This is yet another attempt to scare people into believing that computers are capable of taking the place of humans on their own volition, but they lack the capacity to think for themselves, they require a human to give them instructions to execute any task.
AI is artificial, but it has no intelligence whatsoever. Its intelligence is derived entirely by one or more human beings who provide instructions for the computer to execute. Now there are indeed humans who can produce remarkable instruction sets to emulate the appearance of intelligence, but it is the coders intelligence that you are actually seeing control the computer giving it the appearance of possessing intelligence. Therefore, if you play with it long enough you will provide the correct combination that will completely throw off the ability of the computer to perform the task to provide what was the desired result. At best it can only revert to a message stating that the desired result cannot be derived from the instructions presented, but the coder has to code to catch that event. If he hasn't, than the computer will produce an unpredictable result that is clearly not what was asked for.
...It seems that larger models acquire "emergent abilities" at this point(s).[19][63] These abilities are discovered rather than programmed-in or designed, in some cases only after the LLM [large language model] has been publicly deployed.[3] The most intriguing among emergent abilities is in-context learning from example demonstrations.[64] In-context learning is involved in tasks, such as:
reported arithmetics, decoding the International Phonetic Alphabet, unscrambling a word's letters, disambiguate word in context,[19][65][66] converting spatial words, cardinal directions (for example, replying "northeast" upon [0, 0, 1; 0, 0, 0; 0, 0, 0]), color terms represented in text.[67]
Finally....
Large language models by themselves are "black boxes", and it is not clear how they can perform linguistic tasks.
No one really knows how they work.
(I will respond only to erudite replies) (