Why AI is a Long Way from Human

For over a year, now, I’ve been watching the hype-panic that AI was “going to take over my job” die down to “this can barely produce a believable image” only to reemerge as “we’re nearly there”. (All quotations are non-attributable, of course. Like any AI, I’m making those up as I go…) Basically, the panic is starting to come back, but without the accompanying hype, sort of like the morning-after hangover. Which means we’re back to the great question: when will AI take over the world?

Simply: it won’t. Not unless we let it.

Unlike any of the science fiction you’ve already seen and/or heard about, AI will only do what is is commanded to do. The AI we currently have (and, please, let’s alter this to “Algorithmic Intent”, as there is no intelligence in it, artificial or not) cannot do anything on its own, it must be prompted into action.

That’s our first safe-guard: without a self-driving system, all our AIs can do is sit idle until we ask them to do something.

The second safe-guard is practicality: AIs have no capacity to “understand”. They can categorize, they can relate items to one another, but prompted with a fresh, new problem, all any AI can do is use its previous knowledge. I know what you’re thinking: But that’s what any human can do. Unlike AIs, humans can have “leaps of logic”, they can think “outside of the box”, they can “guess” with confidence. This is beyond the capacity of any algorithm at this point in time.

I’m sure that many philosophers are actively refining the definition of “human” in part due to these existential crises that a machine may replace a human. Spoiler alert, this has been ongoing for eons and will never stop. Humanity, at its core, is an egotistic, narcissistic, and utterly paranoid creature with a massive inferiority complex. Even the littlest threat to one’s position is enough to stir fear and retribution. It’s those qualities that sci-fi writers use as the reason humanity will ultimately lose to the machines we create, because we iron out those flaws.

Truth is that we barely understand those flaws ourselves. Every hour of every day, we’re reminded through social media or the news that the world is filled with a surplus of frightened people with a God complex, terrified that they’ll lose power or control, and therefore everyone else is an enemy. Or someone works for one. All the same, it’s humanity at its best/worst – doing what it does best, which is acting in the worst possible way.

It’s traits like these that reinforce why AI is so far from being human. It’s wonderfully naive, barely having the abilities of a two-year old to understand why a sound uttered in one way means “milk” but in another might refer to a favoured stuffed animal.

But there are more positive traits we can consider, like creativity. And despite everything you’ve seen AI do, it is not inherently creative. Prompting an AI to create something is, at a high conceptual level, a query against an ever-changing database. Yes, the logic is substantially more complicated than select * from tbl;, but the net result is the same: a query that is following the database’s rules to reconstruct the requested output. AIs are (presently) unable to take part of the result, obtain some inspiration, and start to generate from nothing else. What we get is effectively copied and modified from some already known, a process that is not fact-checked or reviewed, leading to what we call “hallucinations”, a polite way of saying that Algorithmic Intent got it flat-out wrong.

The best AI could hope for at this point is fall into the theorem that an infinite number of monkeys at typewriters could eventually turn out the works of Shakespeare, though I would argue that the monkeys, depending on the species, have far more actual intelligence. (The AIs might win solely on having greater attention and patience.)

However, I feel compelled to note that the Infinite Monkey theorem was more or less chucked on its ear back in 1996 by Robert Wilensky, who said: “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.” And Mr. Wilensky hadn’t yet begun to see the horrific depths the internet would yet plunge to in the decades to come.

Lest we forget, though, we’ve already had about 100 billion humans roam the planet over the last 50,000 or so years, and only one of us has ever written the Complete Works of Shakespeare.