Artificial Stupidity: A Linguist's Perspective on AI

n the article “Artificial Stupidity” linguist Geoffrey Pullum states his opinions on artificial

intelligence and how his every day interactions with simple technology affect his

opinions on this. He starts his article by reciting a log of the redundancies of specific

user-interfaces that he encounters throughout a five day period. Most of these

interactions are with simple appliances such as a projector giving a notification that it

will go to standby mode in “one minutes” or a self-checkout machine asking him to

“Insert cash or select payment type” after he has chosen to with cash. He then goes on

to state that based off these interactions, he’s certain that all technology isn’t capable of

communicating with humans on an intellectual level and won’t be for a long time. He

backs this statement up by saying that most user-interfaces are programmed by the
people who weren’t chosen to program a way to make Volkswagen diesel-emission
controls operate only in laboratory tests. He believes that the predictions of artificial
intelligence and robots won’t be able to work with humans without making mistakes and
sounding unintelligent.

What this man doesn’t understand is the process that takes

place to make and program these appliances and the decisions that have to be made

while creating these user-interfaces versus how artificial intelligence and robotics are

developed for interacting with humans. He doesn’t understand that user-interface for

basic, older appliances and artificial intelligence operate on a completely different level.

The first thing that Pullum doesn’t understand is that most user-interfaces weren’t

made for him specifically. They were made for a wide range of people for a large range

of situations. When he talks about the self-checkout line, he points out that the idea of it

asking for him to insert his cash or select payment type after notifying the machine that

he was going to pay with cash is redundant. He may have changed his mind after

selecting cash or accidentally selected the wrong type of payment. This allows users to

know that they can still choose a different payment type even after they have made an

initial selection. This allows for an easier user experience. Without these prompts

and
notifications, there would be more complaints about the difficulty of the user-interface
than the redundancy of a prompt. If they were to develop and install expensive artificial
intelligence at each machine, they could. It could read your movements, facial
expressions, and the items you were buying to mathematically calculate the likeliness of
you changing your mind on your payment. But if you, against all odds and reason,
decided you wanted to pay with credit after selecting cash, you’d still have to notify the
machine of your of heart. At this level, he’s basically asking the checkout machine to
read his mind, which is impossible for a computer or a human to do.
The second thing this man doesn’t understand about creating user-interfaces is the
time and money it would take for people to program these prompts that would satisfy his
user experience with technology; Especially with most of his nitpicky complaints. In his
article, Pullum talked about the redundancy of an elevator someone being alone in an
elevator and it announcing the direction in which it was going. This was obviously added
to allow new passengers a moment to get off before they are swept in the opposite of
the direction that they need to go. This doesn’t really help someone if they are the only
person on the elevator, but to develop the technology to sense the amount of people in
the elevator would be hard to justify. When you like at it from and business standpoint,
you have to ask why you would need this and how could you make money from
developing this. It would be possible to use existing technologies to help quicken the
development of this, but then you’d have to pay the companies that developed those
technologies. Same with the previously mentioned self-checkout machine. From a
business perspective, no one in their right mind would spend the money and time to
develop such things when they could have it run through the simple code of asking or
announcing the same thing every time it’s performing an action. It’s thousands of times
simpler and cheaper, while just as effective.
When it comes to Pullum’s complaints about projectors making mistakes such as
saying “Going to standby in 1 minutes” and microwaves continuing it’s cycle of beeping
five times even after he’s opened its door; It comes down to memory management.
These are basic appliances that are running off of hardware build to make it affordable
and usable. Usually, these appliances’ hardware aren’t designed to hold more
information than what is needed to operate at the basic level. This is because
manufacturers want to keep the cost of production down. Programs take up and use
memory. When the programmers are given the specs for the product the need to work
with, sometimes they have to cut certain features that they could easily include in
products that have better hardware. This being said, most of these petty complaints
aren’t stemmed from the issue of the incompetence of the programmers, but the
hardware they’re working with. It comes down to keeping money in pocket over the
appliance including extra text that most people wouldn’t even notice.
My final and biggest qualm with this article is the statement that based off of his
experiences with older phones and appliances, he believes that artificial intelligence is
not going to be able to communicate with us on a level that we can consider intellectual
by any means. He thinks that even if an artificial intelligence can communicate with us,
it can’t sound half as smart as a human can communicate. What’s funny about this
observation is that you can easily go on the internet and find AI’s that can now identify
how people communicate and mimic us. They use better grammar than the average
human and can find patterns and observations in words and theories that it could take
the most experienced writer a while to recognize. All of the author’s observations are
based off of appliances that are completely unrelated to the theory he has come up
with. It draws a parallel to a man looking at a cheap screwdriver and assuming that it’d
be impossible for another person to build a house with the tools that they may have at
their own workshop.
To wrap things up, my final response to this article is that you can’t let your nitpicky
experiences with basic appliances let you think that computer advancements are
limited. You can’t let the autocorrect on your cellphone or the text on a microwave
display make you think that it’d be impossible for computers to be able to communicate
on a level any deeper than misspelled words and grammatically incorrect sentences.
There are plenty of industries working hand in hand with linguists to make sure that their
technology is able to communicate in a way that many find human. Though, I’d find it
hard to believe that the author of this article would let himself believe that if it weren’t
spoken to him in person by another human.