November 16, 2011 at 13:29Slow but sure Cleverbot bypasses Suzette
Not so long ago, in 2010, the Suzette chatbot won the Löbner Award as part of the Turing Test. However, competitors are not asleep!
Not so long ago in India, a festival Techniche, where the chat bot Cleverbot did nedyuzhy success, actually passing the Turing test. The test was conducted as follows: people chatted with people and a bot. As a result, people took people for people who accepted 63%, and Cleverbot'a - 59%. If we take into account that passing the Turing test is considered to be a result of 50% or higher, then we can say that this is a victory!
After a short conversation, this little devil was almost able to convince me that the bot is me, not him :)
The more the bot communicates with people, the smarter it becomes, daily increasing its “intellectual power”:
The only serious problem, in my opinion, in modern bots is the inability to maintain one topic throughout a long conversation. But at the pace that has been observed recently in the field of their development - I think that this is not far from this.
The second problem, but more general, is the Turing test itself. It is starting to become obsolete. Raising the bar in it no longer makes sense, otherwise people themselves will not be able to pass it. It’s time to change something and create a new chatbot rating system.
Not so long ago in India, a festival Techniche, where the chat bot Cleverbot did nedyuzhy success, actually passing the Turing test. The test was conducted as follows: people chatted with people and a bot. As a result, people took people for people who accepted 63%, and Cleverbot'a - 59%. If we take into account that passing the Turing test is considered to be a result of 50% or higher, then we can say that this is a victory!
After a short conversation, this little devil was almost able to convince me that the bot is me, not him :)
The more the bot communicates with people, the smarter it becomes, daily increasing its “intellectual power”:
The only serious problem, in my opinion, in modern bots is the inability to maintain one topic throughout a long conversation. But at the pace that has been observed recently in the field of their development - I think that this is not far from this.
The second problem, but more general, is the Turing test itself. It is starting to become obsolete. Raising the bar in it no longer makes sense, otherwise people themselves will not be able to pass it. It’s time to change something and create a new chatbot rating system.