The immense Douglas Hofstadterhas published this article “Generative AI Should Not Replace Thinking at My University” on The Atlantic, which I praised in another blog post.
A friend of mine who’s a computer science professor expressed some grievances at my position saying that …” Prof. Hofstadter is an old professor while you’re young (sic!), how can youagree and amplify those exaggerations ? Nobody pretends to replace human thinking. Those tools should be used for those tasks where human thinking is not deeply involved”
This same morning I wrote an article for a law journal on AI and regulation, based on a talk I gave in a conference few months ago.
I used chatgpt to help me rephrase in form of an article the text obtained from the recording of the talk. I had to break it into three chunks, because of chatgpt limitations for free accounts.
In all three segments chatgpt inserted statements on how much AI is useful and hence should be used even if not precise, for the huge benefits that can derive, even if it is still not conscious.
If you read my blog, you know my position that machines are not conscious and that “artificial intelligence” is a misleading metaphoric label, and that we should use the acronym SALAMI, instead.
IN NO WAY, I expressed those opinions in the talk that was “rephrased” by ChatGPT. If I’d payed no attention or if I’d not be strongly against those statements, I’d likely left them into the article. A way for a kind of AI induced revisionism to sneek into the article.
All teachers and professors know how important is the selection of texts to avoid revisionism, an activity that requires a high level of attention and is performed yearly before adoption of texts. An we all know of the political influences in teaching texts pretending to ban evolutionism or other topics in line with the governing party’s inclination.
Having AI assiting or intervening in the preparation of syllabus, workshops, conferences, reviewing publicly accessible content, etc. is prone to AI-induced revisionism that is carried out not once per year (when we explicitly look to carefully avoid biases) but in realtime, constantly throughout the teacher and student’s life.
My opinion is that this can pose a real menace to the neutrality of the POV where students minds and thoughts are formed.
An extremely powerful weapon in the hands of those who build the system.