AI review

Are the demos real? That's the problem we've always had. Are we watching a slideware or a software?

At the AWS London Summit, I found it interesting how demos highlight AI's power using pre-set prompts. However, there's no real way to validate the results on site. These are precise numbers with decimals; they look impressively accurate, but are they accurate?

Are you testing your numbers? Are you comparing the 'old ways' with the 'new ways'? How do you manage reconciliation? Would you keep a person making the same level of approximation in your team?

ps: to look accurate, numbers only need to be odd and with two decimals. That makes them too hard for our brain to verify on the fly, so we trust them.