I have been reading about “artificial intelligence”, and how it will supplant human knowledge workers (think law, and medicine, among others). I have my doubts.
There are several reasons for this. First of all, “artificial intelligence” is based upon “large language models”, wherein the developers feed into their databases tremendous amounts of, well, data. I have read that this takes the form of essentially sweeping up information from the internet (remember this part: it will be relevant) and programming their “AI” to address this information to provide a response to a query.
I have read that long, long ago, in galaxies far, far away, there was a Programmers Axiom, which went “GIGO”, or, Garbage In, Garbage Out. As I understand it, your computer output will not be of higher quality that that which you use to program it. If your program’s assumptions are bullshit, your output will be bullshit as well.
So, should you assume that the temperature observations recorded since the invention of the thermometer (Fahrenheit invented the mercury thermometer around 1714. according to worldhistory.org) reflect the planet’s temperature baseline, then you would notice, with alarm, the warming documented by such readings. Of course, should you factor in the Maunder Minimum, and The Year Without A Summer (1816 ), along with the historical record of Roman occupiers of Brittania growing grapes for wine (a feat not currently A Thing in 21st century England, due to the coolness of the present climate), you might come to different conclusions.
Adding in reports that present “AI” appear to some observers to have a left wing bias (feel free to define that however you want), well, impartiality does not seem to be A Thing.
I have read reports of attorneys presenting legal filings in court that have been written by, or with “assistance” of AI. To my point, it has been reported that these filings have been found to contain case citations that refer to cases that DO NOT EXIST. In addition, upon occasion, case citations refer to cases in support of one legal point or another, said citations that do exist, may DIRECTLY CONTRADICT THE POINT THE ATTORNEY IS ATTEMPTING TO MAKE.
I m certain that legal clients everywhere are filled with confidence at that thought.
To my internet point: Is there anyone of us who, coming across some gem of information on the internet, concerning some topic with which we have some familiarity, shake our heads with how ignorant and ill informed this foolishness is? How many of us have this response every day? Ladies and gentlemen, may I introduce the Gell-Man Amnesia Effect?
So, if I am creating an artificial intelligence, and populating that dataset with “information” from the internet, well, exactly what outcome should I expect? Again, GIGO. Bullshit in, Bullshit out.
Finally, speaking of medicine, a topic with which I have some familiarity, even assuming that the “AI” which takes my job KNOWS ALL with regard to Evidence Based Medicine, there is the issue that some patients did not read the evidence, and might not present with their (let us say) heart attack, in accordance with The Evidence. What will AI make of the relatively fit 60 year old man who complains of a sore throat for weeks on end, worsened by activity? Or the diabetic black woman who complains of tooth pain?
Will the AI inspect her teeth, and, finding them sound appearing, order a troponin and EKG?
After all, what evidence exists that a test for heart muscle injury will reveal information relevant to a dental complaint? Or that a recording of the electrical impulses of one’s heart, will assist in diagnosing and treating their dental disorder?
Finally, “evidence based medicine” assumes (veterans: care to chime in with what happens when we “ass-u-me”?) that all the relevant information-the studies-have, indeed, been published.
How is “AI” to know that studies exist, unpublished, that demonstrate findings contrary to The Narrative?
Should you wonder what that might look like, consider how JAMA and the British Medical Journal had been positively FILLED! with scholarly articles documenting the adverse reactions, contraindications, and adverse drug interactions of the mRNA covid vaccines!
I kid! I kid! Any such studies were unpublished. Perhaps they were erroneous, perhaps they contained facts that prudent clinicians might have wanted to know prior to recommending mRNA vaccines. We will never know, because THOSE studies never saw daylight, and therefore were never subjected to criticism, and scrutiny, and opportunities to be re examined in light of discovery of weaknesses in methodology or technique, so that the truth (NOT ‘my truth”, not “your truth”, but real, objective, third-person-verifiable truth) could be established and documented.
All this to point out that, in my view, “Artificial Intelligence” is NOT ready for prime time. Hell, it ain’t ready for Off Broadway. Indeed, in a show business metaphor, AI likely is not ready for High School Play Competition. (and, I mean no insult to high school thespians: some of their productions are smooth, engaging and entertaining.)
So, I sleep easily, considering that AI appears to be poised to have a significant role in national defense, in medicine, in law.
And, by “sleep easily”, I mean “sleep as if my cold wet bed had been filled with shards of broken glass soaked in pepper spray”, easily.