Jeffrey Shainline is a physicist at NIST working on. Note: Opinions expressed by Jeff do not represent NIST. Please support this podcast by checking out our sponsors: - Stripe: https://stripe.com - Codecademy: https://codecademy.com and use code LEX to get 15% off - Linode: https://linode.com/lex to get $100 free credit - BetterHelp: https://betterhelp.com/lex to get 10% off EPISODE LINKS: Jeff's Website: http://www.shainline.net Jeff's Google Scholar: https://scholar.google.com/citations?user=rnHpY3YAAAAJ Jeff's NIST Page: https://www.nist.gov/people/jeff-shainline PODCAST INFO: Podcast website: https://lexfridman.com/podcast Apple Podcasts: https://apple.co/2lwqZIr Spotify: https://spoti.fi/2nEwCF8 RSS: https://lexfridman.com/feed/podcast/ YouTube Full Episodes: https://youtube.com/lexfridman YouTube Clips: https://youtube.com/lexclips SUPPORT & CONNECT: - Check out the sponsors above, it's the best way to support this podcast - Support on Patreon: https://www.patreon.com/lexfridman - Twitter: https://twitter.com/lexfridman - Instagram: https://www.instagram.com/lexfridman - LinkedIn: https://www.linkedin.com/in/lexfridman - Facebook: https://www.facebook.com/lexfridman - Medium: https://medium.com/@lexfridman OUTLINE: Here's the timestamps for the episode. On some podcast players you should be able to click the timestamp to jump to that time. (00:00) - Introduction (05:56) - How are processors made? (25:15) - Are engineers or physicists more important (27:43) - Super-conductivity (43:31) - Computation (48:07) - Computation vs communication (51:48) - Electrons for computation and light for communication (1:02:32) - Neuromorphic computing (1:27:23) - What is NIST? (1:30:41) - Implementing super-conductivity (1:38:20) - The future of neuromorphic computing (1:57:54) - Loop neurons (2:04:09) - Machine learning (2:18:36) - Cosmological evolution (2:25:44) - Cosmological natural selection (2:43:05) - Life in the universe (2:50:52) - The rare Earth hypothesis
The importance of usability in developing therapy tools should be prioritized in order to create products that are efficient and enjoyable to use. Additionally, incorporating professional therapist help will also improve the overall quality of the product.
This episode explains the basics of quantum computing, how it differs from classical computing, and how quantum processors are built.
The band gap of silicon and the combination of group three and group five compounds make it an elemental semiconductor, which has been crucial in advancing Moore's Law. This intersection of engineering and physics is a beautiful achievement.
Superconducting materials can carry electric current without any resistance, a phenomenon that only happens at extremely low temperatures when the noise of atoms moving in the lattice is low. Josephson junction is a device made from superconducting wires separated by a gap of insulator or normal metal, enabling the study of quantum mechanics at the macroscopic scale through the Josephson effect.
This episode discusses the potential of Josephson circuits, including their ability to process information and sense weak magnetic fields, and why they haven't yet displaced silicon microelectronics in conventional digital computing.
The 90s introduced superconducting logic gates based on Josephson junctions for digital computing, while the brain's computation is based on a network of simple computers performing basic arithmetic operations.
This podcast discusses the differences between communication and computation, and how neurons are capable of performing various inputs, outputs, and computations in different contexts.
The use of superconducting detectors could lead to the detection and manipulation of single particles of light, or photons, which could be used to encode information in a quantum computer. A newly developed device using superconducting wires could allow for the detection of a single photon and trigger the necessary subsequent computations required for quantum computing.
Neurons that are closer to each other are more likely to have faster time scale activities, spreading information throughout the brain's different modules. The brain's temporal activity is consistent with a power law, showing that neuron oscillation and spike production have no defined scale of activity.
The development of artificial intelligence that captures comprehensive and complex human-like activity requires building separate modules, which specialize in different responsibilities, similar to the human brain's structure, as well as studying the human brain's chemistry for processing information.
The brain has an internal mechanism to adjust synapse weights to store useful information, which is similar to supervised learning in neuromorphic computing where the external user controls the hardware's synaptic weights. Growing new neurons doesn't seem to be the primary mechanism by which the brain learns.
A team at NIST is working on advancing light-based communication technology by developing a light source on a silicon chip that could be as simple as a transistor, paving the way for processors to use light for communication above certain distances. The team is comprised of individuals with backgrounds in photonics, nanotech, and electrical engineering.
While silicon chips are excellent in driving light source chips and modulating light intensity, efficiently creating light sources compatible with silicon chips for use in communication requires exotic methods that compromise its interface capability with transistors.
The integration of light sources with superconducting electronics is a promising approach, as it avoids the need for intimate integration with transistors. By placing compound semiconductor light sources on a silicon wafer and growing superconducting electronics on top, signals can be received to produce pulses of light for downstream synaptic connections.
This podcast episode discusses the use of circuits based on Joseph's injunctions in neuron hardware platforms and the impact of low temperatures on achieving advancements in technology.
To match the scale of the human brain, wafers need to be stacked on top of each other with fiber optic communication between them, and 10 billion neurons can be achieved in a space the size of a table. Superconducting single photon detectors can offer analog values with up to 10-bit equivalent resolution, and silicon nitride can be used to form rectangle-shaped paths that target subregions of the network.
The development of loop neurons provides a platform to not only advance neuroscience research, but also explore physics concepts such as critical phenomenon and Ising models through the use of superconducting loops. The loop neurons work through a system of coupling between synapses using transformer technology.
The possibility of moving away from neuromorphic systems, towards physical systems such as superconducting electronics or single photon communication, to better understand human intelligence and general intelligence is being discussed. This comes as large-scale machine learning training systems are being developed, such as Tesla's Dojo.
Dojo is a machine learning focused supercomputer that aims to be modular and scalable for various applications, with its first test being for self-driving technology. Elon Musk's team uses a process of deploying a very basic system, identifying its weaknesses, and then improving iteratively until it becomes more intelligent.
The parameters that define our universe seem optimized to allow for a complex, long-lived universe. In order to simulate such complexity, large scale spiking neural networks with vast communication would have something tremendous to offer.
The laws of physics in a universe could be subject to slight mutations so the offspring universe doesn't have the same parameters as its parent universe. An evolutionary process could lead to the development of the technology that shapes and define a specific universe.
The hypothesis that intelligent civilizations have the ability to produce black holes due to their technological maturity suggests that there may be a sufficient number of similar civilizations in our universe. This is because the parameters of the universe have been selected in a way that makes it possible for them to evolve and ultimately create black holes.
The likelihood of finding at least one intelligent species in the galaxy is highly probable, considering the vastness of the universe, but most of the universe is made up of boring single-cell organisms.
The speaker discusses the possibility of cognitive systems much more capable than humans, residing on, or off, the surface of the earth in environments that are much more suited to them.