 |
RESONANCE THEORY: PART VII What Is Spin?
By Mark Anderson Why Read: It's no secret that many secrets remain in the most fundamental of sciences: physics. And it's no news to the tech-minded members of SNS that physics discoveries provide the most fertile ground for the creation of new technologies and their related products. It should not surprise, therefore, that the nature of "spin" in the quantum world, while incorporated as a design aspect of many new and upcoming technologies, is not well understood. In this week's discussion, we will hope to resolve that problem with a proposal that also sheds new light on many other aspects of the fundamentals of the world around us. - mra ________ [. . .] and books that told me everything about the wasp, except why. - Dylan Thomas, A Child's Christmas in Wales
I learned very early the difference between knowing the name of something and knowing something. - Nobel Physicist Richard Feynman
There are no particles. Let's begin by settling that question forever. Dedicated to Richard Feynman, Elie Cartan, Roger Penrose, John Cramer, and David Brin
(Note: I am using the British spelling of "programme" to indicate a long-term research plan.) The original insight of Resonance Theory, noted in the first paper of the series, was: "The laws of physics derive directly from the physical properties of space." In other words, contrary to then-current belief, space is neither just spacetime nor is it an empty void. The point of the Resonance Programme was to flip the science of physics 180 degrees: stop studying the "particle zoo" and start studying space itself, in order to better understand its many instantiations. (For a full listing of the papers and work to date in Resonance Theory, see "Upgrades" below.)
We're going to start this exploration by looking at the dogma of what is not true but is still taught today. For the sake of simplicity, we will restrict our subject to electron spin. Here is an illustration of a spinning electron sphere, showing spin up and down states:
Let's start by saying that we know, from experiment, that electrons are not little balls, and their spin is definitely NOT like the spinning of a top - i.e., what is shown here. Everyone knows this, yet we still use these pictures. Rather, electron spin is an elementary magnetic aspect of electrons, measurable by passing them through a magnetic field and observing the effect. Essentially, some will turn upward in that field, and some will turn downward - hence the slight misnomer of "up" and "down." This was shown early, in what is called the Stern-Gerlach experiment. It is worth noting that spin is also referred to as spin angular momentum.
By Tatoute - Own work, CC BY-SA 4.0 Stern-Gerlach experiment: Silver atoms travelling through an inhomogeneous magnetic field, and being deflected up or down depending on their spin; (1) furnace, (2) beam of silver atoms, (3) inhomogeneous magnetic field, (4) classically expected result, (5) observed result One can see on the left (4) the portrayal of what scientists first expected to see - a smooth line of deflected particles from top to bottom; and (5) the binary split, from magnetic north to south, of the actual observed particles (in this case, atoms). In this way, physicists first learned that spin was not only not spinning, but it was also quantized, or available only in discrete amounts of magnetic response. What do "up" and "down" really mean, other than this? How does this happen, and what could the actual mechanism of spin tell us about larger questions? Caltech, where Nobel physicist Richard Feynman spent most of his professional life, is arguably a world leader in the study of physics, and of the theory behind electrons and light. Here is a good Caltech rundown about assistant professor Chip Sebens, written last year, on the question of whether electrons spin. You can imagine you are a young new PhD student in his class: Deep inside all matter in the universe, electrons are buzzing around and behaving as if they are twirling around on their axes like spinning tops. These "spinning" electrons are fundamental to quantum physics and play a central role in our understanding of atoms and molecules. Other subatomic particles spin, too, and the study of spin has technical applications in the fields of chemistry, physics, medicine, and computer electronics. But many physicists will tell you that electrons are not really spinning -they merely act like it. For example, electrons have angular momentum, which is the tendency of something to keep rotating - like a moving bicycle wheel or a spinning skater - and because they have this property, one might conclude they are spinning. Further evidence comes from the fact that electrons act like little magnets, and magnetic fields arise from rotating charged bodies. The problem with the notion that electrons spin is, due to their miniscule [sic] size, the electrons would have to be spinning faster than the speed of light to match observed angular momentum values. (Think of an electron like a spinning skater with their arms folded inward: the smaller the overall size, the faster it spins.) Chip Sebens, assistant professor of philosophy at Caltech, wants to go back to the drawing board and rethink this notion. As a philosopher of physics, he wants to figure out what is really going on at the deepest levels of nature. "Philosophers tend to be attracted to problems that have been unsolved for a really long time," Sebens explains. "In quantum mechanics, we have ways of predicting the results of experiments that work very well for electrons and account for spin, but important foundational questions remain unanswered: Why do these methods work and what's happening inside an atom?" Summary: Electrons are not spinning balls, and no one knows how they work.
The industrial uses of spin today: Spin's applications also include electron spin resonance (ESR); below is a selection of fields for this use, starting on the bio side. If you imagine nuclear magnetic resonance (NMR) and magnetic resonance imaging (MRI), this is similar - getting information from materials from their electron spin characteristics:
Here is a brief description of ESR: Electron spin resonance (ESR) spectroscopy has been widely applied in the research of biological free radicals for quantitative and qualitative analyses of reactive oxygen species (ROS) and reactive nitrogen species (RNS). The ESR spin-trapping method was developed in the early 1970s and enabled the analysis of short-lived free radicals. This method is now widely used as one of the most powerful tools for free radical studies. In this report, some of the studies that applied ESR for the measurement of ROS and RNS during oxidative stress are discussed. Of course, there are many applications of spin beyond bio and med, and most of them seem to be in electronics. Specifically, there is the general applied field of "spintronics": "Spintronics (a portmanteau meaning spin transport electronics), also known as spin electronics, is the study of the intrinsic spin of the electron and its associated magnetic moment, in addition to its fundamental electronic charge, in solid-state devices. The field of spintronics concerns spin-charge coupling in metallic systems; the analogous effects in insulators fall into the field of multiferroics. Spintronics fundamentally differs from traditional electronics in that, in addition to charge state, electron spins are used as a further degree of freedom, with implications in the efficiency of data storage and transfer. Spintronic systems are most often realised in dilute magnetic semiconductors (DMS) and Heusler alloys and are of particular interest in the field of quantum computing and neuromorphic computing." For instance, spintronics is vital to a new era of computer memory: A spintronic device utilizes the spin degree of freedom in storing and transmitting information. Such devices can display highly useful properties, such as non-volatility, low energy consumption, fast processing speed, and scalability to tiny dimensions that allow for high density. A spintronic device usually comprises several magnetic layers and non-magnetic layers stacked together. A typical spintronic device is the magnetic tunneling junction (MTJ) [. . .] for an in-plane MTJ (i-MTJ) where the magnetization in both FMs lie in-plane and for a perpendicular MTJ (p-MTJ) where the magnetization lies normal to the plane. The MTJ is the main building block of the magnetic random access memory (MRAM), an emerging non-volatile memory among others, such as resistive random-access memory (RRAM), phase-change memory (PCM), and ferroelectric random-access memory (FeRAM). An MTJ consists of two FM layers separated by a thin tunneling dielectric layer, usually magnesium oxide (MgO) for achieving high TMR. The resistance of the MTJ, which represents the magnetic bit, is determined by the relative magnetization orientations of the two FM layers. Typically, parallel magnetizations result in a larger probability for tunneling in the MTJ and hence a lower resistance ("0" state), and anti-parallel magnetizations result in a lower probability for tunneling and hence a higher MTJ resistance ("1" state). The magnetization of one FM layer is fixed in one direction, whereas the magnetization of the other FM layer can be altered by using current induced magnetic fields or spin polarized currents generated from spin-transfer torques (STTs) or spin-orbit torques (SOTs). Typically, an MRAM bit cell is composed of one MTJ and one or several access transistors. [. . .] The density of the memory not only depends on the size of the MTJ, but also on the footprint of the access transistors, which, today must be larger in area than the MTJ, to generate enough current to switch the MTJ reliably. So, spin matters - scientifically, technically, commercially. Summary: Spin, in many forms, is providing a base technology for many of the leading technologies just coming online today in medicine, agtech, bio, computing (processing and memory), and related fields. Indeed, spin may be the most important conjoining of science and technology since the chip and the laser, and there is no reason to think our increased understanding of spin will not improve both.
Spin is at the center of how atoms are built. Electron pairs of opposite spin are at every level of atomic energy. It turns out that electrons at given energy states must have different up/down spins, as pairs: [. . .] no two electrons can be in exactly the same state (this is known as the "Pauli exclusion principle," as it was first proposed by Wolfgang Pauli). This is the property that makes multi-electron atoms different from each other: as you try to construct the lowest-energy state for an atom with a given number of electrons, you're forced to "fill up" the allowed states that correspond to particular electron orbits, with each orbit getting two electrons with different spin states. Hydrogen has a single electron in the lowest-energy spatial wavefunction with, say, spin up. Helium adds a second electron to (more or less) the same wavefunction, with its spin down. Lithium, the next element up, can't fit its third electron in there, so it needs to go into the second-lowest energy state, with spin up, and beryllium puts a second electron in there with its spin down, and so on. If you took chemistry in high school and recall weeks of drawing little up and down arrows in groups of 2, 2, 6, 2... what you were doing was applying that weird spin rotation rule. This pattern of sorting electrons into "shells" is what determines an element's position on the periodic table: they're sorted into columns by the number of electrons in their outermost shell. The column position of an element determines the number and type of bonds it will form with other elements, which is the basis for everything in chemistry." - Forbes, "Three Everyday Things That Couldn't Exist Without Quantum-Mechanical Spin" In other words, spin is the core aspect of all of chemistry.
In one of Feynman's many books on the subject, he talks about the behavior of light and electrons. In one section, he breaks it all down into a few easy steps. He says of spin (and I paraphrase): "Well, here you just have to believe me. When you turn the spin all the way around, 360 degrees, you are only halfway home, and you have to turn it another 360 degrees, or 720 degrees total, to come all the way back around. You just have to believe me on this, I don't know why it is." Right, now we have your attention. How could that possibly be? I chose this excerpt from Forbes on precisely this issue, because it was easier to find than in my Feynman library: The strangest and most subtle of the properties of spin has to do with what happens when you rotate these particles. If you take something with ordinary angular momentum, like a spinning molecule, and apply torques to it that rotate the direction of its axis through 360 degrees, at the end of the process, you're back to where you started: the wavefunction for the particle after a 360 degree rotation is mathematically identical to the wavefunction before the rotation. Do the same thing to an electron spin, though, and you don't end up back where you started - the wavefunction of a spin after a 360 degree rotation is the negative of what you started with. You need to rotate through 720 degrees - two full rotations of the spin axis - before you return to the original state. How can we resolve this apparent 2x revolution requirement? This has had me stymied, I will admit, for over a decade.
Recently, I realized that, functionally, I was envisioning the metaphor of a Mobius strip, which also has doubled periodicity. It didn't take but a few minutes to discover that this formalism had been around since early last century, and applied to spin: [There are mathematical constructs called spinors, which we might introduce at this stage, first invented by Elie Cartan and perfected by mathematician Roger Penrose. Early in my research, the spinor struck me as having the potential to be one of the primordial mathematical objects from which we could derive all physical particles. Since Roger had done brilliant work on both spinors and twistors, I visited him at Oxford to see whether he agreed - and, of course, he did, after giving me a quick Penrose quiz on physics.] In this vein, here is an illustration of Mobius spin with a 720-degree period, having gone halfway around:
A spinor visualized as a vector pointing along the Mbius band, exhibiting a sign inversion when the circle (the "physical system") is continuously rotated through a full turn of 360. Note that, since the sign is inverted on its first 360 degrees, it must make another 360-degree turn to come to home again. Or: In geometry and physics, spinors (pronounced "spinner" IPA /spɪnər/) are elements of a complex number-based vector space that can be associated with Euclidean space. A spinor transforms linearly when the Euclidean space is subjected to a slight (infinitesimal) rotation, but unlike geometric vectors and tensors, a spinor transforms to its negative when the space rotates through 360 (see picture). It takes a rotation of 720 for a spinor to go back to its original state. But we are not dealing with vectors now; we are looking at real waves moving through real space with real resonant characteristics. While the above constitutes a better model than the one at the top of this paper, it is still just another construct - say, for class demonstration. The reality, in this proposal, is much more interesting, translating the behavior of this model into actual wave mechanics, in three dimensions. What is the solution? Clearly, our new proposal will need to explain this conundrum. But wait - there's much more in the "Physics' Unexplained Phenomena" mystery box we call "spin" . . .
Everyone in computing has heard of quantum computing, and most know that this is possible thanks to a (complete black-box of a) property called "entanglement." This describes the ability of two electrons, for example, to have spin values that are opposite and locked together - or entangled. It is perhaps the least understood (Einstein hated it), most important unknown in physics today. It's all the more important because it's a future key to breaking most of the world's encryption schemes, essentially overnight, and to moving computing into an entirely new world order of speed and accuracy, and revolutionizing telecommunications, and so affecting banking, and warfare, and - you get the picture. And, at the heart of entanglement? Spin. The person (or company, or country) who figures out spin properly will have a massive advantage in moving forward in entanglement. Well, that adds a bit of spice to our proposal today, doesn't it? And then, to add (unnecessarily) even more to the importance of this project, we have an unresolved science quandary that dates back to about 1801 . . .
Ever since the first double-slit experiment of that date - in fact, dating back to the Ancient Greeks - scientists have been unable to resolve the debate about the basic nature of the building blocks of the universe. Are they particles or waves? Now, these days, you could also mutate this conversation into: Are they particles or fields? But we're going to make that question obsolete as well. Before we get to our proposed solution, let's add one more data waypoint to our search. Members may recall, from our paper on "How to Build an Atom" (see "Upgrades"), our coverage of Anne L'Huillier, co-winner of the 2023 Nobel Prize in Physics. Anne has spent much of her time figuring out how to take the very first photograph of an electron - something we all might be interested in, given the above issues we're trying to clean up. To do this, she is using shutter speeds in laser pulses of "attoseconds." An attosecond is 10-18 seconds. Here's something interesting: in describing the first pictures of an electron, she was asked what they were showing. To paraphrase her response: "I believe we are seeing what appears to be the leading edge of the electron wave." Just to be a bit mischievous on our way to the final proposal, I'll add one more impediment faced by those who, until now, thought electrons were particles rather than waves: according to almost all physicists, it is impossible for the point electron to be spinning fast enough (yeah, we're back to the spinning-top metaphor) to generate spin effects, since this would mean its outer edge would be traveling faster than light. Now, there's a non-starter. But of course, we have already given up on the whole spinning thing. Let's review.
We'll first summarize the main contributors to this proposal that need to be addressed by our proposed solution:
As SNS members can likely tell, I have been thinking about this problem for a while; so, naturally, there are additional requirements and facts that have gone into this current proposal. And let me say, at the outset: there is no way that this is a final answer. Rather, while I believe it is absolutely correct as far as it goes, the primary value of this proposal is to change the fundamental questions being asked and the basic physical issues being considered. I do think that this work will allow a great deal of new research to be done, of a different quality and type, than we are doing today, and with a corresponding acceleration in learning about space itself, about particles, about waves, and about the interaction of light and matter.
Now, we should be free to move beyond the particle zoo of the standard theory and into a deeper world of wave mechanics and interactions, driven by the properties of space; and start defining and describing the many ways in which space can dance with itself. Before I wrote this issue, I wanted to test these ideas on a few credentialed physicists for their reactions. It turned out that David Brin, a PhD physicist and leading science-fiction author, had just been awarded Fellow status from his alma mater, Caltech, on a recent day when we were having dinner outdoors at his house. Over coffee, I explained the Resonance spin proposal; it took about four minutes. His first reaction: "Very interesting" (thoughtful, looking up at the sky above us). His second reaction, a minute later: "I can see how this might be right" (getting a bit more into it). Finally, another minute later: "I can't see why this wouldn't be correct - I like it! It explains so many things . . ." "But you know what matters even more?" he added. "I think Feynman would have loved it."
Your comments are always welcome.
Sincerely, Mark Anderson
Information and material presented in the SNS Global Report should not be construed as legal, tax, investment, financial, or other advice. Nothing contained in this publication constitutes a solicitation, recommendation, endorsement, or offer by Strategic News Service or any third-party service provider to buy or sell any securities or other financial instruments. This publication is not intended to be a solicitation, offering, or recommendation of any security, commodity, derivative, investment management service, or advisory service and is not commodity trading advice. Strategic News Service does not represent that the securities, products, or services discussed in this publication are suitable or appropriate for any or all investors.
We encourage you to forward your favorite issues of SNS to a friend(s) or colleague(s) 1 time per recipient, provided that you cc info@strategicnewsservice.com and that sharing does not result in the publication of the SNS Global Report or its contents in any form except as provided in the SNS Terms of Service (linked below). To arrange for a speech or consultation by Mark Anderson on subjects in technology and economics, or to schedule a strategic review of your company, email mark@stratnews.com. For inquiries about Partnership or Sponsorship Opportunities and/or SNS Events, please contact Berit Anderson, SNS COO, at berit@stratnews.com.
The Resonance Theory Programme This is the seventh white paper on Resonance Theory. All those to date are linked below, as well as additional related pieces previously published in SNS. Resonance Theory has been based on two main pillars:
Resonance Theory has been laid out in previous issues of SNS:
Related Materials Video: FiRe 2022: "Redefining the Cosmos: The Resonance Approach: David Brin interviews Mark Anderson" (52:35) "SNS: Special Alert: A New Resonance Era for the Cosmos" (8/28/22) "SNS: How to Build an Atom" (9/20/23) Additional Archives Almost all of the ideas and work described in these issues were originally concluded, after two years of effort, in 1979, with the first dated and serial-numbered submission to Physical Review Letters D being in 1981. Of course, PRL didn't publish it, since it used additional new stringlike wave geometries made of otherwise-empty space, rather than the fields the peer reviewers were accustomed to. The original paper can be found here: https://www.stratnews.com/resonance/ And the supporting documents here: https://www.stratnews.com/resonance-support/
Gabriela Cowperthwaite | Writer, Director & Producer, The Grab; Director, Blackfish Member #: 20831
Gabriela Cowperthwaite has been directing both documentaries and narrative films for over 20 years, rising to prominence with the 2013 release of Blackfish, which examined the plight of orcas and their trainers at SeaWorld. Blackfish quickly became one of the highest-grossing documentaries of all time, whilst giving rise to the expression "the Blackfish effect" - a label that described the film's ability to permanently change SeaWorld's practices and disrupt its business model, while changing captivity laws worldwide. Blackfish was shortlisted for an Academy Award; nominated for a BAFTA, a Broadcast Critics' Awards, and an International Documentary Association Award; and won the Satellite Award for Best Feature. In early 2024, Gabriela's space-station thriller, I.S.S., starring Chris Messina and Academy Award winner Arianna DeBose, premiered worldwide. Her critically acclaimed film Our Friend, starring Casey Affleck, Dakota Johnson, and Jason Segel, was released in 2020. Based on a true-life story, Our Friend told the story of a couple struggling through a diagnosis alongside their best friend. Megan Leavey, which premiered in 2017, is based on the true-life story of a Marine corporal (Kate Mara) whose unique bond with her military combat dog saved many lives during their deployment in Iraq. Last year, Gabriela helmed Children of the Underground, an FX/Hulu series highlighting Faye Yager, a woman who started a dark underground network that has been hiding children from their abusers for decades. The series was nominated in 2023 for an Independent Spirit Award for best documentary series. Gabriela is currently writing an original script for Imagine Entertainment. Her latest documentary, The Grab - a geopolitical thriller - follows a shadowy world of powerful entities who are grabbing up the world's most precious resources while we all look the other way. The Grab opened TIFF in 2022. The film, released theatrically in June 2024, is available on demand and streaming and is being hailed as the "holy sh*t documentary of the year." SNS Connection: The Grab has been selected to be this year's FiRe Featured Film; it will be screened for all FiRe attendees on Oct. 21. A live Q&A will follow with both Gabriela and Nate Halverson, senior producer and reporter for the Center for Investigative Journalism, whose investigative reporting is featured in the film. Watch the "Academy Conversations: Oscars Contenders" interview with Gabriela (June 2024, 12:04)
"[. . .] quantum-mechanical spin is sneakily one of the weirdest properties in quantum physics, and also the most important." - Chad Orzel, Forbes
"Sometimes progress in physics requires first backing up to reexamine, reinterpret and revise the theories that we already have. To do this kind of research, we need scholars who blend the roles of physicist and philosopher, as was done thousands of years ago in Ancient Greece." - Asst. Professor Chip Sebens, Caltech, in an Aeon essay
"No one fully understands spinors. Their algebra is formally understood but their general significance is mysterious. In some sense they describe the 'square root' of geometry and, just as understanding the square root of -1 took centuries, the same might be true of spinors." - Michael Atiyah, as recounted by Paul Dirac biographer Graham Farmelo; Wikipedia
"We used to enjoy the best of both worlds. Now we're losing on both ends." - A VC who has worked funding companies in both China and the US; quoted in the New York Times That's what comes from sleeping with the enemy.
"It's basically the same fracking technology. The difference is that we're going after clean heat instead of hydrocarbons." - Cindy Taff, Sage CEO and 36-year Shell Oil veteran, on using geothermal heat to power data centers; ibid.
"We are the number one people's party in Thuringia. We have achieved a historic result." - Bjorn Hocke, regional leader of the AfD (far right) party in Thuringia; quoted in the Washington Post
"It is impossible to have a wave without a medium that is waving." - Mark Anderson, on the initial insight behind Resonance Theory, back when the dogma was that space was just empty space
Ed. Note: Some of the letters below have been republished to include subsequent replies.
Subject: SNS: THE TECHNOLOGY ENTROPY TRAP: PRODUCTIVITY VS. CHAOS Mark, This is super interesting - technophilosophy - many thanks - Yrs,
Kim Stanley Robinson
Mark adds: Stan, We're very excited to see all of the good work you are doing on global warming, and hope that all our members are not only reading your books, but also following you closely online. We look forward to having you with us next year at FiRe 2025, again, and catching up - And thank you for adding a new word to my vocabulary - technophilosopher. Got it.
Mark Anderson
Mark, I'm so pleased you covered the complexity of GenAI. I'm working in this space and started with a bottom's up and top down view of the math, the industry and the practitioners. It is scary. It is very similar to when the walled garden of cell phones was opened and everyone could make apps. Everyone did, but most practitioners had no idea of the underlying infrastructure. Even the big tech guys initially made ridiculous errors due to poor understanding of mobile systems. But the overall complexity of those systems could be studied and understood. And the rate of change was measured in years. They are well understood by experts and eventually the apps were well bounded and behaved (technically). GenAI is different in two key areas. (1) Even the experts, of which there are very few, do not understand and can't effectively study the technology directly due to the sheer scale and complexity of the models and associated algorithms. (2) Unlike mobile technology and every other major wave, the acceleration rate of complexity and scope of GenAI is off the charts. Even the experts can't keep up. Is it too late? Is this genie out of the bottle not to be put back in? Thank you for continuing to think deeply and differently. It's important humanity continues to do so lest we perish more quickly. Warmest,
Mary Jesse CEO, Hexagon Blue LLC
Mark adds: Mary, We agree. There is a lot of room for great products and companies in ML and AI, but the needle is moving rapidly to the metrics of Trust and Transparency. I think we have no choice, given adoption rates of GPT; that we have to go forward, even as companies are finding that the "guardrails" are not built-in by vendors, but are in their decisions on which areas of operations do not require T and T.
Mark Anderson
Subject: "LLMS ARE ALREADY DONE: WHO WILL WIN - AND LOSE - IN THE NEXT PHASE OF MACHINE LEARNING" Berit, I keep referring to AWS as a distant third in a two horse genAI race that Google is busy losing to Microsoft, who is busy proving to the startups why it has the market cap it does. My other comment would be, as always, to look at who survey answers come from: those Predibase statistics are interesting as much as for what they say about who they're measuring (anyone who says they have two LLMs in deployment and is even considering the cost of training is a machine learning engineer and not in charge of company strategy). I took a fairly in-depth look at genAI adoption recently: https://www.cio.com/article/2096857/expectations-vs-reality-a-real-world-check-on-generative-ai.html - so in depth that my editor had to cut out my comments on a very similar survey from Intel. "Talking to IT teams like the AI professionals in Intel's 2023 ML Insider survey suggests a rather lower figure for organizations launching genAI solutions in production in 2023: just 10%. However, the fact that a quarter had deployed genAI models in production and the high numbers who view infrastructure as the biggest challenge in using LLMs suggest the study may cover those more likely to build than buy, which takes longer." Best to everyone!
Mary Branscombe Freelance Technology Journalism
Mary, Fantastic point about the Predibase stats, and thank you for sharing the Intel study. I had an interesting conversation yesterday with the CEO of a GAN company, who will be at FiRe, about what exactly you can count on LLMs to deliver without issue. GANs being, I think, one of the brighter points of LLMs as far as improving efficiency, information democracy, and access within organizations. Their perspective is to keep LLMs siloed to language parsing surrounding the delivery of information and to rely only on internal data for the actual information. This avoids accidental error caused by hallucinations and other issues, which still are and seemingly always will be a problem, given the probabilistic nature of LLM data generation. Thank you for taking the time to write in!
Berit Anderson
Mary adds:
Berit, GANs, RAG, a whole range of techniques (every provider likes the one they sell)... But what almost every company who wants to make generative AI useful needs to do is go back and audit 20-30 years of company data before they feed their GenAI tools on it because hardly anyone has done their access permissions and sensitivity labelling right! It's tedious, it's plumbing and it's utterly critical if you want to manage the information GenAI can use. Managing "hallucinations" means understanding tokens and understanding the difference between running out of tokens so you get results back that are no longer constrained to your prompt and the pure stochastic nature of LLMs. Oh, and investing in staff by giving them actual training that teaches them to use GenAI responsibly and verify the output. In other words, it's not magic and organisations have all the usual skills, governance and UX issues to deal with - but that's so much less exciting to talk about,
Mary Branscombe
Mark, We almost never have really understood all the ins and outs, implications, even enabling physics of our approaches, systems etc... What you write is an old observation, updated with the latest issues, but the basics are the same........we are always, at various us levels, have been, are, will be surprised....... Perhaps AI will help reduce surprises, because it knows FAR MORE than individual humans.
Dennis Bushnell HFAIAA, FASME, FRAeS
Mark adds:
Dennis, I think that AI (GPT) today will not do us much good in the way of avoiding surprises, except in the limited frame you are describing: in datasets that have far too many contextual variables for us to track with our brains, our systems will be able to provide such context, albeit often riddled with hallucinations or inconsistencies (like math problems). To avoid the "unknown unknowns" Donald Rumsfeld worried about, we will likely need to be using Explainable AI (XAI), which is not part of GPT and not commonly available (as compared with being commonly marketed) today. Pattern has it in two flavors (disclosure: I am the CEO), under the TrueXAI umbrella: ICE/XAI and Neural/XAI. These could help, by making new discoveries that would then not be surprising later.
Mark Anderson
Subject: CYBER-ATTACKS: THE NEXT-TECHNOLOGY GENERATION [A Conversation with Gilman Louie and Jody Westby] this was a great conversation, very illuminating. Thanks.
Kim Stanley Robinson
* On October 20-23, Mark will be speaking on a variety of subjects, and hoping to see many of our SNS members in person, at the FiRe 2024 conference at the Terranea resort in Palos Verdes, California. * On November 24, he will be at the Bio Europe conference in Stockholm. * On December 11, he will be attending the Fortune Brainstorm AI Conference in San Francisco, where he will yet again be trying to break through the walls of Valley bull***t and move toward trust and transparency in AI.
In between times, he will be looking for Strawberry Island, somewhere north of Green Bay.
Copyright 2024 Strategic News Service LLC "Strategic News Service," "SNS," "Future in Review," "FiRe," "INVNT/IP," and "SNS Project Inkwell" are all registered service marks of Strategic News Service LLC. ISSN 1093-8494 | |
