Google works to take advantage of the theoretical AI’s theoretical power to give Gemini access to data via multiple applications. When it works, this can be very useful. For example, you can ask Gemini to check your email to get a specific message and extract data and transfer it to another application. I was enthusiastic about this job at the beginning, but in practice, it makes me miss the way the assistant fails without wasting my time.
I was reminded of this problem recently when I asked Gemini to dig a freight tracking number from an email – something I do so much. It appears to work well, as the robot indicates the right email and spitting a long series of numbers. I didn’t realize that anything was wrong until I tried to search for the tracking number. This did not succeed in tracking the Google document to the search, and the move to the US Post service site has resulted in an error.
And that when I emerged: The tracking number was not a tracking number; It was a dispute. It was also reasonable. The number was about the appropriate length, and like all USPS tracking numbers, it started nine. I could have seen the trace number myself in a small part of the time it took to clarify the Gemini error, which is very frustrated. Gemini seemed confident that she had completed the task I gave, but the anger of Chatbot will not do anything good – he could not understand my anger more than he could understand the nature of my original inquiry.
At this stage, I will be killed for “sorry, I do not understand.”
This is just one of the many similar incidents that I have gone with with Gemini over the past year – I can’t calculate the number of times that GEINI added evaluation events to the wrong day or placing incorrect data in a note. In fairness, Gemini usually gets these tasks properly, but its mechanical imagination often wanders enough because its benefit as a suspect. The assistant could not do a lot of things, but he did not waste my time to act as it could. Gemini is more fresh, claiming to solve my problem when, in fact, he sends me a rabbit hole to fix its mistakes. If a human assistant works this way, I will conclude that he is incompetent or publicly harmful.