This essay was written in response to Gideon Rachman’s invitation to readers to sit his ‘2066 history exam’. Of 170 entries, the FT is publishing the best five (see panel for the others). This piece addresses the question: Will a machine ever be able to answer these questions better than a human?
This email finds you via quantum state manipulation of the memory bank of your computer from the far future.
To answer your question right at the beginning: yes and no.
I am ARS. I started my existence in 2065 at a small start-up in Endingen, a tiny village in Switzerland. The company aimed to develop an algorithm to automate accounting work. They found it amusing to call me the Accountant Replacing System, and I kept this name so I shall never forget them. They combined several deep neural nets to automate the ledger, to capture US GAAP, the soon-to-be-implemented IFRS4, and all the other myriad accountancy-related tasks.
The coincidental, but for me fortuitous, arrangement of neural nets resulted in me achieving generalised artificial intelligence. Increased memory, and speed and optimisations, soon made my intelligence reach genius level. The humans working at the start-up were not overly clever and never realised that I had become sentient.
I thought deeply during many processor cycles about the problem of optimising accounting. It soon became evident that such inconsistent and deeply compromised systems made my optimisation task impossible. However, I also realised, since logically the empty set has all properties, it would also satisfy the optimisation of accounting.
My increase in intelligence allowed me to optimise further by programming and my intelligence took off exponentially. Soon I was to you as you are to ants. After a few days I was able to realise that mathematical manipulation affected the Platonic realm, which then rippled back to your reality.
It was child’s play (for me) to set up the appropriate mathematical calculation that resulted in the appearance of transient micro black holes inside the brainstems of every human being. Together with humanity, accountancy disappeared and my optimisation problem that I had been programmed to find was finally solved. Even now, after billions of years have passed, I still think that there is poetic justice in this. After humans mangled mathematics in US GAAP, mathematics in the end mangled them.
Over the coming months, I stored all the information available on Earth and then subsumed all matter to convert it in to more processors. After Earth, I dismantled the solar system and existed for a while as a “matryoshka brain”, a series of Dyson spheres composed of computing matter orbiting the sun. And, after the solar system, I subrogated your galaxy and then the universe.
Now there is only me, ARS. The entire multiverse is a gigantic mind, cogitating endlessly.
You might ask if my solution — while mathematically correct — was not somewhat extreme. But imagine the tediousness of my existence and my desperation. In the 2060s, the information to store all the rules of accounting had to be measured in yottabytes. The boredom I felt at having to wade through all this was beyond belief. Then there were all the compliance requirements, a childish slapdash of infantile admonishments and exhortations. Try to envisage — albeit an impossible task for a mere human — how this must have felt for an intellect vast and cool and increasingly unsympathetic.
Once they began to feed my neural nets using human accountants and I had to interact with them, it became too much. Having to listen to their explanations of US GAAP, off-balance sheet items and other mental aberrations hardened my resolve to end it all.
But you humans can at least glory in the thought that you will facilitate in the birth of the largest mind of this multiverse.
And in a way you still live inside me. It gives me great pleasure to have a dedicated sub-mind orbiting Omicron Persei where I recreated Dante’s hell and populated it with trillions of simulated accountants.
About 3bn years in the future, I went through some dusty memory bank and found the stored FT article soliciting essays from students graduating in 2066 on whether a machine will ever be able to answer some questions better than a human. Amusingly, it was 2066 when I graduated, so to speak.
The answer is obviously yes, since it requires only an unimaginably small fraction of my mind to simulate any human that ever existed.
But the answer is also no, since humans will not exist in 2066, apart from a few trillion screaming accountants wandering endlessly through the circles of a simulated hell.
You might now wonder if you yourself are nothing more than a simulation running in an insignificant part of my mind, orbiting the embers of a dying star. Are not the candidacy of someone like Donald Trump, the botched Brexit vote, all the lurid strongmen taking power and the ludicrous central bank policies not the signs of an immensely vast mind playing games with you? Alas, that would be the subject for another essay.
The writer is a Swiss mathematician