings and so on.', 'start': 2711.248, 'duration': 5.723}, {'end': 2723.535, 'text': "We don't have an intermediate language of neuroscience that is kind of the thing that is above the level of individual neuron firings.", 'start': 2717.231, 'duration': 6.304}, {'end': 2725.916, 'text': 'and below the level of psychology, basically.', 'start': 2723.915, 'duration': 2.001}, {'end': 2731.979, 'text': "And I think that that's the question of can you find that and when can you find that is a very interesting question.", 'start': 2726.476, 'duration': 5.503}, {'end': 2737.982, 'text': 'I think one of the things that is a science point that when you have computational irreducibility,', 'start': 2732.5, 'duration': 5.482}, {'end': 2744.266, 'text': 'one of the necessary consequences of computational irreducibility is, within a computationally irreducible system,', 'start': 2737.982, 'duration': 6.284}, {'end': 2750.309, 'text': 'there will always be an infinite number of specific little pieces of computational reusability that you can find.', 'start': 2744.266, 'duration': 6.043}], 'summary': "Understanding the brain's intermediate language in neuroscience is a challenging question with implications for computational irreducibility.", 'duration': 39.061, 'max_score': 2711.248, 'thumbnail': 'https://coursnap.oss-ap-southeast-1.aliyuncs.com/video-capture/z5WZhCBRDpU/pics/z5WZhCBRDpU2711248.jpg'}, {'end': 3121.082, 'src': 'embed', 'start': 3088.785, 'weight': 4, 'content': [{'end': 3093.751, 'text': "And that's something that is really a matter of natural science, of neuroscience and so on to determine.", 'start': 3088.785, 'duration': 4.966}, {'end': 3096.448, 'text': 'But there happens to be an alignment there,', 'start': 3094.127, 'duration': 2.321}, {'end': 3103.913, 'text': 'possibly because the underlying sort of McCulloch-Pitts-level architecture is quite similar between our brains and what happens in neural nets.', 'start': 3096.448, 'duration': 7.465}, {'end': 3112.457, 'text': "And I think that's something that has been fortunate.", 'start': 3104.613, 'duration': 7.844}, {'end': 3121.082, 'text': 'I mean, for me, I first looked at neural nets back in 1981 or 1980.', 'start': 3112.597, 'duration': 8.485}], 'summary': 'Neural nets and brain architecture align, possibly due to similar mcculloch-pitts-level structures, as observed since 1980.', 'duration': 32.297, 'max_score': 3088.785, 'thumbnail': 'https://coursnap.oss-ap-southeast-1.aliyuncs.com/video-capture/z5WZhCBRDpU/pics/z5WZhCBRDpU3088785.jpg'}, {'end': 3202.721, 'src': 'embed', 'start': 3176.171, 'weight': 2, 'content': [{'end': 3181.792, 'text': "Yeah Well, once you can do that, once you've got actual computation in the loop.", 'start': 3176.171, 'duration': 5.621}, {'end': 3192.015, 'text': "The thing to emphasize, perhaps, about what I think we're adding to the world of large language models with Wolfram Language and Wolfram Alpha,", 'start': 3182.692, 'duration': 9.323}, {'end': 3198.299, 'text': "is it's both the depth of computation, the kind of possibility of irreducible computation,", 'start': 3192.015, 'duration': 6.284}, {'end': 3202.721, 'text': 'and the structured knowledge of the world that we actually do have.', 'start': 3198.299, 'duration': 4.422}], 'summary': 'Wolfram language and wolfram alpha add depth of computation and structured knowledge to large language models.', 'duration': 26.55, 'max_score': 3176.171, 'thumbnail': 'https://coursnap.oss-ap-southeast-1.aliyuncs.com/video-capture/z5WZhCBRDpU/pics/z5WZhCBRDpU3176171.jpg'}], 'start': 2653.47, 'title': 'Ai role in science and computational irreducibility', 'summary': 'Discusses the potential role of ais in advancing science and the need for developing human-intelligible sciences. it also explores computational irreducibility and its implications on predictability and inventions, with potential applications in machine learning and neural networks.', 'chapters': [{'end': 2731.979, 'start': 2653.47, 'title': 'Developing human-intelligible sciences', 'summary': 'Discusses the potential role of ais in advancing science and explores the need for developing human-intelligible sciences to understand complex computational processes, such as neuroscience, in between individual neuron firings and psychology.', 'duration': 78.509, 'highlights': ['Developing human-intelligible sciences to understand complex computational processes, such as neuroscience, in between individual neuron firings and psychology.', 'Questioning the possibility of developing sciences resembling zoology to understand complex computational processes.', 'Exploring the potential for AIs to advance science and build better versions of themselves.', 'The challenge of finding an intermediate language of neuroscience that transcends individual neuron firings and psychology.']}, {'end': 3449.589, 'start': 2732.5, 'title': 'Computational irreducibility and predictive behavior', 'summary': 'Explores the concept of computational irreducibility, highlighting the infinite number of specific pieces of computational reusability within a computationally irreducible system, and its implications on inventions and predictability, with a potential application in machine learning and neural networks.', 'duration': 717.089, 'highlights': ['The concept of computational irreducibility leads to an infinite collection of specific pieces of computational reusability within a system, despite the overall computational irreducibility, potentially impacting inventions and predictability.', 'The potential application of ideas from physics, such as quantum mechanics, in machine learning, by considering all possible paths and utilizing concepts like continuum space and event horizons in training space, may lead to valuable insights, despite current practical irrelevance.', 'The alignment between neural nets and human generalization, possibly due to the similarities in underlying architecture, presents an area of fortunate convergence and a potential avenue for further exploration in neuroscience and machine learning.', 'The addition of Wolfram Language and Wolfram Alpha to large language models emphasizes the depth of computation, irreducible computation, and structured knowledge, offering a representation of real-world entities and a departure from mere statistical occurrences on the web.', 'The exploration of the computational phase of matter in biological tissue, characterized by meaningful structure in the transport of molecules, suggests a potential link between biological processes and computational understanding, although the precise integration of these aspects remains unknown.']}], 'duration': 796.119, 'thumbnail': 'https://coursnap.oss-ap-southeast-1.aliyuncs.com/video-capture/z5WZhCBRDpU/pics/z5WZhCBRDpU2653470.jpg', 'highlights': ['Exploring the potential for AIs to advance science and build better versions of themselves.', 'The challenge of finding an intermediate language of neuroscience that transcends individual neuron firings and psychology.', 'The addition of Wolfram Language and Wolfram Alpha to large language models emphasizes the depth of computation, irreducible computation, and structured knowledge, offering a representation of real-world entities and a departure from mere statistical occurrences on the web.', 'The concept of computational irreducibility leads to an infinite collection of specific pieces of computational reusability within a system, despite the overall computational irreducibility, potentially impacting inventions and predictability.', 'The alignment between neural nets and human generalization, possibly due to the similarities in underlying architecture, presents an area of fortunate convergence and a potential avenue for further exploration in neuroscience and machine learning.']}], 'highlights': ['Dr. Stephen Wolfram revolutionized science with Mathematica, Wolfram Alpha, and Wolfram Language.', 'Wolfram announced integration with ChatGPT, providing computational superpowers.', 'Mathematica, launched in 1988, is a cornerstone for researchers and innovators worldwide.', 'Wolfram Alpha, a computational knowledge engine, is utilized by millions of users worldwide.', "Wolfram's focus is on the Wolfram language, a powerful programming language.", 'Dr. Wolfram became the youngest recipient of the MacArthur Fellowship at age 21.', 'Wolfram earned a PhD in theoretical physics from Caltech by the age of 20.', 'ChatGPT has potentially discovered a semantic grammar of language, enabling the creation of semantically reasonable content, potentially fostering computational creativity.', 'The combination of Wolfram Alpha and ChatGPT could result in a hybrid Turing machine, representing a potential advancement toward artificial general intelligence.', 'Large language models, including ChatGPT, capture a static snapshot of the world, setting a standard for knowledge work using existing human-created content.', "The 'ruley ad' concept has seen significant breakthroughs, contributing to our understanding of fundamental physics and computational terms.", 'Computation serves as a giant escape valve, enabling the exploration of a vast number of new possibilities.', 'The concept of Rulliad as the computational structure underlying physics and mathematics, with constraints on computation and persistence in time resulting in the observation of physics as we know it.', 'The phenomenon of simple programs leading to extremely complicated behavior is a fundamental scientific result with wide-ranging implications, exemplifying the significance of recognizing previously unnoticed occurrences.', 'The concept of computational irreducibility and its implications in predicting complex behaviors in computational universe. It explains the limitations in predicting complex behaviors and the importance of following through each step in understanding complex systems.', 'The principle of computational equivalence and its ubiquity in systems exhibiting non-obviously simple behavior. It highlights the ubiquitous nature of computational equivalence in systems and its implications for understanding computational sophistication.', 'Computational irreducibility hinders neural net training as it requires explicit computation, which goes against the approach of predicting outcomes.', 'ChatGPT serves as a new form of linguistic user interface, enhancing communication with the computational world.', 'Neural nets are trained on the things humans produce and may have a finely tuned sense of how humans think about things.', 'Large language models may be used to identify mathematical theorems considered interesting by humans and to find unnamed concepts in the computational universe.', 'The importance of conceivability and intelligibility in computational modeling, emphasizing the need for a cognitive bridge between human thinking and computational possibilities, is discussed.', 'Exploring the potential for AIs to advance science and build better versions of themselves.', 'The challenge of finding an intermediate language of neuroscience that transcends individual neuron firings and psychology.', 'The addition of Wolfram Language and Wolfram Alpha to large language models emphasizes the depth of computation, irreducible computation, and structured knowledge, offering a representation of real-world entities and a departure from mere statistical occurrences on the web.', 'The concept of computational irreducibility leads to an infinite collection of specific pieces of computational reusability within a system, despite the overall computational irreducibility, potentially impacting inventions and predictability.', 'The alignment between neural nets and human generalization, possibly due to the similarities in underlying architecture, presents an area of fortunate convergence and a potential avenue for further exploration in neuroscience and machine learning.']}