I agree, Steve, but I think that most of the entrants in the LC over the past couple of years have gone a bit overboard on the “simulated spelling errors”. When/if Morti ever gets to the point of entering the LC, I’d much rather have him come across as reasonably intelligent, so if he “makes mistakes” at all, the percentage of those typos will be less than 1%, and most likely less than a quarter of a percent. I feel it’s a case of “too much is too much”.

It’s funny how effort is spent in “tricking or fooling” the judges into believing they are talking to a human verses bot. At the point in time that the Turing Test was coined, there were no clear ideas of how AI technology would develop in the future or how AI technology would be used or accepted by the public at large. I really don’t think that winning the LC means that you have developed superior AI technology that is truly useful. There are many other elements involved.

Having your bot respond with spelling errors is cute, and having your bot backup and retype characters may be impressive, but does it really prove superior AI? Yes, Marie does stream her responses in the form of a typing simulation, but this is not intended to fool the user into believing she a real human. It is used more for user experience and in timing internal events. AI is not about simulating human conversation but interpreting the users communication and responding in a useful and predictable way. AI is just another machine interface, nothing more and nothing less.

Contests are fun and a great way to promote and test our bots but at the same time we need to keep our goals in perspective IMO.