The Artificial In Intelligence

What in the world is it?

Quartz recently published an article on the scope of what we mean when we say Artificial Intelligence. You may read it here, but broadly there are two schools of thought.

There are the generalists who define the term broadly & these include systems such as Siri & many other such personal assistants. In this broad definition, as long as a system uses an AI component, understands commands & completes tasks, it qualifies as AI.

Then there are the literalists. This group cares much more about the architecture of the system in that it should be one end-to-end computational unit. Stringing together a bunch of algorithms to achieve a degree of home automation does not qualify. These are marvelous feats of design without being AI.

To make a hard thing more difficult, Douglas Hofstadter, has this to say about the Jeopardy winning Watson:

Watson is finding text without having a clue as to what the text means. In that sense, there's no intelligence there. It's clever, it's impressive, but it's absolutely vacuous.

In the same interview, Hofstadter says about IBM & Google:

They're thinking about how we can get these computers to sidestep or bypass the whole question of meaning...

Making meaning, even where there is none

Yuval Noah Harari, speaking about his much acclaimed book Sapiens, posits that the progress of human civilization may be attributed to cooperation basis shared beliefs propagated by stories. Money is a proxy for economic worth, and gold is universally valued. Such value, as you know, isn't intrinsic but the beliefs run deep & we have created global systems basis such shared beliefs.

Searching for and attributing meaning to all kinds of phenomenon & rallying around these points of views, says Harari, is the reason behind where we find ourselves in the world today.

Wait, I thought this was about AI

It is. But if we plonk humans intelligence on some kind of a pedestal and think about how AI looks in comparison, things start getting very murky. Because, in a comparative sense, human intelligence isn't just computational ability that AI can out-calculate. In humans, the metaphysical is mixed up with the modular and data conflicts with the doctrinal.

With humans as yardsticks, what manifests as intelligence draws from multiple, sometimes conflicting, faculties.

So you want proof?

The neuroscientist Antonio Damasio had noticed an unusual pattern of symptoms in patients who had suffered damage to a specific part of the brain - the ventromedial prefrontal cortex (vmPFC). Their emotionality dropped to nearly zero while they retained full knowledge of right and wrong and showed no deficits in IQ.

Yet this liberation of reason from the thrall of the passions rendered them completely incapable of any decision making, including in purely analytic and organizational tasks.

The head can't do head stuff without the heart.

Look, just tell me what you mean

If human faculties are anything to go by, conceiving intelligence as an isolated, analytical engine which can be augmented boundlessly by applying copious computing power seems deeply flawed.

I'm somewhat resigned to the notion that AI will disrupt jobs & economics. Automation is already doing this. But I wrote this article to express hope that there are limits to what AI can achieve in the way it is conceived today.

References:

1. The Righteous Mind by Jonathan Haidt
2. Sapiens by Yuval Noah Harari
3. The Rise Of The Robots by Martin Ford
4. Descarte's Error by Antonio Damasio