Computers don't understand daydreams
On medieval medicines, machine learning, and one of the reasons to apply extreme caution to the use of computers for judgment
Cambridge University has unveiled the results of a remarkable project to catalog and digitally archive a large collection of medieval medical manuscripts. It's a fascinating undertaking, compiling 8,000 recipes for supposed cures from 186 different texts. Much of the text is unintelligible to the modern reader and speaker of English, but some transcripts are available.
■ Beyond the pure curiosity factor, it is a fascinating project because it reveals something important about human nature that distinguishes us (we, of "organic" intelligence, as opposed to the artificial type) from the many computers upon which we have come to rely so thoroughly in our modern age. The central matter is this: A computer doesn't understand time. Nor does it recognize changes over time as having value. And there's little reason to believe that digital computing machines ever will. They can measure the passage of time, of course, and they can be programmed to indicate time as a meaningful variable. Even a microwave oven can do that.
■ But when it comes to finding answers, the very nature of binary programming is that there is either a current answer or not. And if an answer is old, it is no longer the answer, and is probably to be discarded. That's adequate if a person is asking Google "Is there a tsunami warning in effect for Honolulu right now?".
■ What makes us distinct from the machines we build -- even the large language models and "machine learning" products we can design -- is that human beings can recognize the change in knowledge and understanding over time, and can hold conflicting, expired, or misleading information in mind when arriving at conclusions.
■ Decisions that were made by medieval practitioners of medicine (or perhaps it would be more accurate to call them "healing arts", since there was little of what we know as "science" involved) were often radically different from those made today, but those changes themselves have value. What led, for example, from mystical claptrap about "humours" to immunotherapy today?
■ Knowledge took a path from the past into the present, and choices that are objectively wrong today may have been conditionally right in the past. Likewise, the "right" answer to a question can easily change depending on whether we're asking a kindergartener, a high school sophomore, or a Ph.D. candidate -- what you learn as a child may be subjectively or conditionally right for what you can comprehend at the time, even setting the stage for you to be objectively accurate later on.
■ There are more than seven colors of the rainbow, but that doesn't make a child wrong when they recite the words behind "Roy G. Biv". A stage of knowledge may be important to get right, even if it will later be judged wrong.
■ It's important for the purposes of human judgment to understand what changed and why over a time period, and that's something you can only do if you have an organic understanding of time. There's little reason to believe that any digital machine ever will. That doesn't mean we should negate their use -- but it does mean we should develop a nuanced understanding of the theory of knowledge itself before making ourselves dependent upon the choices of any device that has never felt the pressure of a timed test or daydreamed while staring idly into space.
"But when it comes to finding answers, the very nature of binary programming is that there is either a current answer or not.But when it comes to finding answers, the very nature of binary programming is that there is either a current answer or not."
There is a very important qualifier you use there. *Binary* programming has that limitation, but as we move toward the potential of quantum computing, would that not allow, at least theoretically, multiple states to exist simultaneously... thereby replicating the way you speak of human, or biological, interpretation of facts and time?