6 more things AI Can’t Do
Learning when to use AI - and when not to - is increasingly part of being a teacher… and a human!
AI has been the hot topic in education (and elsewhere), resulting in some hilarious memes:
Each of these memes alludes to the limitations of AI which can result in humorous mistakes…. Or much more dire consequences. As educators, we must clearly understand the capabilities of AI, but also its constraints.
To that end, I share a continuation of my last blog post…
6 more things AI Can’t do
6. Chatbots make inexplicable logical mistakes.
This may be counterintuitive, but it seems Chat GPT is bad at following logical series of rules.
Here are a couple of surprising examples. Chat GPT tried unsuccessfully to play chess – see the ridiculous results in this Youtube short:
Chat GPT Fails at Chess!
(Chess fans, click here for the longer, hilarious version)
Users on this Reddit Chess forum had similar experiences, with Chat GPT breaking rules and making inexplicable errors - always with maximum confidence. One user reported Chat GPT ended up resigning from their game with the following message:
"I'm sorry, as a language model AI, I don't have the ability to keep track of the game state, it would be best to continue playing with a chess software or with a human opponent. I apologize if my mistakes caused inconvenience, please let me know if there is anything else I can help you with."
I had a similar experience when I asked Chat GPT to write me some sentences with indirect objects, and it substituted direct objects. This surprised me, as this grammar rule is fairly objective. Read more about my prompts and Chat GPT’s responses here.
And, a very simple example: When I asked it to grammar check a French sentence, ChatGPT prompted me to add Spanish punctuation.
I’ve never had a French student at any level make such a ridiculous error.
7. AI Can’t Make Moral or Ethical Judgments.
“When I asked it to grammar check a French sentence, Chat GPT prompted me to add Spanish punctuation.”
Various organizations have experimented with using AI Chatbots to provide medical advice. These projects were epic failures; the AI doctor gave terrible medical advice including advising patients with eating disorders to lose weight (source) and advising test patients reporting pain to commit suicide! (Source) Researchers concluded AI is not ready to give unsupervised medial advice and pulled the plug on the chatbot doctors. The same could be extended to education and other areas requiring both technical expertise and good judgment. As Forbes concluded, “Chatbots and artificial intelligence might not be ready for prime time. Especially when lives are at stake.” (Source)
8. AI Can’t Provide Physical Assistance
This may seem obvious, but it bears mentioning. My job as a teacher includes administering first aid, helping students organize their binders, giving high fives and fist bumps, supervising unstructured time in common areas, and showing a new student where her class is. AI can’t replace any of that.
9. AI Can’t Provide Emotional Support
As with physical assistance, I also spend a lot of my day listening, providing sympathy, and coaching students in managing their emotions. Chat GPT can’t give any real help to a student whose dog died, to a student who is angry with a classmate, to a student who doesn’t know how to study, or to a student who feels like they have no friends. I model mature behavior and give free counseling on a daily basis. We learned during lockdowns that teaching is more than uploading knowledge into students’ brains, and research is proving that face to face conversation fires a lot more neurons than Zoom interactions. (read the study here) AI can’t replicate the human touch I bring to the classroom.
10. AI can’t challenge, push back, or disagree with you
Human relationships are messy and complicated. Avoid all this with clean, sterile, virtual relationships! Researchers are alarmed as young people increasingly opt out of human society through (for example) AI girlfriends who can be designed to order and are consistently enthralled with your affection. The artificial girlfriend, it turns out, is much more accommodating and less demanding than the flesh and blood version. (read more here). Of course, the relationship is an illusion which can only lead to loneliness and isolation.
Part of being human is learning to negotiate, manage disagreement, and interact with different personalities and points of view. AI makes it possible to sidestep that challenge, but in doing so we forego the opportunity to grow into better versions of ourselves.
That’s not to say AI doesn’t have its place… but it has to be managed properly.
11. AI Can’t Survive a Water Spill
If, at any point, you feel threatened or worry that AI has become too powerful… dump a glass of water on it, watch the sparks fly, and bask in the knowledge that as a teacher, your skill and expertise will always be an essential part of society.
How do you address the limitations of AI? Share in the chat!
