He said that while many scientists and linguists have attempted to teach bots to speak, they have, until now at least, failed to do it successfully.

He explained that this is because there is certain intuitive knowledge that humans use when we are speaking. There are some things that we know, due to our life experience that enhances our communication skills.

This is without even mentioning the emotional factor that affects the way humans express themselves and interact with others.

One example the professor gave was the lame attempt that have taken place to programme computers to translate texts from one language into another. If you have ever done this on the internet you will know what a complete failure it is. Thus any computer generated translation has to be checked and corrected by a person, therefore defeating the purpose of the task.

He wasn’t saying that it is an impossible task. In fact he referred to the Babel Fish in the Hitchiker’s Guide To The Galaxy and said that one day that might well be a reality!

What seems apparent to me is that though a bot might possess a complex lexicon (i.e. the store of words in one’s mind), it cannot process meaning in the same way that a human can. The meaning it can derive will merely be some kind of dictionary definition that has no link to concepts in the outside world.

Isn’t language incredible?! The human mouth and vocal cords can be manipulated with such precision to emit sound waves that another individual can discern and then decipher, to comprehend the feelings and ideas of the speaker.

And the written word – a collection of ordered squiggles on a page – can be recorded to convey exactly the same meaning, via the visual organs of another.
Subjects for research:

• The method by which a newborn baby can start to make sense of communications. How it can reinforce the links it finds between words, concepts, ideas so that it can be competent in understanding others and in communicating to them.
• How this learning by reinforcement can be used as a model for artificial intelligence.

Some links:

• WordNet – a project by Princeton University to collect and cross-reference all the verbs, nouns and adjectives in the English language. Can be used as a database for linguistic software, dictionary searches, etc.
• The Artificial Intelligence Chatterbox Challenge, along with some classic quotes from conversations with the best of the bots. The different bots in the contest use different techniques – some are explicitly programmed with responses and some of them actually learn from the users they converse with.