ChatGPT is specifically bad at these kinds of tasks because of tokenization. If you plug your query into platform.openai.com/tokenizer, you can see that"egregious" is a single token, so the LLM doesn't actually see any "e" characters -- to answer your question it would have had to learn a fact about how the word was spelled from it's training data, and I imagine texts explicitly talking about how words are spelled are not very common. Good explanation here if this still doesn't make sense: twitter.com/npew/status/152590084988886…, or check out Andrej Karpathy's latest video if you have 2 hours for a deep dive: youtube.com/watch?v=zduSFxRajkE IMO questions about spelling or number sense are pretty tired as gotchas, because they are all basically just artifacts of this implementation detail. There are other language models available that don't have this issue. BTW this is also the reason DALL-E etc suck at generating text in images.

Unfortunately, the token-based approach that transformers use leaves them vulnerable to certain tasks like math, dissecting/modifying characters within a word, consistently tracking syllables etc. Check out Peter Welinder's Twitter post on how he taught GPT-3 to reverse a word; a similar technique might yield the results you seek. twitter.com/npew/status/152590084988886…

Check It -- : ) "gpt-3 was never real, openai has faked all its output by simulating it with a large language model" - Joscha Bach 16 May 2022 twitter.com/Plinz/status/15262687458023…