Trying to be human is getting in AI's way
Thoughts on bots
Back in 2017, on the fifth-floor of a downtown LA highrise, I was on hour five of combing through code in the corner of a coworking space that smelled of “corporate but cool because we allow dogs.” Despite not having taken a math class since high school and my only real notable work experience having been as a production assistant in television, I found myself employed at a tech startup programming chatbots.
I was searching for rogue commas or out of place zeros or backwards symbols that were causing bugs in a chatbot for Hyundai cars or the Gorillaz or Jack in the Box or Steve Aoki or… whatever company or entity that had taken the bait of our VP of Sales’ promise that chatbots would make them money.
I, an interactive producer or writer, title dependent upon if we were in a meeting with a client or I was asking for a raise, was getting bored. Not just with my computer screen, but with our product.
This was five years before ChatGPT, so we were too soon for the mainstream culture to understand chatbots in a non-customer-service capacity and dollar signs were getting in the way of innovation in terms of how our company leadership sought to use our technology. It wasn’t long before I left, the company went belly-up, and I was back on a studio lot. But with artificial intelligence now firmly in the cultural zeitgeist, the hand-wringing of Congress, and the essay-writing process of my fellowship applications (whatever!), my thoughts have circled back to that weird short-lived time working in tech.
It didn’t take much internet rabbit-holing to land on some intimidatingly intelligent and impressively young historical figures whose intellectual feats caught my attention and juicy personal lives kept me around.
While I can’t tell you a thing about how the program works that Ada Lovelace published in the 1840s making her the first published programmer 100 years before the first machine resembling a computer was even built, I can tell you about her troubled gambling on horse races and her supposed extra-marital affairs and her lifelong obsession with her absent father, the famous poet and philanderer Lord Byron who left her and her mother when she was just five weeks old.
I haven't a clue how Alan Turing’s “thinking machine” worked but I know how it was essential to breaking the Nazi code that led to the end of WWII and that the British government, despite his contributions to their war success, essentially ended his life when they found out he was gay.
Gossip and personal tragedy aside, Ada and Alan saw the future. Ada, in the 1840s, when the most advanced still only theoretical steam-powered machines were just capable of printing math tables, predicted that one day machines would be able to write music. Nobody in her day was thinking like that. It’s a little hard to succinctly describe Ada’s contribution to technology since what she contributed was a book rather than a machine, so for now you’ll just have to trust me she was important. Luckily, for the sake of describing Alan Turing’s impact on the world, there’s a pretty good movie from 2014 called The Imitation Game that I can point to (given leeway for creative liberties of course).
But what I’m focusing on today is the two of them and their points of view on artificial intelligence 75+ and 150+ years before its arrival on the scene. As well as my point of view on it, because I am the once in a generation mathematical mind missing from this equation.
Ada
Ada thought machines, as much as they would eventually be capable of, would never be able to have original thought, as they will only ever do what humans tell them to say or do or think. They will always be reliant on human instruction.
Alan
Alan believed that with enough computational power and the right algorithms, his “thinking machine” (which functioned as an early computer), would eventually achieve parity with the human mind. He didn’t believe there would be a limit for Artificial General Intelligence.
Well, a quick google tells me that ChatGPT-4 passed the Turing Test, which caused shockwaves in the AI world as finally a machine reached this covetable achievement of technically having parity with the human mind. But having used it myself and conducted my own rigorous research of casually talking to friends, I’m suspicious.
The most advanced AI systems today, although available to be used by anyone with a keyboard and an internet connection, still require sophisticated prompting to return the best results. While this progress is impressive nonetheless, my friends and family do not need very sophisticated prompting to spit out an original thought. In fact, usually the simpler the better and often no prompting is required at all. So, I would argue that our current reality lies somewhere in between Ada and Alan.
After a rewatch of The Imitation Game and then sitting through a rambling and likely incoherent lecture from me about the history of computers and the personal lives of long-dead mathematicians, my poor watch partner wiped his blurry eyes and reminded me of this art installation:
Can’t Help Myself by Sun Yuan and Peng Yu
This anthropomorphic machine, capable of performing “dance” tricks when commanded, constantly leaked red fluid. It was programmed to do one job—sweep the fluid back into a predetermined area when it went too far. The machines task and its position behind glass as if in a cage made people feel some kind of way. “It’s sad,” he said, recalling the piece. I shrugged, “The machine isn’t sad though.”
So then came my theory.
Sam
It’s not really about what the machine is capable of, it’s how we, humankind, evaluate the machine.
Some may argue that this is the point of the Turing Test and not actually an original theory, but I differ in that I don’t think it’s important if we ever see the machine as “human” per the Turing Test or having “original thought” per Ada Lovelace. Forget human parity. Making that the sign of success is getting in the way of what’s actually important for its own longevity and survival–making us buy into it emotionally. We don’t need to ever feel like we’re talking to a human, we just need to overcome the mental barrier of feeling emotionally connected to a machine.
Here’s a poor analogy that would easily be torn to bits if I thought about it for five minutes longer – so many people, especially in the US, see pets as family or best friends. Even if they could have talked, I can tell you that the conversations I would have had with the pets I’ve adored through my life would probably not have been very intelligent. In fact, having a human conversation might have gotten in the way. It’s not essential for connection.
We decide when a machine cleaning up fluid makes us sad. We decide when our pets are our children. And we’ll, at some point, decide when computers have rights. Artificial general intelligence doesn’t need to pass as human, it just needs to make us care.