Many of you may have seen this posted in Facebook or other social media. It was posted to our internal Cisco Webex chat, and it cracked me up. What do you think of this? Is it always going to be easy to trick AI? In this case, you'd probably have to tell the AI never to list those sites, or at least anticipate the trick question. Now imagine a way to get around such limitations for every possible topic.
If there's a way to avoid anticipating every opportunity to trick the AI without having to code for every exception, I'd like to hear it.
Let's discuss. See more below for an example of hard-coding exceptions.
Snapchat now has an AI (I don't use Snapchat - I got this from a site called "Not The Bee"). There are several examples of how this AI is biased, but they're all politically sensitive, so I tried to pick one that is least likely to offend someone. I post it here as an example of how people can code the AI for exceptions:
