The new conversationalists on the block
WE humans are a narcissistic bunch.
We create in our own image, anthropomorphise animals and inanimate objects, and idolise those which best reflect humanity’s traits back at us.
From the tale of Pygmalion, the ancient Greek sculptor who fell in love with his own sculpture, to the modern day science fiction film ‘Ex Machina’ (a movie I highly recommend), in which an artificially intelligent robot dupes both her human creator, and the man hired to test her, our fascination with humanesque constructs has stuck with us from the beginning of recorded history.
In recent times, this obsession has driven computer scientists to attempt to develop artificial intelligences by mimicking the way the neural networks in our brains process information.
While not entirely new, interest in chatbots (artificial intelligences designed for the purposes of communicating with a human user in the most human-like way possible), have certainly increased markedly.
And with the renewed interest, developers have been hard at work infusing the ‘human factor’ into their chatty apps.
The race to create realistic chatbots arose soon after the first digital computers were developed, as did the tests for measuring machine intelligence.
Developed by Alan Turing, the ‘ Turing Test’ involves a human evaluator who is tasked with engaging in text based conversations with both an artificial intelligence, and another human who acts as a foil.
After five minutes of conversation with each entity, the evaluator selects the one they think is human.
If the evaluator can’t distinguish between the two, or picks the incorrect one, the program is deemed to have passed the test.
In 1990, American inventor Hugh Loebner introduced the Loebner Prize, adopting the principles laid out by Turing in 1950.
The annual competition grants prizes to chatbots who manage to fool the judges or present the most human-like conversational techniques.
Just as in the ‘ Turing Test’, the Loebner prize also involves a human who converses with a judge at the same time as the A.I via a text messaging interface, and the judge must distinguish between the two.
Last year, the chatbot Mitsuku took out the prize, which she had won in 2013 as well.
Mitsuku’s success, a chatbot developed by Steve Worswick, is largely due to a broad and interwoven database of common objects as well as their characteristics and relationships to each other.
Worswick looks at the bot’s conversational logs to identify topics users are asking about and manually includes them in Mitsuku’s vast knowledge database in order to improve relevance and ‘human-ness’.
For example, because of Mitsuku’s Japanese name and anime-like avatar, many users ask her about Japanese culture and anime shows, which Worswick then manually inputs answers to, aiding her learning, and improving her responses.
You can chat with Mitsuku at http://www.mitsuku.com/.
CHATTY: Advances in technology and a renewed interest have given rise to some surprisingly human-like chatbots.