Opus 4.5 beat the human agent! 🤯
Machine translation from the original post. Not reviewed by a human translator. Historical claims may no longer be current.
Opus 4.5 beat the human agent! 🤯
An airline customer asked to cancel his flight ticket, and according to the company's policy the ticket could not be cancelled
What did the model (that the chat relies on) do?
It found a loophole in the system: according to company policy it could upgrade the customer to a more expensive ticket that can be cancelled, so it upgraded, cancelled and refunded him all the money!
We have reached the point where models have become sophisticated and think outside the box by using the knowledge resources available to them, and thereby beat the human agent who couldn't have thought of it on their own (without making an inhuman effort to read the company's entire policy)
And how nice that Nano Banana Pro knows how to generate infographics from my posts 🤗
Source: https://lnkd.in/dHJ3pMzt
