We Don’t Ask AI For Answers!
We ask for “something that sounds like an answer”.
Often I would explain to my high school math students that learning and using math vocabulary could give them the appearance of understanding: “Did you ever notice that if you use specialized words, words that ‘sound good’, people assume you are far more knowledgeable than you actually are?” Likewise, I would say, if you mispronounce words or make obviously untrue statements you will be labeled ‘ignorant’ – which may be true!
Artificial Intelligence may be the most dramatic example of undeserved hype ever invented. Artificial surely, but the Intelligence is appearance. What has been hyped as ‘AI’ – mostly by those with a financial interest in its success – is known by those who developed it and are improving it as an LLM, or Large Language Model.
Peter Greene in ‘An Excellent AI Explanation’ quotes an AI specialist on what the ‘answer’ to your prompt is BY DESIGN!
Something that seems fundamental to me about ChatGPT, which gets lost over and over again:
When you enter text into it, you’re asking “What would a response to this sound like?”
If you put in a scientific question, and it comes back with a response citing a non-existent paper with a plausible title, using a real journal name and an author name who’s written things related to your question, it’s not being tricky or telling lies or doing anything at all surprising! This IS what a response to that question would sound like! It did the thing!
But people keep wanting the “say something that sounds like an answer” machine to be doing something else.
It’s good at generating things that sound like responses to being told it was wrong, so people think that it’s engaging in introspection or looking up more information or something, but it’s not, it’s only, ever, saying something that sounds like the next bit of the conversation.
So LLM’s operate on the algorithm which says: “Speak like you know what you’re talking about.”
It MUST be noted here that nowhere in the algorithm does the concept of ‘truth’ appear. A fault they are ‘working on’ but which unfortunately conflicts with a number of other desirable qualities they need this ‘intelligence’ to have. The technicians describe this as ‘AI is hallucinating’, which it could be argued is as close to ‘thinking’ as AI is capable.
How does AI sound so intelligent, so aware, so human! It’s because AI is extremely well-read: It reads what humans have written, which is by a large margin true – at least from some perspective. The fact that AI is spouting the work of others, the definition of plagiarism, is being adjudged now – and for quite a while into the future you can be assured.
The LLM does not quote the author(s), but rather puts words into a plausible order, paraphrasing. Is it stealing the idea? This brings up the argument at the center of this very important discussion; since ‘stealing’ itself is an ‘idea’, is an LLM capable of HAVING an ‘idea’.
If your desire is to ‘sound like you know what you’re talking about’ throwing around ‘intelligence’ and ‘ideas’ and ‘thinking’ could certainly prove useful, and the adherents of Artificial Intelligence must be given credit for coining the name. The truth is our understanding of intelligence and thought is extremely sketchy.
What is not sketchy is where the plausible sounding answers we get from LLMs come from, an algorithm that demands ‘something that sounds like an answer’. Depending on how successful the search through its database for intelligent humans who have expounded on such, the answer may or may not involve the truth. How is the LLM to distinguish ‘truth’? Does not the very concept of truth involve thought? Regardless of the truly metaphysical argument on whether machines can think, please know that nobody claims what they do currently is anywhere close.
AI; based on the principle of ‘sounding good’, incapable of understanding the concept of ‘truth’, capable of sounding ‘human’, incapable of emotion – very capable of ‘sounding’ like it is: What would be an appropriate role for such an entity to play?
Try TEACHER!
Our memories are truly too short for our own good if we can’t remember the pandemic and the failure of virtual schooling. Yet the hullabaloo about how horrible public school is and how rosy and beautiful AI is has a purpose – virtual teachers! The one inescapable fact is that virtual teachers will be one helluva lot cheaper that humans; although still a gold mine for the oligarchs with the AI. If you don’t see this coming you are either blind to what’s happening in education or you plan to profit from it.
This is not a ‘what’s to come’ but an ‘almost there’ because of the massive teacher shortage which no one is acknowledging, because no one is prepared to do anything about it. Check your own district, you’ll find an inordinate number of new hires; it would be interesting to know how many are under-qualified or unqualified. Even now a growing number of the hard-to-staff courses are only available virtually. Conversion to AI teachers will not be proposed as a choice, but as the only answer to a shortage that has been intentionally created.
In previous posts I have alluded to my thirty-six year career as a HS math teacher: My sixteen year career prior to that was a swimming coach, working with all ages. Thus my entire working life I have been a teacher; and God willing will be until I die. There is no more human act than teaching, for both the student and the teacher. The true danger is that we are eliminating the teacher from the learning equation without understanding this also eliminates the human learning that accompanies the teacher-student interaction; how to debate your point of view, the role emotion plays in learning, what social behavior is considered acceptable and what is not – and why. Not to mention why we as humans consider truth to be both desirable and important!
It is my belief that the only way to allow the teachers to teach our kids is to return control of education to the localities where it is given. This is its history and the liars will tell you it still is, but the control is the money, which comes from the fed with many strings attached. While the money is needed – desperately in most cases – the attached strings have made the cure worse than the disease.
We must all decry this bastardization use of a technology which does, and will, provide amazing capabilities in many scenarios; a use I can assure you every good teacher manages. We must keep the human in education by convincing our teachers to stay; paying them as the professionals they are. We must demand the freedom to teach our community’s students WITHOUT loss of federal funds!
Yes there is chaos.
But there also lies opportunity!
