https://arxiv.org/abs/1705.08807 https://arxiv.org/abs/1705.08807

Is there much point in us bothering? AI will outperform humans in just about everything by the middle of this century, finds a survey of machine learning researchers. This could just about be the most important topic of the modern age.

The good news: AI may prove to be an essential aid in our quest to prolong life. The bad news: many of the people reading this article may live to see the day – the day when AI makes us just about redundant.

It is just that the experts may be too cautious. When the New Scientist Magazine tweeted its take on the story yesterday, with the headline “AI will be able to beat us at everything by 2060”, Elon Musk re-tweeted suggesting that 2040, or even 2030 is a more likely date.

Or is that a deadline – humanity’s deadline for remaining relevant?

There is good news for surgeons, AI won’t be able to outperform them until 2053 – so if you have aspirations to be a surgeon, or if indeed you are one, you have still got a 35-year career ahead of you.

Not so good for book authors – AI is expected to be able to be better at producing a best-selling book than a human being by 2049, that’s hard to believe isn’t? Will AI really be able to do a better job that Dan Brown?*

The survey also found that experts reckon AI will be able to outperform language translators by 2024, writing high school essays by 2026, driving a truck by 2027 and working in retail by 2031.

There is one big flaw in this research, looking back at the quality of the authors’ high school essays, it would be a surprise to learn that AI can’t already do a better job.

The survey found that researchers believe there is a 50 per cent chance of AI outperforming humans in all tasks in 45 years and of automating all human jobs in 120 years.

Intriguingly, Asian respondents are expecting these dates much sooner than North Americans.

There is a well know saying, sit a monkey in front of a typewriter and give it an infinite number of years, then at some point during this long-time frame, it will produce the complete works of William Shakespeare. An idea that was taken to its logical conclusion by Douglas Adams’s Hitch Hikers Guide to the Galaxy, on board the infinite probability space ship: “There's an infinite number of monkeys outside who want to talk to us about this script for Hamlet.”

In reality, it probably wouldn’t happen – a monkey may never be able to replicate Shakespeare by random. But if you put in some kind of control mechanism, in which the monkeys are in some way rewarded every time they do something that is one step closer to producing a Shakespeare play, that may be different. Evolution has such a control mechanism, it is called survival of the fittest, and as a result it is the greatest innovative force we know of – albeit rather slow. Nature, thanks to evolution, has crated solutions to puzzles that scientists still don’t understand. For that matter, evolution created a monkey called William Shakespeare.

The point about computers these days is that it in many ways their development can be quite similar to natural evolution – machines design machines to design machines. Throw big data into the mix, and we already have AI that can beat the best player in the world at the game Go.

All this does not necessarily mean we will create a machine with free will, or self-awareness – we have no way of knowing if this is possible – but if they can do all the stuff we can do, even if they don’t understand what they are doing, better than we can do it, does that not rather undermine us?

And while an AI system that wants to rule the world may never develop, hand control of the military to AI – after all, one day it will be able to do this better than we can – then we have no way of predicting what it will do with that power.

This is the single most important issue of this age. And if you think you have a good chance of surviving the next three decades, or care about what the world will be like when your kids are older, then the time to start talking about it is now.

*”Renowned curator Jaques Sauniere staggered through the vaulted archway of the museum's Grand Gallery” could an AI ever be able to create poetry like that?

Reference for survey