Sorry… after droning on a some length about the “I” in AI, I realize that I was dragging a LibreOffice topic off the rails.
I’m moving it here as a Lounge topic instead.
TL;DR: It’s hard to agree or disagree - because we’re talking about the meaning of an adjective in the ever-evolving context of our spoken word.
The long version…
I believe much of the angst over which adjectives we ought to reach for when describing AI stems from the fact that these words entered our lexicon well before any modern technology was introduced. And as such, they come to us with much cultural and emotional baggage.
Adjectives like intelligent, smart, and knowledgeable are descriptions of an inherent internal mental state which we can never observe directly. Instead, we infer them from external behaviors and identifiable work products.
For example, we typically conclude whether a person is intelligent by judging their mastery of language and apparent grasp of what’s going on around them.
From what we can observe externally, we draw inferences about their internal mental capacities. And we then reach for subjective qualitative terms like intelligent, smart and knowledgeable if we are sufficiently impressed. In truth, such terms are about as subjective as words like nice, friendly, helpful. Or pretty, handsome, beautiful.
Even an IQ test produces a metric based entirely on external behaviors (picking answers), not on direct inspection of one’s internal mental processes.
And so today, as LLM advances in mimicking language proficiency (while actually relying upon probability models - not cognition nor consciousness in any human sense), we are drawn to associate their communication “skills” with intelligence… without actually being able to directly observe the algorithmic internal processes.
In fact, even their programmers have lost track of the observable internal details given the scale of their models’ processing power and the complexity of their neural networks. (Alpha GO perfectly illustrates where we now stand on this front.)
So we’re relegated to selecting our adjectives from the limited set which evolved for us to describe intangible human traits… not to accurately characterize what’s going on inside these new electric appliances.
Perhaps if we developed a new set of adjectives, we could more easily draw (less emotionally charged) distinctions between the capacities of humans to those of our new machines.
Of course our tools can do a far better job at certain tasks than humans can. When it comes to math, the abacus proved that millennia ago. For breaking ground on a data center, so does the shovel.
Meanwhile, humans will always prevail at other activities, like lovingly cuddling and nursing a newborn. Or sacrificing our well-being and self interest for those about whom we deeply care.
As a capitalist, I should also note that not being able to use the term “intelligent” to describe AI models would likely undermine their owners’ abilities to raise capital, market their wares to other businesses, and spook the rest of us about the need to get on board before it’s too late.
