Does your AI assistant annoy you daily?
Subscribe to enjoy similar stories. For all the PhD-level intelligence they’re supposed to have, our AI assistants can be maddeningly bad at fulfilling requests smoothly. The productive time lost in trying to figure out what is going wrong is so significant that many companies are now evaluating whether the much-touted AI is delivering any real return on investment.
As a user of any large language model (LLM), you’ll face annoyances on an almost daily basis. From making things up, to seemingly refusing to obey instructions, to making mistakes that are hard to believe—it can be exhausting. There are serious hallucinations that can cause real damage in high-stakes situations, and sycophancy is another big time waster, other than being dangerous at times.
But one of the most common problems in working with an AI chatbot is the error loop. The AI just gets something incorrectly, and no matter how many times you try to correct or tweak it, it just does it wrong all over again. There’s no point arguing with an AI, but there are a few workarounds worth trying to end up with the result you’re looking for—and preserve your sanity.
A friend of mine recently spent the better part of two days trying to get the recording from a meeting with multiple speakers present, transcribed. He gave the job to an AI chat assistant, which happily began on the task. Hour after hour passed, but the chat assistant kept saying it was “still working on it".
But it probably wasn’t. Always look for the virtual power button that shows whether the chatbot is actually working on something. The chat assistant was extremely apologetic.
Read on livemint.com