An uncharitable interpretation of Donald Hoffman might view him as claiming that a completely deluded agent has an advantage over an agent with some grip on reality. I hope he does not actually take this extreme view.
Hoffman likes to use the graphic user interface of a PC as an analogy. Computer users do not wish to know everything about what is happening in their PC. They find it much easier to deal with an abstract representation that in some ways corresponds to the internal workings of the computer, and in other ways is just made up to make sense to the user. An icon's image or location on the desktop says nothing important about it at the deep level. These are just handles that the user can manipulate for reasons that have nothing to do with internal operations of the computer. But there are some important connections.
We can interpret his metaphor to cast the conscious mind in the role of the user, reality as the computer hardware, and the unconscious mind and perceptions as the GUI. Some aspects of what we perceive correspond indirectly to reality, others result from processes within the unconscious. Does this get us to agreement with Hoffman, or did he overstate his case?
Hoffman does not explain his work in a way that makes it easy to agree with him. But we should keep in mind what we are agreeing with and what we are disagreeing with.
Hoffman presents some very provocative results. He describes them verbally in absurd terms. I would summarize his conclusions as "evolution favors deluded agents over agents who perceive reality correctly" and "consciousness is the foundation of reality". These both seem ridiculous at first glance.
But Hoffman does not give us sufficient information to judge him fairly. These conclusions summarize the results of a computer simulation and a mathematical theory. We must understand those before we can give Hoffman a charitable interpretation. Does Hoffman interpret his models and his application of their terms to our reality accurately? Without the details, we can't know, and Hoffman does not provide much detail in his interviews for the popular press.
Does the following example parallel Hoffman's idea? Humans evolved a response to danger in our ancestral environment. Evolution favored strategies that may generate more non-fatal errors so long as they tend to avoid fatal errors more reliably. So we respond to dangers that do not actually exist more often than we ignore real dangers (in the ancestral environment). The costs of the two sorts of error differ; and heuristics that sacrifice one measure of accuracy to avoid the more costly sort of error can make sense. We can imagine simulating this idea and having nervous nellies outcompete others who more accurately predict danger but who make fatal mistakes more often.
Now imagine that Hoffman's simulation does something similar, but only allows perception to vary. All agents use the same strategy for avoiding danger, but have different abilities to perceive reality accurately. The perceptions that evolve may also reflect the difference in the cost of error. Hallucinating danger more often would be an acceptable cost, if it is counterbalanced by a sufficiently lower chance of ignoring potentially fatal danger.
I found this counterintuitive at first. I was tempted to think that more accurate perception always gives us an advantage over less accurate, and we should analyse any difference in strategy separately. But this artificial distinction misleads us. Scientists no longer think the brain works that way; The line between unconscious thought and perception is fuzzy. (E.g. the sensitivity of the rods and cones in the eye respond to our emotional state.)
Even if we assume a separation between perception and strategic response, the speed and cognitive cost of perceptions could impact fitness in addition to their accuracy. Increases in accuracy may trade off with these other factors. We can't just assume that more accuracy always improves fitness. We have to count the cost of the accuracy in terms of other factors sacrificed.
So I can't be sure that I agree with Hoffman, but I am not sure I disagree either. Can we interpret what happened in his simulation unambiguously to support his conclusion? How far from accuracy could this process lead? I don't know.
Hoffman's mathematical model of a consciousness-based universe seems even more difficult to criticize from a simple verbal summary. He seems to say that he has some math that derives quantum reality from more primitive entities that can be interpreted as consciousness. It is up to mathematical physicists to say whether the math works. Can we interpret the math without understanding it?
Is the math correct? Does the interpretation work? We can view Hoffman's argument as a reductio ad absurdum. This gives us a choice; If the intervening logic works, we either accept the result or criticize the assumptions. Which assumptions specifically must we reject? Hoffman has not really made this clear in his popular summaries. We would need to read Hoffman's serious papers to find out what he assumes.
No comments:
Post a Comment