Maybe we shouldn't have called it AI
by Gene Michael Stover
created Sunday, 2021 November 14
updated Sunday, 2021 November 21
Problems with AI
Problems with the term “artificial intelligence”
- “The definition of intelligence is
Right off the bat, when you use the term “artificial
intelligence”, you've invited a debate. Unless what you
want is to debate the definition of intelligence, that debate
will take time & energy from whatever you really did want
to discuss (such as recent successes, failures, or
implementation questions that haven't been solved yet).
- If you mention the intelligence of a program or other
machine, you've opened the door for someone else to argue
about whether a machine can be intelligent. As with the
previous point, unless this is what you wanted to debate, it's
a waste of time &
- Someone will invariably point out that whatever
A.I. technique you want to discuss isn't how brains work.
Again, we're now in time-sucking debate territory that
detracts from the interesting topic you wanted to discuss.
- “Artificial intelligence” inspires wariness in
some people, making it difficult for them to grok whatever
point you're trying to make
- “Artificial intelligence” inspires ethical
questions in some people, making it difficult for them to grok
whatever actually interesting point you're trying to make
- We probably don't even have an agreed set of requirements
that many of us would agree is intelligent (even if you
discount the people who question whether a machine can be
intelligent at all). So again, why waste time with the debate
that'll probably follow?
These problems with the term “artificial intelligence”
apply to its predecessor “cognitive simulation”
& probably other predecessors I'm not aware of.
“Machine learning” has the same problems. In
addition, it's often considered a subset of A.I., so it wouldn't
be an appropriate replacement.
Instead of AI
Do you even give a fuck whether it's intelligent, or do we
just want & need software that solves hard problems that we
didn't program it to solve explicitly?
Instead of “artificial intelligence”, maybe we
should use a term that's more indisputably descriptive.
Seems that a descriptive term might be “flexible,
expandable, & resilient” systems. It's not catchy,
& the abbreviation is F.E.A.R., which doesn't help against
the worried & ethically concerned people.
Someone needs to come up with a descriptive, indisputable,
& catchy term to replace A.I.
And for gods's sakes, let's stop suggesting that an automaton is
or will be intelligent. It distracts from the important
- 00. definition is controversial
- 01. philosophical debates
By the way, very smart people have debated these ideas for
thousands of years without finding solid answers. So you
won't find a solid answer, either. And even if you do (and
you won't), the next step will be for lawyers, artists, &
the rest of civilization to take it to the next step. Your
individual word won't count for much. And like I said, you
won't solve the philosophical question anyway. So unless you
want to debate it because you enjoy debating it, don't bother.
(And few of us — including you & me — know
shit about debating, so don't bother to debate anything at
all. It's a waste of time.)
- 03. a final thought
I would not be surprised if the term
“intelligence” limits how we think of the problem
& its solutions. If we jetisoned the word from
practitioners's associations with the systems they are trying
to build, wouldn't that allow them to imagine more possible
solutions. That jetisoning could take a full generation or 2.