Eps 20: My argument about whether the brain is a computer or not
| Host image: | StyleGAN neural net |
|---|---|
| Content creation: | GPT-3.5, |
Host
Franklin Steward
Podcast Content
We contend that the ongoing brain-computer metaphor controversy is really just a disagreement about semantics, since the brain is either literally a computer, or it is obviously nothing like a computer at all, depending on ones definition. Not everybody believes neural networks confirm the view that our brains are like computers. A handful of cognitive scientists--most notably Antony Chemero at the University of Cincinnati, author of the book "Radical Embodied Cognitive Science" --now completely reject the notion that human brains function like computers.
Unfortunately, the majority of cognitive scientists and neuroscientists who study the brain are still operating on -- and even venerate -- this limited brain-as-computer model. This has led to efforts to try and design computer architectures that more closely mimic the brain. In addition to practical applications of brain science to computers, attempts to develop computer electronics that mimic brain circuitry could result in better understanding how the brain itself really works, on the most basic level. Neuroscientists can, and are, studying how the brain implements different algorithms, how it handles computationally intractable functions, and how it does things such as storing, compressing, or conveying information.
While tech engineers understand the parts that go into making up a computer easily, cognitive scientists have no idea about the nitty gritty of how the brain does even the most basic tasks, like memory storage, learning language, or identifying objects. A closer look at the workings of the brain exposes a few more fundamental flaws with the theories behind computation. If parts of the brain that we consider to be essentially human--not only intelligence, but also self-awareness--are emergent properties of the brain, not function-specific ones, as seems probable, then computational theories of the mind become weaker still. If we accept a definition of a computer as based on a usage outside of computing , the brain is not a computer, and computers serve, at best, as a weak metaphor for just one narrow slice of human cognition.
A corollary is that brain behavior is not exhaustively described by a set union of all of its cognitive capacities, and thus, even if all cognitive capacities turned out to be computable, brains are not simply computers. If, as mathematical physicist Roger Penrose has suggested, humans show an ability to overcome some of these fundamental limits to computation, then the brain would have to interface with systems existing outside of a logic-algorithmic universe. Now, that does not necessarily mean an artificial human brain is not possible -- it is just that programming one would be far more like embedded systems programming, rather than computer programming. It is far less clear that there is any easy path towards a computer geared toward the kinds of emergent properties that make a human brain different.
Related might suggest that the appearance of novel, insightful metaphors about the brain and its operation is contingent upon future technical discoveries, in the same way as plumbing, the telephone exchange, or computers. As the buildings fuse, the distinction between a digital computer and a brain would be a matter of technology, not of nothing in common. If we ever reach the point of understanding pieces, we may be able to deploy a computing system much more resembling the brain than anything that is out there now. To even start imagining how this could work in practice, we will need both a knowledge of neural functions far beyond anything we can imagine at present, and it will take an unimaginably large amount of computing power and simulations that accurately simulate the structures of the brains under study.
The mathematical physicist Roger Penrose has suggested that quantum-mechanical processes may exist that could yield answers to questions in ways no current computer model--classical or quantum--would permit, and that the human brain could potentially interface with them via the tiny structures, or microtubules, within neurons. Marcus goes on to suggest that neurosciences challenge is reverse engineering the brain, in much the same way one would examine a computer, looking at its components and their connections in order to decipher how it works. A computer is capable of processing information exponentially faster than the human brain. Research has shown that a brain has higher computing efficiency by orders of magnitude compared to an electronic computer.
Some brain scientists have argued that insights about the brain would come from studying behaviors rather than by comparisons to computers. They argue that what we want most to learn about the brains functions, like the mechanisms of consciousness and the nature of sleep, are emergent properties, and therefore cannot be learned from us if we continue to attempt to find insights into the brain through the lens of relevant computer components. Cognitive scientists have told us almost nothing about why these parts of the brain lit up, or why this would matter. Although my colleagues and I are not literally suggesting the brain is a field-programmable gate array, our hypothesis is that the brain may similarly be composed of a highly organized collection of basic building blocks, like the computational primitives for building sequences, retrieving information from memory, and routing information among various locations within the brain.
The failure suggests that, while there is appeal in the metaphor of the computer, and that brains indeed do process information and somehow represent the outside world, we will still have to make substantial theoretical discoveries in order to progress.