@IntuitMachine's Threads
I read a ton of books, but Kevin Kelley's book "Inevitable" is a must-read for anyone trying to understand the present and the future: amazon.com/Inevitable-Und… .
"Wayfinding", "Scale", "Possible Minds", "Beginning of Infinity", "Surfaces and Essences" are my top five recent other books worth reading.
"The Book of Why", "Other Minds: The Octopus...", "From Bacteria to Bach", "The Tangle Tree" and "Homo Deus" are books I would also recommend. The value of books is that they have more depth and are encapsulated with a higher-order structure.
OpenAI's experiments in co-evolution (deep reinforcement learning and self-play) openai.com/blog/emergent-… reveals the emergent complexity of intentional agents competing and cooperating among themselves.
This kind of technology cannot be undersold. It's a paradigm shift not only in technological terms but also in scientific terms. Detractors who keep pitching expectation moderation are in fact performing a disservice to society as a whole.
Many investors lost their shirt when the dot-com burst. Yet society as a whole benefited wildly from the bust. Society discovered an extreme surplus of cheap fiber optics that accelerated progress in many other fronts.
If you can't simulate and understand biology then you can't overcome the limits of AGI. Well that's the opinion of this paper: researchgate.net/publication/33… #ai
I'm going to argue however that simulating biology is a harder problem than AGI. So I expect AGI to happen before we solve human biology. Kurzweil's immortality will have to wait for AGI to happen.
But why is it that we can bypass the complexity of biology and invent more streamlined general intelligence? That's because biology creates Rube Goldberg like mechanisms. Biology is gratuitously complex.
UBI is currently a hard sell for the prevailing world view that existed from the dawn of homo sapiens. That is, you got to work to survive. Nevertheless, I am hopeful that this antiquated notion of human existence will eventually be obsolete.
I've learned a lot since writing my first book on AI (deeplearningplaybook.com). But indeed it is a surprise that @AndrewYang recognized the problem and took this message to the national consciousness. Humanity First is a difficult idea to accept because it is so rare!
@AndrewYang The more one studies Artificial Intelligence, the more one cannot avoid the meaning of being human. Being human revolves around our "vocation". It's derived from the word "to call". It can be interpreted as an occupation but also as a way of living.
I wonder if boredom is like hunger. We intuitively understand that experiencing hunger can have a beneficial effect in one's outlook, but what about boredom? nytimes.com/2019/02/02/opi… .
In today's information society, boredom has become a new scarcity. One can always find an external device to extricate oneself from boredom. However, is extricating oneself without external tools a cognitive skill that is vital?
People with minds overflowing with ideas rarely encounter boredom. Is boredom a symptom of a mind with insufficient ideas worth exploring? Is boredom that realization that there's very little that there is novel in one's thoughts?
It is essential to frame Complex Adaptive Systems from the lens of a learning system. The dynamism of complex systems are a consequence of its learning algorithm. Natural Evolution and Cultural evolution are both learning algorithms.
Evolution consists of generative models that not only generates learning systems but also learns to improve its generative models. We find 2nd order feedback not on just in the system itself but in the generative model of the system.
Perhaps the only pragmatic way to understand a complex system is to understand the dynamics of the generative model. Focusing on the generated model is like focusing on just the trajectory of a Lorenz attractor but ignoring the generating function.
Philosophically, it (Deep Learning) is every rationalist’s nightmare: the triumph of the inexplicable over the clear, and that of the probabilistic over the deterministic... thenewcentre.org/seminars/parad…
"Here Charles Sanders Peirce’s tychism and Georg Simmel’s dynamism will help us understand what these (Deep Learning) developments mean for the future of human (and machine) subjectivity." #deeplearning #ai
en.wikipedia.org/wiki/Tychism - Peirce formulates four reasons in support of his hypothesis: Growth and increasing complexity, Variety and diversity, Regularity (laws of nature),
Mind/consciousness/feeling
Mind/consciousness/feeling
Reductionist methods are as limited in explaining the emergence of abstract thought as explaining liquid water knowing that it is composed of H2O. #ai #agi
It's difficult enough to derive emergent properties from inanimate components. Now imagine how difficult it would be to derive emergent properties from animate (or intentional) components. medium.com/intuitionmachi…
It is easier to discover models that demonstrate how (generative) than finding models that explain why (explanatory). This is deep learning approaches that can build generative models move us an incremental step forward in understanding complex systems.
DeepMind now researching the connection of Sleep to Deep Learning: deepmind.com/blog/article/r… #ai #deeplearning
Biological brains require sleep, but what exactly does sleep have to do with cognition? medium.com/intuitionmachi…
But why do brains require re-experiencing events to synthesize its understanding of these events? It seems that what humans remember about experiences are the summaries of these experiences played backward. That is how we build our causality networks!
In this interview by @lexfridman of @ylecun Lecun says the math of deep learning has more to do with Cybernetics than that found in computer science. Norbert Wiener formulated Cybernetics based on the idea of negative feedback.
@lexfridman @ylecun This was before electronic computers were invented. This was when Electrical Engineering was seeing a split into power engineering and control engineering. It's the latter that Wiener's work had its influence.
@lexfridman @ylecun Artificial Intelligence was coined 15 years later and Wiener was not invited. The A.I. proponents wanted to focus on the emerging tools of computer science and not the older tools of continuous mathematics.
"Immune cells called microglia don’t just fight pathogens. In the developing brain, they sculpt neuron branches, helping them to mature." nytimes.com/2019/08/29/sci…
Indeed there's a possibility that the meta-learning program in the brain originates from the microglia. I've written about this in 2016: medium.com/intuitionmachi…
Immune systems, even the CRISPR system, are intrinsically decentralized pattern matching systems. In fact, T-cells have been shown to grow into neuron-like structures.
Deep Learning (DL circa 2012) big breakthrough is that it demonstrated a new kind of Artificial Intuition. Prior to DL, computers provided programmable Artificial Logic (circa 1945).
AlphaGo big breakthrough is that it demonstrated that with the use of DL one could beat the best human in a game of intuition (i.e. Go). .
AlphaZero big breakthrough is that you could train a system from scratch (without human knowledge) and it could be the best human system Stockfish that was programmed using logic. medium.com/intuitionmachi…
The entire history of computer science can be framed from the perspective of scaling specification into robust implementation.
Attempts using analog circuitry had difficulty scaling due to accumulative errors in their iterative operations.
This was solved through the transition of digital based computation, however the original sequential behavior of algorithms remained.
It's ironic that Von Neumann described the circuits of his new invention (i.e. the computer) in terms of McCulloch-Pitts logic for neurons.
Walter Pitts was homeless at 13 and hid in the local library. In the library, he read all 3 volumes of Bertrand Russell's Principia Mathematica. nautil.us/issue/21/infor…
Godel however proved later that Russell's ambitions to frame all knowledge using mathematical logic had fundamental limits in its completeness. This sentiment was also expressed by Norbert Wiener (while as a student of Russell).
Reinforcement learning does not scale. Efficient environmental sampling does. One aspect of intelligence is the ability to predict the world while observing so little. #deeplearning #ai
To observe so little implies that there's an imaginary simulation that is able to extract new information without interacting with the environment. In short, a game is played between imagination and reality.
Deep Learning has revealed to us what methods work very well in reality. This leads to imagined new architectures that will be exponentially more capable than the previous architectures. #deeplearning #ai #agi
