SFSD-X

Round 3 - Notes

Well this was a tough one to write, and I didn't have much of an idea where it was going when I wrote the first few paragraphs. I decided early on that it would be Richard the Third as an Artificial Intelligence, because that would allow me to have him survive the battle at the end of the play.

After a few days thinking about it I thought it might be fun to have some AI researchers watching what was going on. Then the story came slowly, with the idea that the opening monologue was a diary or memoir, that 'Richard' could somehow 'escape' from the lab. And if he did escape, well in the play he states that he wants to make everyone miserable, because he cannot be happy. So, of course, he's going to want to destroy the world isn't he?!

So I was lucky again, because the lyrics to the song that formed part of the prompt are all about death and destruction, and when Richard gets control of the Trident submarines, he has the 'machines to feed'!

From a hard scifi technical point of view, the question of AI on a digital computer is something that has been debated at least since Turing's famous paper. We already have programs that can fool humans that they are real people. So in that respect, they pass 'The Turing Test' to some extent.

The issue of whether or not 'true' consciousness can be realised in a digital computer is much more difficult. John Searle's 'Chinese Room Argument' is an excellent thought experiment that is well worth looking up and reading if you are interested in the topic. Personally, I find it compelling, but there are many great thinkers who follow the 'Strong AI' school, which claim consciousness is an 'emergent property' that manifests in any sufficiently complex system. Searle posits a distinction between syntax and semantics. McCarthy, Dennet, Hoftstadder and others claim the distinction illusory.

I've used a little bit of computer jargon, but it's all accurate so can be referenced if required. Hopefully it doesn't detract from the story, but it's actually quite hard to write something factually correct about computers without it...

It was tough as well to reconcile what the 'entities' as I've called them might experience 'themselves'. Would they be able to 'see' each other? What would that mean for a computer program, that clearly does not have eyes. I've talked about the space they inhabit - the halls and chambers, the clothes they wear, that Richard is disfigured. Visual properties of their experience that transfer via the 'Cameras' and 'Screens'. If the prompt was longer then I think I would have had to justify including this aspect in the story - perhaps some detail about the programming which allowed such language to be used even though the entities are entirely software based. 

I stated that Richard was a 'mutant', in order to correspond with his hunchback appearance in the play. I originally intended to expand on that, but it wasn't going anywhere. I left it in though, because it was a reference to the play and I liked the line about the maggot having contiguous DNA :) It also falls short, I think, in that there is no explanation about the pundits that discuss the news feeds, and the extent of the Kingdom. And another short-coming is that I haven't made it clear about the historical time that they inhabit. When I wrote the bit about the veil shifting automatically, I was intending to set the scenes in a virtual future, and I was going to include virtual machines and technology. The idea was to have a world inside the computer programs that mechanical hovering horses, and floating drones that interacted with the characters. That fell by the wayside as I ran out of words for the story.....

The 'neural-network module' was a late addition. Neural-networks make use of hidden nodes that interact with neighbouring nodes by forming connections with each other based on values that reflect the strength of the connection. The are 'black boxes', in that you can apply inputs, and measure outputs, but the values within are meaningless. A neural-network module needs to be 'trained' - you apply known inputs and outputs and allow the program to settle into a state where it's outputs correspond to what what's required. Subsequently, you can apply inputs for which you don't know what outputs you desire, and take the given outputs as a solution to the problem. They work well for 'fuzzy' problems, where an algorithm is not available. 

So for example, a problem such as recognising an organic shape such as a tree in a photograph is clearly a problem where you cannot easily develop a mathematical formula to solve the problem. So you present the network with photographs (coded so that you have a set of inputs) and each time you get the response you want (yes a tree, no not a tree) you accept that state and eventually the network is able to identify which of your training set contain trees. In theory, you can then present new photographs, which the network has not been trained on, and hopefully it picks out the ones that include trees.

In the early days, there was a well known study where the research team trained a neural network to identify aerial photos showing military tanks. The network was able to identify with 100% accuracy which photos in the training set pictured tanks, but when they showed new pictures it didn't work. They soon realised that in the training set, all the pictures in the training set that included tanks were taken on a cloudy day, and all the others on a sunny day! So they had inadvertently trained the network to distinguish between sunny-day photos and cloudy-day photos!

It's a fascinating area of study, and I suspect one day (given the funding) some research group somewhere might well produce a system that uses neural-networks to perform some amazing human-like problem-solving.

I wanted to include as well the idea that creativity is born of constraint. In a lot of ways, that is what these contests are about - we are given constraints with the prompts, and it helps us to write because we are not staring at a blank sheet. Hoftstadder (who I mentioned above) wrote a book about this titled Le Ton beau de Marot  about a poem he found from the French writer Marot. It's a poem in 22 lines, with three syllables per line. He wanted to translate it into English, and ended up with well over a dozen different versions, all true to the metre and rhyme, and translated by different people. It's a fascinating read.

Finally, I wanted to write something that did not develop the characters - there's no romance, we don't get any backstories to the characters; the three main characters have names, and we see a little bit about how they interact. But overall, it's just a plain scifi story about what might happen if an Artificial Intelligence was able to become conscious. And the prompt allowed it to be malevolent, and to realise it's full evil, given that it began as an evil character from a play. 

I don't know if it works that way, or if it's a bit boring. There's no twist at the end, it just gradually builds from something the researchers want to pursue, and ends up with full-out thermonuclear distruction :)

It was interesting learning about Trident too - I don't think that part of the story is plausible, in that there are comprehensive fail-safes to prevent an unauthorised launch. But I was amazed to find that at any time, there is a submarine out there with eight missiles, each carrying twelve nuclear warheads that can be independently targeted. The missiles have a range of 7,000 miles, and the warheads have a yield of 100 kilotons. Scary eh?

So, a great prompt, lots of fun to write, and finished in time with an hour to spare! Yay!

(BTW - Ann Hathaway was, of course, Shakespeare's wife (I visited her birthplace once in Stratford-on-Avon :) )

(Actually, my family moved to Stratford-on-Avon when I was four, and moved to Devon when I was ten. So I spent six of my formative years in Shakespeare country. And I'm ashamed to say I never once saw a play during that time...)

Apart from the literature on AI and machine consciousness, influences on this are obviously stories and movies where machines become conscious. The Terminator films have Skynet. When I was younger, me and my friends were enthralled by the film War Games. Before that there was Proteous, and the 60s film (that I can't remember the name of now) where a computer took over missiles in America, and then communicated with its counterpart in Cold War USSR. Also, there was an episode of TTNG where the Holodeck allowed a virtual Professor Moriarty to take over the Enterprise! It's a well-trodden path in scifi terms... 

Having Shakespeare as a prompt was fantastic for the imagination! And even though I thought the music was absolutely dreadful (sorry AngusEcrivain!) the lyrics were perfect for Richards final interactions with the poor researchers that gave him a mind of his own!

What a great prompt! (And too bad for all the typos and misplaced grammar!)