A similar debate has been sparked by Github's Co-Pilot a few months ago and those who haven't made up their mind yet, might be interested to read a comprehensive recap of some opinion pieces here.
Personally, I am not overly concerned about GPT3 in particular. Obviously, it can solve classroom assignments and high school teachers as well as university lecturers will have to worry about fair grading more than ever. On the other hand, every student should be aware that the purpose of homework is exercise for them. Evidently, one doesn't get in shape and grow abs just by watching fitness influencers on Instagram, and neither does letting GTP3 write your term paper help one's education. However, the risk of cutting corners & cheating should be balanced against the fact that having a personal tutor, who can explain things differently or in more detail, can also be very supportive.
That being said, I entirely depend on other people's knowledge for my entire life. I search at least a 100 times a day information on the internet and use many products daily. By buying and using those items, I profit from all the know-how that went into their manufacturing, without even being aware of all the development effort. I don't bother about the contents of my toothpaste and the regulations governing cosmetic ingredients, I have no clue about oil drilling, farming and agriculture, the challenges of running and maintaining a power grid, about logistics and supply chain management etc. I just trust that some people anywhere in the world responsibly take care of all of that for me. In this sense, GTP3 will yet become another product people can make use of, while blissfully ignoring the complexity that enabled it in the first place.
What bothers me more than the technology leaps are the shortcomings of our human brains and the fragility of our civilizations. GPT3 now closely resembles what Douglas Adams in 1987 envisioned in his novel Dirk Gently’s Holistic Detective Agency: A computer program that is able to compellingly justify any decision retroactively. This and other AI-generated content (like Deepfakes) enables more sophisticated manipulations than were possible ever before, yet also previous, way more primitive methods were sufficient to persuade people into fascism or to fall for totalitarian ideologies.
David Hume was already aware that our brains are biased to preferably trust information provided by our friends or peers and to cherry-pick whatever information can justify our emotional and moral preferences. Benjamin Franklin as well noted that "So convenient a thing it is to be a reasonable creature, since it enables one to find or make a reason for everything one has a mind to do". The connectivity and speed of the internet is in my opinion the much bigger risk than AI itself: Social networks in particular have potentiated the options for allurement and the possibilities for delusion, and it seems that even people in liberal democracies could forfeit a lot of societal progress out of wounded pride, stubbornness, vengefulness and other emotional reasons. As humanity, it is our eternal duty to strive for unity, cooperation & cohesion and actively resist dichotomous thinking and groupish sentiments. The ability to build communities and to enable cooperation across large groups in unprecedented complexity is what really sets us apart from all other species on this planet, yet it is also the most fragile asset of ours. Many civilizations have already collapsed in the course of history and solving the challenges of this century peacefully will be much harder than training AI to write a function or an assay.
Therefore, I personally hope that the 21st century will once not be remembered as the century of AI and technology, but predominantly as the century of flourishing social and political sciences, which eventually fostered global cooperation and democracy despite enormous global challenges.