Donald Hoffman, a cognitive science researcher, has been advancing a theory he calls the Interface Theory of Perception. It challenges the widely held claim (one that I've made myself in philosophical arguments in this forum) that evolution promotes the development of sense organs and brains that result in generally veridical perception of reality. (At least, veridical perception at the scale of the organism itself, not necessarily at smaller or larger scales of no direct use or concern, such as the earth's movements through space or the atomic makeup of materials.) Basically, the standard claim is if you don't accurately perceive the lion, or you don't accurately perceive the possibility of a lion you don't see because a boulder limits your view, you get eaten and don't pass on your misperceiving genes.
What appears to set Hoffman apart from eons of philosophical speculation from Plato's Cave to Bishop Berkeley to Hume, is that he's done some investigations of the question "does evolution lead to veridical perception?" using simulated evolution of simulated perception in simulated worlds. He claims that the results have been clear and consistent: that "perception of reality goes extinct." Organisms whose perceptions are accurate succumb to competition from organisms whose perceptions are tuned directly to fitness.
(I don't have a good understanding of what some of those phrases e.g. "tuned to fitness" really mean in the context of the actual experiments he ran. I might have to buy his book to find that out. It's possible that "tuned to fitness" is a cheat, an abstraction that can be built into a simulation of a world but could not be accomplished by a system actually functioning within that world. As a crude example: clearly a perceptual system that would automatically know the most beneficial choice to make in any situation with no sensory input at all—call it an oracle—would out-compete one that had to imperfectly sense and perceive the present state of reality and figure out the best choice from that. But such an oracle is probably not possible, and near-veridical perception could be the next best thing, superior to a system that had to figure out the best choice from an inaccurate perception. So if Hoffman's tests allow for some possibly much subtler form of oracle, it would undermine his findings.)
Hoffman's favorite analogy, after which he appears to have named his "ITP" hypothesis, is to compare the world we perceive to a GUI on a computer. The characteristics of the icons on the GUI desktop—their shape, color, position, etc.—do not represent any actual characteristics of the files in the computer; they merely present the underlying reality of the computer files in a useful way. Space, time, and objects are components of our own naturally-evolved fitness-maximizing interface with reality, while reality's actual nature could be completely different and presently unknown. Reality could be discoverable by further research once the counterproductive falsified hypothesis of veridical perception is discarded. He further suggests that the nature of consciousness cannot be understood in terms of perceived objects (e.g. neurons which are just more icons in the interface) or fundamental particles (which are just the pixels of the interface) but might require scientific revelation of the underlying reality instead.
I think he's wrong*, but I'm really curious about how his simulations actually worked. Even if he's wrong, the question of why those experiements produced the results he says they did could be very interesting and revealing about cognitive evolution. Also, maybe he's not wrong, which would also be really interesting.
Here's his TED talk: https://www.ted.com/talks/donald_hof...ge=en#t-655679
His 2019 book is provocatively titled "The Case Against Reality: Why Evolution Hid the Truth from Our Eyes." (The publisher has chosen to highlight an endorsement from Deepak Chopra in the publisher's copy on Amazon. :p)
Is anyone familiar with this, or has anyone run across it before? Any thoughts?
*The question "how does perceiving neurons and the functioning of neurons, when using microscopes and sensitive electronics and advanced imagining technologies, increase fitness?" looms large. Our evolving surviving ancestors did not perceive neurons, they perceived squishy meat with no obvious purpose. So why do we perceive them now? How did they get added to our present perceptual "interface" unless they correspond to characteristics of reality?
ETA: Oh, and I've posted this in R&P instead of SMMT because it might be of more interest to regular posters here. To me what's notable here is the science, but until we can get more details of those investigations the discussion will probably tend to drift toward the philosophical.
What appears to set Hoffman apart from eons of philosophical speculation from Plato's Cave to Bishop Berkeley to Hume, is that he's done some investigations of the question "does evolution lead to veridical perception?" using simulated evolution of simulated perception in simulated worlds. He claims that the results have been clear and consistent: that "perception of reality goes extinct." Organisms whose perceptions are accurate succumb to competition from organisms whose perceptions are tuned directly to fitness.
(I don't have a good understanding of what some of those phrases e.g. "tuned to fitness" really mean in the context of the actual experiments he ran. I might have to buy his book to find that out. It's possible that "tuned to fitness" is a cheat, an abstraction that can be built into a simulation of a world but could not be accomplished by a system actually functioning within that world. As a crude example: clearly a perceptual system that would automatically know the most beneficial choice to make in any situation with no sensory input at all—call it an oracle—would out-compete one that had to imperfectly sense and perceive the present state of reality and figure out the best choice from that. But such an oracle is probably not possible, and near-veridical perception could be the next best thing, superior to a system that had to figure out the best choice from an inaccurate perception. So if Hoffman's tests allow for some possibly much subtler form of oracle, it would undermine his findings.)
Hoffman's favorite analogy, after which he appears to have named his "ITP" hypothesis, is to compare the world we perceive to a GUI on a computer. The characteristics of the icons on the GUI desktop—their shape, color, position, etc.—do not represent any actual characteristics of the files in the computer; they merely present the underlying reality of the computer files in a useful way. Space, time, and objects are components of our own naturally-evolved fitness-maximizing interface with reality, while reality's actual nature could be completely different and presently unknown. Reality could be discoverable by further research once the counterproductive falsified hypothesis of veridical perception is discarded. He further suggests that the nature of consciousness cannot be understood in terms of perceived objects (e.g. neurons which are just more icons in the interface) or fundamental particles (which are just the pixels of the interface) but might require scientific revelation of the underlying reality instead.
I think he's wrong*, but I'm really curious about how his simulations actually worked. Even if he's wrong, the question of why those experiements produced the results he says they did could be very interesting and revealing about cognitive evolution. Also, maybe he's not wrong, which would also be really interesting.
Here's his TED talk: https://www.ted.com/talks/donald_hof...ge=en#t-655679
His 2019 book is provocatively titled "The Case Against Reality: Why Evolution Hid the Truth from Our Eyes." (The publisher has chosen to highlight an endorsement from Deepak Chopra in the publisher's copy on Amazon. :p)
Is anyone familiar with this, or has anyone run across it before? Any thoughts?
*The question "how does perceiving neurons and the functioning of neurons, when using microscopes and sensitive electronics and advanced imagining technologies, increase fitness?" looms large. Our evolving surviving ancestors did not perceive neurons, they perceived squishy meat with no obvious purpose. So why do we perceive them now? How did they get added to our present perceptual "interface" unless they correspond to characteristics of reality?
ETA: Oh, and I've posted this in R&P instead of SMMT because it might be of more interest to regular posters here. To me what's notable here is the science, but until we can get more details of those investigations the discussion will probably tend to drift toward the philosophical.
via International Skeptics Forum https://ift.tt/2zwAcUB
Aucun commentaire:
Enregistrer un commentaire