MONKEY WRITES HAMLET, GETS NOBEL PRIZE FOR LITERATURE
A monkey with a typewriter and an infinite amount of time will sure enough by sheer probability recreate any existing literary work, right? This statement, which is part of the Infinite monkey theorem, is a household name in lecture slides, being used to get a laugh out of its student viewers. Though it is theoretically possible and also unfunny, one would not deem it probable after a quick glance at the equation. Computer scientist Robert Wilensky keenly put it as: “We’ve heard that a million monkeys at a million keyboards could produce the complete works of Shakespeare; now, thanks to the Internet, we know that is not true.” I would pose this joke not only rings true for internet users, but for the capabilities of artificial intelligence.
“We’ve heard that a million monkeys at a million keyboards could produce the complete works of Shakespeare; now, thanks to the Internet, we know that is not true.”
Computer scientist Robert Wilensky
WELCOME TO YOUR DRIVING EXAM
Let’s dial it back a notch. I cannot find the original video on Youtube anymore (much to my agony), so I would like to very strongly express that the following experiment is not made by me!
Pretend you are driving down a lane in your car (if you don’t have a license, just imagine it’s an illegal joyride or something). Suddenly, there is an object in your lane that you want to dodge. Your quick reaction skills save the day, you decide to switch to the lane on your right. Here comes the question: How do you move the steering wheel for this maneuver? Make sure no one is watching you slowly go insane and then play out the movements. If you made a tug to the right and then back to neutral, you’re incorrect and have crashed into the cars that are two lanes over. Why? Well, you were supposed to make a tug to the left as well to align with the next lane. Kudos to whoever got it right now, though rest assured most would have done it correctly in a real situation.
So why didn’t you do it correctly now? The answer is Polanyi’s paradox: We know much more than we are able to express and than we ‘know’. This makes it extremely difficult to create AI which mimics our capabilities, we don’t even know what we are able to do and how we do it!
Additionally; AI is just not ‘talented’ enough for some tasks (just like I will never be able to become an Olympic swimmer). Suppose we gave an AI the task of understanding what or who ‘his’ is in the following sentence: Bill passed the Gameboy to John because his turn was over. It is very likely it wouldn’t know because it doesn’t try to reason or understand the sentence as we do, and it probably won’t (proficiently) for quite some time to come.
RISING UNEMPLOYMENT AMONGST COMPUTERS, LOOKING FOR JOBS WITHIN SKILLSET
So if AI can’t do my primary school fill-in-the-dots Dutch sentence exercises is it all useful?!
Computer scientist and artist James Bridle has the following postulate ‘We are inclined to define intelligence time and time again as ‘that which humans do’’. It would be a good look for us if we reflected on the strong suits of computers and what they should do for us within these parameters, redefining the intelligence in AI.
One possible road would be the creations of artist Refik Anadol. By creating abstract digital works with AI, he has delved into a niche of computer art that is not trying nor needs to compare with human capability. Also, I mean come on, it looks pretty rad.
Looking for something more practical instead? Take a gander at the movie ‘Her’ Without spoiling the story for those interested, the role of AI in this movie is not to become truly and completely intelligent in the human sense. Its function is rather to support and nurture humans insofar as a computer can.
Bottom line:
Introspection on reasonable use cases of AI might take us much farther than we are right now. After all, do we really need AI-generated Hamlet?
References
https://en.wikipedia.org/wiki/Infinite_monkey_theorem#In_popular_culture
https://en.wikipedia.org/wiki/Polanyi%27s_paradox
https://www.groene.nl/artikel/de-wereld-anders-denken
https://en.wikipedia.org/wiki/Winograd_schema_challenge
https://cs.nyu.edu/~davise/papers/WSOld.html
https://refikanadol.com/
I have learnt quite a few things from your post (and, hopefully, one or two things would be of help with my driving skills or whenever I am at a logical dilemma of “Where should I NOT crash my car?”) But, my question to you is slightly trivial — do you really think that the main idea of the movie ‘Her’ is to show that AI functions to help us, poor humans, with daily tasks? The idea to me seems highly exaggerated, at least judging by the trailer. I mean, isn’t being in love is already perceived as something too “intelligent” for an AI like Samantha to do? So, in a way, it’s no longer a story about support but about dependency (and not a healthy one). Also, do keep in mind that linguists were once behind programming languages — you saying, “AI is just not ‘talented’ enough for some tasks” misses one key word: YET. “AI is just not ‘talented’ enough [YET] for some tasks”.
Your question is not at all trivial! I think you’re correct the movie Her is an extreme example. Nonetheless, the core idea of a virtual mental support system that morphs to your needs is not too far-fetched, we can see it happening right now in the MedTech industry. On the topic of AI being talented enough or not, it depends where we draw the line. Will AI be able to excel at any and all tasks we give it, I still don’t think so and am not alone in that thought. Will it be possible to do slightly to a bit more than it can now, probably.