Answering your questions about business and technology
If these AI tools are so advanced then why do they keep getting simple, obvious things wrong?
~ Darren
Hi Darren,
That’s difficult to answer without specific examples.
That also helps to feed my answer. One thing that we see a lot is that people don’t give the AI tools (chatbots, agents, etc.) enough information to carry out the task(s).
It can be challenging for people because they are told to “talk to AI like it’s a person” but at the same time, when you start leaving out information that another person would know, the AI tools suddenly start to “hallucinate” more often.
We tell people to think about how to make the implicit, explicit. In other words, if you don’t get back a response that you want from an AI tool, go back to your prompt and find the things that you left out because they were “obvious”. They might be obvious to a person, but not obvious to a language model. If there are other possibilities, the models may choose them even if a person would never consider them.
I realize this is vague, but this is, ironically, why I opened by mentioning that I don’t have enough information…
Hope that helps
~ Damien