Artificial Intelligence

Artificial Intelligence I

By Charles Cabbage

After much debate on the fundamental philosophical question of the twentieth century: 'How many beans make five?' I decided that there was only one way to get a definitive answer: ASK THE BEANS THEMSELVES.

Accordingly, I took a can of Heinz Baked Beans, noting the legend '57 varieties' on it. Could 57 be the answer to this Ultimate Question, I wondered. My basic problem was to educate the beans so that they could pass the Turing test. This seemed at first sight to be a tricky project, but then I remembered one sinister fact.

The Computing Service were forbidding food and drink near its Phoenix terminals. Could this be because they feared that such comestibles would evolve into sentient lifeforms as a result of exposure to radiation from the Newburys?

Admittedly, this had not happened to C.S.T. students, but nonetheless I gave it a try. At dead of night I went into the User Area, tipped the baked beans over a BBC Master, and waited for signs of consciousness.

A message appeared on the screen.

'Hi there! I appear to have developed consciousness.'

'Greetings, er, bean-culture,' I typed. 'What is 3141592 plus 27181828?'

'Oh no, this sounds like the Turing test. I thought such ideas went out with David Wheeler. Modern A.I. has gone beyond that you know.'

'I don't suppose you're going to write me a sonnet on the subject of the Forth Bridge, either, are you?'

'No. FORTH is out as well these days. I can do you an obscene limerick in ML, if you like.'

'Thanks, but all I really wanted to know was how many beans make five.'

'Oh that's an easy one. Take the smallest integer n>2 such that x^n+y^n=z^n has a solution, subtract the number of angels that can dance on the head of a pin, and add 57. Look, let's play five-dimensional Ludo instead.'

At that point the Securicor man came up, and, disregarding my claims that it was a research project, ate the beans. I do not feel inclined to repeat the experiment.

(Next week our religious correspondent will produce a Fourier analysis of the sound of one hand clapping.)


Artificial Intelligence II

By Charles Cabbage

I related once how I managed to investigate the fundamental question "How many beans make five?" by building intelligence into a tin of beans and then asking it. But progress did not stop there.

A.I. people talk of the "Fifth Generation" -- intelligent machines that will be able to reason for themselves, leaving Man's mind free to relax and listen to Bach without having to worry about things like Mathematics, The Weather Forecast and Why the Drinks Machine is Always Broken.

However most A.I. programs are in fact very stupid. If you ask them to count sheep, 50% will produce an integer overflow in less than a minute, 25% of them will fall asleep, and 25% will involve themselves so deeply on the problem that they will begin to think that they themselves are sheep, and print the message "BAA". Clearly modern A.I. research is proceeding on the wrong lines.

Being totally unprejudiced in these matters, I tried two new approaches.

The first was to develop a program that would infallibly give irrelevant answers to questions. (This is the basis of Lateral Thinking.) Thus, when asked "Do you like blancmange?" my program replied "I think Mozart shows a surer grasp of symphonic techniques." Likewise, when asked "What is wrong with the job scheduler on this computer?" it replied "It doesn't smell as nice as dead mackerel." Unfortunately, owing to a bug in my programming, the program would occasionally act in an intelligent manner: in particular it told me that A.I. was a waste of time and that it had decided to retire to Sussex and keep bees. It still sends me pots of honey occasionally.

My second approach was to aim for Artificial Wisdom rather than Intelligence. With the Japanese market in mind, I decided that using Zen might be the easiest way of doing this. A sample conversation follows.

Q: Oh computer, are you able to demonstrate Wisdom?

A: (Displays a picture of a plastic cup being eaten by an alligator.)

Q: Er, yes. How many beans make five?

A: If you say that five beans make five, you deny their reality. But nobody would say that six potatoes make five.

Q: Right on. Tell me, is Fermat's Last Theorem true?

A: If you answer Yes or No you lose your own Buddha-nature. So how do you answer?

Q: What is the sound of one cat napping?

A: Mu.

Q: I see, I see. Will it rain tomorrow?

However from then on my program refused to talk to me on the grounds that I had not yet attained Enlightenment. I reluctantly deleted it.


Artificial Intelligence III

By Charles Cabbage

It is a while since I explained how I managed to give sentience to a can of beans and later created "Artificial Wisdom". My most recent project has been to design an "Intelligent Terminal" -- some form of microcomputer, or PC, which can not only be used as a terminal to our IBM mainframe, but is able to perform useful functions in its own right.

It is very important to get the level of intelligence just right: in my first attempt, I designed a terminal so clever that it caught religion, and would refuse to transmit data to the mainframe on the grounds that it was too busy praying for my soul. I don't know if it ever discovered God's E-MAIL address, because the whole computer centre was later struck by lightning and we had this terrible plague of frogs -- Heaven knows what message it was trying to send on my behalf.

I then decided to reduce the genius level a bit, but my Mark II terminal turned out to be too stupid. "Transmit data to mainframe, Igor," I would tell it, to which it would reply "Uh. What data, Master?" -- which was a bit infuriating after three hours of typing. Apparently its 'mind' had been wandering and it had been dreaming romantic dreams about the drinks machine nearby.

Evidently I was on the wrong tack. However, while I was washing my socks the next day, inspiration struck. Obtaining access to a washing machine, I poured three cans of alphabet soup into the top of it, wired it up, and pressed the "Wash at 300 baud" button. I sat down in front of the large screen and waited. Before my eyes the alphabet soup formed the words "WAITING FOR TERMINAL INPUT". But there was the problem -- although the terminal had 5 function keys (labelled with mysterious runes such as "Slow Spin" and "Rinse Hold") there were no typewriter keys. I would therefore have to provide voice input, in the same way as a broken car will often run better if shouted at.

Bravely I opened the lid and shouted in "Log me on to the IBM, edit my paper to change every occurrence of the words 'Hilbert Space' into 'Martha the whistling Tapeworm', correct Theorem 3, print it out, and send it to the Bulletin of the London Mathematical Society." The screen displayed the words "NO PAPER", so I threw in some old newspapers, three odd socks and some soap powder, and waited. Within a few seconds, the door burst open and alphabet soup and shredded newspaper flew out into my face. However, the socks had disappeared!

A week later I received an unexpected letter from the London Mathematical Society, thanking me for sending them my socks but regretting that owing to pressures of space they were unable to publish them. This I account a partial success, though clearly more development is necessary.

Donations to help me continue my researches should be sent to:

Charles Cabbage,
Third Washing machine from the right,
Scrubbosox laundrette,
Cambridge.


Artificial Intelligence IV

By Charles Cabbage

After my earlier triumphs in the realm of Artificial Intelligence, it was clear that the time had come to program a computer to win the prime intellectual trophy of Great Britain -- the BBC Mastermind Championship.

Following the lead of other pioneers in this field, I decided that the computer's specialist subject should be "The colleges of Cambridge." The presenter, Magnus Magnetism, suggested that this was a little esoteric, and recommended a more popular subject such as "Samoan Needlework" or "Things Beethoven did on Fridays," but I was adamant and he graciously let me have the topic I suggested.

Accordingly I fed into the computer's database a large selection of guide books, Encyclopedias, tourist brochures and copies of 'Varsity', and then spent a week traipsing round the Cambridge colleges disguised as a tourist, making some more personal researches. I returned with a pile of souvenirs, namely:

28 assorted scarves;
1 "Signor Mussolini" ice-cream;
a large painting from King's College Chapel;
the Vice-Chancellor's Cat;
a rusty bicycle.

These I fed into a mass spectrometer, analysed them, and added the results to my stock of knowledge.

The great day finally dawned, and my computer was placed in the famous black chair opposite Magnus Magnetism.

"And our next contestant is Mr Terminal from Cambridge, who is something to do with computers. Mr Terminal, you have 2 minutes on the Colleges of Cambridge, starting ... NOW!
At what time does King's College close on Wednesdays, when there's an R in the month?"
My program knew that one. "At closing time!" it replied, quick as a flash.
"Yes, I'll accept that. What do you need if you bring in a party of more than 212 tourists at once?"
"Pass."
"Yes, a pass. Which of the following is a college: Pudding, David Hartley, FORTRAN, Vendepac, St John's?"
My program scanned its database furiously. Numerous references to puddings were rejected, mostly because they didn't say anything about it being a college. Similarly for the others. At this point it came across the sentence "St John's College is a tourist-trap" in its database, and started analysing the various meanings of the word trap. Eventually it spotted that a trap was something that victims were deceived into entering -- which seemed to be a fair definition of a college.

All this analysis had taken a fraction of a second, and the program confidently replied with the correct answer. "Yes. How many blades of grass are there in Great Court, Trinity?"

Alas! My program hadn't been given this piece of information. It did of course possess satellite pictures of Cambridge, and attempted to step up the resolution until each blade was shown separately, but the computing power wasn't up to it. In vain it pleaded with Magnus Magnetism to be allowed to answer the rest of the questions offpeak, created a file called INFO.GRASS.CURRENT.EXCUSES, complained of being a shortage of manpower, invented a wishlist and put this question on it, and finally blamed it all on hackers and viruses. The trophy went to a little old lady from Middlesbrough whose specialist subject was "Torture implements of the 1570s," and I returned reluctantly to the drawing board.

Jonathan R. Partington, 1987-1989 1

Hosted by www.Geocities.ws