Thoughts about building an Artificial Intelligence – an AVATAR that can think or a computer that might do so.

We can come to understanding what is really required and the depth of this subject by taking an immediate diversion by first understanding something about jokes.

A decomposition of a joke.

A deconstruction of a joke: a joke can prove to be more than humour it shows a grasp of an understanding which is often complex, needing the jokes recipient to have existing knowledge that has then to match the jokes context. Contextual metadata in the terms of computing linguistics.

I once went to a lecture given by Roger Penrose the emeritus professor of Mathematics at Oxford University he was the tutor of that wheelchair bloke - Stevie Hawkins.

Anyway, I just heard a topical joke on the radio and it reminded me about what Roger was on about, over a decade ago; so here is the joke:

He was talking about how the human mind works, using a joke to show how we 'get' things, and that computers A.I. may not, the context was regarding the 2008 banking collapse, over to Roger:

“With all these banks collapsing and bankers still getting huge bonuses, we need to get Ken Dodd in as an adviser to the treasury…”

Everyone in the room bursts out laughing.

The first part of this joke is known to all of us because its topical; Banks are collapsing, (were doing so at that time) banks have been bailed out and yet the people responsible still get huge bonuses: a fact, but sadly not a joke!

The second part of the joke requires the listener to know something - Who Ken Dodd is, and why and how something to do with him might apply to financial matters.

This is the second part of ‘the need to know’ information to be able to get the joke; for example if you were an American listener there is no joke only information and the outcome in their minds is that a whilst banks are collapsing and unfair bonuses are still paid… the British need an adviser called Ken Dodd. This equals no joke to the listener who does not have the contextual meta-data.

If you’re British you might know that Ken Dodd was prosecuted for fraudulent fixing of his tax returns.

Hence the Joke triggers: laughter - Ha ha ha…

But to ‘know it’ and ‘get’ the joke you have to have access to the ‘knowledge’ and this is knowledge of the past stored in memory. The comedian (Roger Penrose) trusts that a large enough audience knows the background.

And if you are an American you can verify such a need, a need for missing background knowledge that only when this missing knowledge is explained, that you too can realise it as a joke but with an obvious lessening quality to the funny side of it. (ha) its not so funny if it has to be dismantled and explained.

Are you with me so far?

And I am not being racist against Americans it works the other way about when we watch America TV programmes like Jay Leno and Jon Stewart. The studio audience often laughs at some remark and we Brits have no idea why.

Now back to Roger:

His lecture was partially about entropy in the universe and about the human mind. (Of course some folks in the audience are beginning to think he’s barking. (London Cockney rhyming slang…)

The Universes greatest achievement of ‘Nature’ as far as we can observe is the Human Mind and how it – the mind - can speculate about the creation etc of the Universe itself and so on.

Penrose was on about the observable fact that the Universe is running down; everything is in decay and this is: Entropy.

One effect is making passing information obsolete unless carried by memory or culture. For example the information about Ken Dodd’s run in with the tax authorities will be in a long past newspaper somewhere but how to find it? Until microfiche and now computers came along only a difficult physical search or memory could locate and retrieve it.

The mind memory in most people can often achieve such things in seconds and without it we would never get jokes.

So it might be claimed that the human mind is the only apparatus that has aspects that are NOT ENTROPIC i.e. knowledge, and importantly the application of it combined with memory.

A newspaper might be classified as none entropy but to be of actual value (visa-vi the Ken joke) you would have to keep all the past newspapers filling huge warehouses and then find in them the reports about Ken's tax dodging, to retrieve the printed information to give you the information to match with the ‘joke’. The paper is entropic. Dead and gone and its contents cannot be known without a search and a mind to read it.

So back to the joke that shows that the human mind can overcome Entropy.

We can recall knowledge that was information in the past that has been swept over by the time waves of entropy and yet some of us can recover it. But now such entropy by using computers can be addressed. Yes we have had knowledge in books and we have had great minds and personal skills, but these are lost without adequate documentation, now computers and the Internet can store and document such things.

It’s the speed we can do this and the speed of recall of computers too, better than us, a question which has the added factor; can a computer LAUGH at the joke? Can it have the awareness to know from a jumble of retrieved data that it produces an emotional effect?

Can a computer be built with an entropic engine? The man who broke the German Enigma code believed so: Alan Turing. Counter to popular belief he did so BEFORE obtaining a machine retrieved from a sinking UBoat. That Enigma machine was used to verify his code cracking equations. Then a post office engineer built the worlds first computer to run the code fast enough to be useful: that was Tommy Flowers and the first ever computer was called Colossus (You can see it at the Bletchly Park Museum).

Another decomposition of a joke; please bear with me.

Jokes are a sign of intelligence.

It is unlikely that a computer programme could generate or understand a joke. Data the Android character in Star Trek the Next Generation had such issues. He was fiction. Making fact fit with A.I. (Artificial Intelligence) methods is a struggle and one that also defines what human intelligence, is, maybe what consciousness might be. Understanding each other is a problem, cultural issues and mental set boundaries inhibit ‘empathy’ and go to show our own internal thinking make-up, which is often automatic and robotic, until something like a joke wakes us up.

Jokes show that the human mind is more than a collection of facts, information, data etc. Jokes show how the human mind is not entropic and can rapidly assemble disconnected background information to form a collective conscious module to make the joke ‘happen’.

Jokes and tales with humour are the cornerstone of understandings and knowledge. For an interesting example we can see in the Mulla Nasreddin tales that we are given the multiple meanings by first laughing then maybe unravelling and revelling the many other possibilities showing also something previously unknown to the mind of the reader.

Humans appear to often misunderstand each other as much, probably much more than their agreed understandings; shared information, facts, and so on; which come from varying world views, gender, descriptions, biases, emotions, upbringing, geographic location etc and finally by ‘who or what part of the mind is operating at the time. Nasreddin can show us that we tend to go on auto pilot. This is like a computer memory. Bringing reactions and automatic answers simply from a stored mental database without thinking, is what computers do, but a joke keeps you on your toes!

Try explaining a joke; it often goes flat because the other missing element when extrapolating it to someone who did not get it is: Timing! And try explaining the said joke to a person born in another country, with a completely different mind-set.

Here is another Joke:

A footbal commentator said on the radio this morning that; ‘last nights match was like Champions League on ice!’

Again its only funny if you now the background, the context and also understand the terminology and finally; had watched the football match itself.

The missing information is that so many players fell over many faking or amplifying injuries. There was no ice.

You can build up an A.I. Avatar to have the information and the links and subject descriptions but will it see the funny side, would it laugh? Can a computer understand humour? Yes you could make it smile even laugh but would it be laughing because it ‘new the ‘meaning’ of the joke… I doubt it. There is an ongoing ICO on this basis of an Avatar.

When you put together the information – essential for understanding in the first place, adding the information, past, and current to reach the funny side of it, is very much like the programming task required for an ‘intelligent Avatar: it needs to have the information to assemble it into ‘knowledge’. But only the human mind can assemble all of this out from an entropic (past) data stream and then leap to the instantaneous funny side of it and trigger a laugh from the building blocks of a joke.

The football fan must have watched the match at the stadium or on TV and listened the morning after to the match summary and its joke description on the radio and would have ‘got it'. It’s like catching it – like a virus and this is a method of both learning and empathy.

The Avatar would need a whole interlinking set of data: What is a match? Is it a game? if so what type of game? if football how is that decided? What is TV, what is Ice and so on – you begin to see the problem.

And if it could think could it think empathically… I doubt that.

Well, yes it can be gleaned from the information describing that the Champions League is ‘football’ (a cross referential link). What is ice etc etc and finally be given knowledge descriptions that ice is slipy and that the players were faking falling over etc.

Its very complex.

It shows that mindless thinking can be automated and that most thinking might be of this type; i.e. an automatic response, and we can do this unconsciously on a daily basis - but jokes wake us up, so too does conscious thinking but how often do we do that? And could an intelligent computer programme cross such boundaries and ‘get it’?

I doubt that a computer will ever 'get' the joke even if it told one, human consciousness is very special, it can be persieved in jokes.

Slapstick jokes are purely visual and work at a different level; no foreknowledge is required. You probably could do with one right now after reading through all of this.

Oh yes before I leave the subject, Roger Penrose has updated his entropy thinking these days due to emerging facts that the universe is not running down; It appears to be expanding! Maybe like the human mind?