When the Cranes Cry

The crane has a symbolic resonance in Celtic mythology. A magician, assuming an elaborate pose—one eye open and one leg drawn up—was said to see into the otherworld, just as the crane itself moved from sky to land to water. But there is the other meaning of the word crane: the ancient lifting contraption that helped build Greece and likely had a role in Egypt and Sumeria before that. And now they protrude into the urban sky, raising up our buildings and even other cranes as we densify our cities. It was this mechanical meaning that Dan Dennett at Tufts chose to contrast with conceptual skyhooks, the unsupported contrivances that save protagonists in plays by dangling gods above the stage. For Dennett, the building crane is the metaphor we should apply to the mindless, simple algorithm of evolution. The algorithm raises up species and thus creates our mysterious ideas about meaning and purpose. No skyhooks or Deus ex Machina are needed.

Dennett passed away at 82 in Maine leaving a legacy as a public intellectual who engaged in the pursuit of reason throughout his adult career. He was committed to the idea that this world—this teeming ensemble of living matter—is intrinsically miraculous, built up by something dead simple into all the convolutions and perilous ideas that we now use to parse its mysteries. He was one of the Four Horsemen of the Apocalypse during the so-called New Atheism craze of 2008-2010, along with Richard Dawkins, Christopher Hitchens, and Sam Harris, but even then he was committed to the crane metaphor to displace these ancient skyhooks of belief rather than, say, a satirical impact-analysis of religion a la Hitchens.

There is another phrase that Dennett championed in Darwin’s Dangerous Idea: Evolution and the Meanings of Life: universal acid.… Read the rest

Begging the Pseudo-Question

 

I recently got involved in an “audiophile” online discussion thread replete with devious trolling, commenter bans, incivility—the works. I do this from time to time because raucous argumentation forces one to think in tactical and strategic ways that are not the norm in everyday life. I also learn new things. In this case, I went on several quests, hunting down papers on the ability of Chinese language speakers to disambiguate tones in Gaussian noise, how distortion artifacts impact our perception of spatialization in binaural audio presentations, and even Rayleigh wave detection by sand scorpions (I actually worked on a simulator for that as a late undergrad). One of the key disagreements in the thread was over the notion of “science.” There were several perspectives on this, with the first one being that science requires experimentation and therefore using scientifically-derived tools for investigating the performance of audio equipment does not amount to science. This is obviously a shrugger and a distraction. The other primary perspective is always that science is in constant revision and there may be new insights that prove this-or-that subtle hearing capability since human hearing is just sooooo amazing. We are sooooo amazing.

There’s a bit of a Two Cultures-like tension in this universe of audio equipment aficionados: while engineering and science brings them audio gear, they want it to be poetic and ineffable and the work of mastery based in genius rather than Fast Fourier Transforms. Graphs are boring. Listening is beautiful.

Part of the reason for the disagreement is clearly that we just don’t have shared meanings about concepts like science. We circle around them and try to triangulate using metaphors, analogies, and explore the logical consequences of limits and extensions to their meaning.… Read the rest

Inferred Modular Superparrots

The buzz about ChatGPT and related efforts has been surprisingly resistant to the standard deflationary pressure of the Gartner hype cycle. Quantum computing definitely fizzled but appears to be moving towards the plateau of productivity with recent expansions of the number of practical qubits available by IBM and Origin in China, as well as additional government funding out of national security interests and fears. But ChatGPT attracted more sustained attention because people can play with it easily without needing to understand something like Shor’s algorithm for factoring integers. Instead, you just feed it a prompt and are amazed that it writes so well. And related image generators are delightful (as above) and may represent a true displacement of creative professionals even at this early stage, with video hallucinators evolving rapidly too.

But are Large Language Models (LLMs) like ChatGPT doing much more than stitching together recorded fragments of texts ingested from an internet-scale corpus of text? Are they inferring patterns that are in any way beyond just being stochastic parrots? And why would scaling up a system result in qualitative new capabilities, if there are any at all?

Some new work covered in Quanta Magazine has some intriguing suggestions that there is a bit more going on in LLMs, although the subtitle contains the word “understanding” that I think is premature. At heart is the idea that as networks scale up given ordering rules that are not highly uniform or correlated they tend to break up into collections of subnetworks that are distinct (substitute “graphs” for networks if you are a specialist). The theory, then, is that the ingest of sufficient magnitudes of text into a sufficiently large network and the error-minimization involved in tuning that network to match output to input also segregates groupings that the Quanta author and researchers at Princeton and DeepMind refer to as skills.… Read the rest

Follow the Paths

There is a little corner of philosophical inquiry that asks whether knowledge is justified based on all our other knowledge. This epistemological foundationalism rests on the concept that if we keep finding justifications for things we can literally get to the bottom of it all. So, for instance, if we ask why we think there is a planet called Earth, we can find reasons for that belief that go beyond just “’cause I know!” like “I sense the ground beneath my feet” and “I’ve learned empirically-verified facts about the planet during my education that have been validated by space missions.” Then, in turn, we need to justify the idea that empiricism is a valid way of attaining knowledge with something like, “It’s shown to be reliable over time.” This idea of reliability is certainly changing and variable, however, since scientific insights and theories have varied, depending on the domain in question and timeframe. And why should we in fact value our senses as being reliable (or mostly reliable) given what we know about hallucinations, apophenia, and optical illusions?

There is also a curious argument in philosophy that parallels this skepticism about the reliability of our perceptions, reason, and the “warrants” for our beliefs called the Evolutionary Argument Against Naturalism (EAAN). I’ve previously discussed some aspects of EAAN, but it is, amazingly, still discussed in academic circles. In a nutshell it asserts that our reliable reasoning can’t be evolved because evolution does not reliably deliver good, truthful ways of thinking about the world.

While it may seem obvious that the evolutionary algorithm does not deliver or guarantee completely reliable facilities for discerning true things from false things, the notion of epistemological pragmatism is a direct parallel to evolutionary search (as Fitelson and Sober hint).… Read the rest

Sentience is Physical, Part 3: Now with Flaming Birds

Moving to Portland brings all the positives and negatives of urban living. A notable positive is access to the arts and I’m looking forward to catching Stravinsky’s The Firebird this weekend with the Oregon Symphony. Part of the program is a new work by composer Vijay Iyer who has a history of incorporating concepts derived from African rhythms, hip hop, and jazz into his compositional efforts. I took the opportunity this morning to read his 1998 dissertation from Berkeley that capped off his interdisciplinary program in the cognitive science of music. I’ll just say up front that I’m not sure it rises to the level of a dissertation since it does not really provide any significant new results. He notes the development of a microtiming programming environment coded in MAX but doesn’t give significant results or novel experimental testing of the system or of human perceptions of microtiming. What the dissertation does do, however, is give a lucid overview and some new insights about how cognition and music interact, as well as point towards ways to test the theories that Iyer develops during the course of his work. A too-long master’s thesis might be a better category for it, but I’ve never been exposed to musicology dissertations so perhaps this level of work is normal.

Iyer’s core thesis is that musical cognition and expression arise from a physical engagement with our environments combined with cultural situatedness. That is, rhythm is tied to a basic “tactus” or spontaneously perceived regular pulse or beat of music that is physically associated with walking, heartbeats, tapping, chewing, and so forth. Similarly, the culture of musical production as well as the history that informs a given piece all combine to influence how music is produced and experienced.… Read the rest

Entanglements: Collected Short Works

Now available in Kindle, softcover, and hardcover versions, Entanglements assembles a decade of short works by author, scientist, entrepreneur, and inventor Mark William Davis.

The fiction includes an intimate experimental triptych on the evolution of sexual identities. A genre-defying poetic meditation on creativity and environmental holocaust competes with conventional science fiction about quantum consciousness and virtual worlds. A postmodern interrogation of the intersection of storytelling and film rounds out the collected works as a counterpoint to an introductory dive into the ethics of altruism.

The nonfiction is divided into topics ranging from literary theory to philosophical concerns of religion, science, and artificial intelligence. Legal theories are magnified to examine the meaning of liberty and autonomy. A qualitative mathematics of free will is developed over the course of two essays and contextualized as part of the algorithm of evolution. What meaning really amounts to is always a central concern, whether discussing politics, culture, or ideas.

The works show the author’s own evolution in his thinking of our entanglement with reality as driven by underlying metaphors that transect science, reason, and society. For Davis, metaphors and the constellations of words that help frame them are the raw materials of thought, and their evolution and refinement is the central narrative of our growth as individuals in a webwork of societies and systems.

Entanglements is for readers who are in love with ideas and the networks of language that support and enervate them. It is a metalinguistic swim along a polychromatic reef of thought where fiction and nonfictional analysis coexist like coral and fish in a greater ecosystem.

Mark William Davis is the author of three dozen scientific papers and patents in cognitive science, search, machine translation, and even the structure of art.… Read the rest

Sentience is Physical

Sentience is all the rage these days. With large language models (LLMs) based on deep learning neural networks, question-answering behavior of these systems takes on curious approximations to talking with a smart person. Recently a member of Google’s AI team was fired after declaring one of their systems sentient. His offense? Violating public disclosure rules. I and many others who have a firm understanding of how these systems work—by predicting next words from previous productions crossed with the question token stream—are quick to dismiss the claims of sentience. But what does sentience really amount to and how can we determine if a machine becomes sentient?

Note that there are those who differentiate sentience (able to have feelings), from sapience (able to have thoughts), and consciousness (some private, subjective phenomenal sense of self). I am willing to blend them together a bit since the topic here isn’t narrowly trying to address the ethics of animal treatment, for example, where the distinction can be useful.

First we have the “imitation game” Turing test-style approach to the question of how we might ever determine if a machine becomes sentient. If a remote machine can fool a human into believing it is a person, it must be as intelligent as a person and therefore sentient like we presume of people. But this is a limited goal line. If the interaction is only over a limited domain like solving your cable internet installation problems, we don’t think of that as a sentient machine. Even against a larger domain of open-ended question and answering, if the human doesn’t hit upon a revealing kind of error that a machine might make that a human would not, we remain unconvinced that the target is sentient.… Read the rest

We Are Weak Chaos

Recent work in deep learning networks has been largely driven by the capacity of modern computing systems to compute gradient descent over very large networks. We use gaming cards with GPUs that are great for parallel processing to perform the matrix multiplications and summations that are the primitive operations central to artificial neural network formalisms. Conceptually, another primary advance is the pre-training of networks as autocorrelators that helps with smoothing out later “fine tuning” training programs over other data. There are some additional contributions that are notable in impact and that reintroduce the rather old idea of recurrent neural networks, networks with outputs attached back to inputs that create resonant kinds of running states within the network. The original motivation of such architectures was to emulate the vast interconnectivity of real neural systems and to capture a more temporal appreciation of data where past states affect ongoing processing, rather than a pure feed-through architecture. Neural networks are already nonlinear systems, so adding recurrence just ups the complexity of trying to figure out how to train them. Treating them as black boxes and using evolutionary algorithms was fashionable for me in the 90s, though the computing capabilities just weren’t up for anything other than small systems, as I found out when chastised for overusing a Cray at Los Alamos.

But does any of this have anything to do with real brain systems? Perhaps. Here’s Toker, et. al. “Consciousness is supported by near-critical slow cortical electrodynamics,” in Proceedings of the National Academy of Sciences (with the unenviable acronym PNAS). The researchers and clinicians studied the electrical activity of macaque and human brains in a wide variety of states: epileptics undergoing seizures, macaque monkeys sleeping, people on LSD, those under the effects of anesthesia, and people with disorders of consciousness.… Read the rest

Intelligent Borrowing

There has been a continuous bleed of biological, philosophical, linguistic, and psychological concepts into computer science since the 1950s. Artificial neural networks were inspired by real ones. Simulated evolution was designed around metaphorical patterns of natural evolution. Philosophical, linguistic, and psychological ideas transferred as knowledge representation and grammars, both natural and formal.

Since computer science is a uniquely synthetic kind of science and not quite a natural one, borrowing and applying metaphors seems to be part of the normal mode of advancement in this field. There is a purely mathematical component to the field in the fundamental questions around classes of algorithms and what is computable, but there are also highly synthetic issues that arise from architectures that are contingent on physical realizations. Finally, the application to simulating intelligent behavior relies largely on three separate modes of operation:

  1. Hypothesize about how intelligent beings perform such tasks
  2. Import metaphors based on those hypotheses
  3. Given initial success, use considerations of statistical features and their mappings to improve on the imported metaphors (and, rarely, improve with additional biological insights)

So, for instance, we import a simplified model of neural networks as connected sets of weights representing some kind of variable activation or inhibition potentials combined with sudden synaptic firing. Abstractly we already have an interesting kind of transfer function that takes a set of input variables and has a nonlinear mapping to the output variables. It’s interesting because being nonlinear means it can potentially compute very difficult relationships between the input and output.

But we see limitations, immediately, and these are observed in the history of the field. For instance, if you just have a single layer of these simulated neurons, the system isn’t fundamentally complex enough to compute any complex functions, so we add a few layers and then more and more.… Read the rest

One Shot, Few Shot, Radical Shot

Exunoplura is back up after a sad excursion through the challenges of hosting providers. To be blunt, they mostly suck. Between systems that just don’t work right (SSL certificate provisioning in this case) and bad to counterproductive support experiences, it’s enough to make one want to host it oneself. But hosting is mostly, as they say of war, long boring periods punctuated by moments of terror as things go frustratingly sideways. But we are back up again after two hosting provider side-trips!

Honestly, I’d like to see an AI agent effectively navigate through these technological challenges. Where even human performance is fleeting and imperfect, the notion that an AI could learn how to deal with the uncertain corners of the process strikes me as currently unthinkable. But there are some interesting recent developments worth noting and discussing in the journey towards what is named “general AI” or a framework that is as flexible as people can be, rather than narrowly tied to a specific task like visually inspecting welds or answering a few questions about weather, music, and so forth.

First, there is the work by the OpenAI folks on massive language models being tested against one-shot or few-shot learning problems. In each of these learning problems, the number of presentations of the training data cases is limited, rather than presenting huge numbers of exemplars and “fine tuning” the response of the model. What is a language model? Well, it varies across different approaches, but typically is a weighted context of words of varying length, with the weights reflecting the probabilities of those words in those contexts over a massive collection of text corpora. For the OpenAI model, GPT-3, the total number of parameters (words/contexts and their counts) is an astonishing 175 billion using 45 Tb of text to train the model.… Read the rest