Monday, July 7, 2014
Turing bots are garbage: a short conversation with "Cleverbot"
So I was looking around for a good online Turing bot to chat with.
I found Turinghub.com, which seemed to have an interesting idea: you chat with multiple entities, and try to figure out which is human and which is a bot. That seemed an interesting idea, but as it turns out there are no humans online, and the two entities that I did chat with were both quite obviously bots. I nailed it in something like 3 sentences.
Why was I so sure?
Because they had no opinion about Lana Del Rey.
No really. If you're connected to the internet, then you have an opinion about Lana Del Rey.
Here's one way to easily determine if you have a bot or not on the other end: ask it for its opinion on something it should definitely have a strong opinion of. The usual bot will be programmed to change the subject - not respond in any way that indicates even a semblance of sentience.
For example, here's a screenshot of my very very short convo with Cleverbot, another "Turing-winning" bot which supposedly got a lot of press a few years ago by being very convincing on the Turing test:
Really, when every single statement in a conversation is responded to with a gross tangent, it's a sign that the programmers aren't even trying. This is just the goddamn Eliza script, 50 years on.
Honestly. Do these people think that human beings just talk at each other? Is there no such thing as responding to what another person says?