This is easy to circumvent, however, as more complicated/obscure natural language constructs can be used to confuse the bot (botfuscation). What you're suggesting is that in a few years we'll have bots that are perfectly capable of parsing natural language, understanding the problem asked, and solving it. This means they'll have to perform image recognition as well for the problems that have pictures. I believe this technology is still in its infancy and is nowhere near 100% accurate. But maybe you're more up on the research than I am?