Further, the problem with this example is it relies on a comparison against public data.
Most of these AI start failing pretty hard when you ask it to do the same task on something completely novel to it (like a company document). Sometimes they'll get it right. Other times, they'll spit out gibberish that's clearly some generic answer.
Most of these AI start failing pretty hard when you ask it to do the same task on something completely novel to it (like a company document). Sometimes they'll get it right. Other times, they'll spit out gibberish that's clearly some generic answer.