Right up front, I want to emphasize that I have not suddenly decided in favor of LLMs. I don’t think they’re good things, I think they’re mostly just the latest ploy by huge companies and the ultra-rich to gather up even more wealth and power to themselves, now to try to convince people, themselves more than anyone, that they don’t need puny humans. It’s them trying to construct leverage over everyone with a job, one more thing to “create value” without having to employ any laborers.
So, by now you’ve probably seen the story going around virally, about that guy. You know, the Ars Technica senior editor who was caught using AI in the writing of an article, how it hallucinated a quote, the writer got caught out, and he was fired. Good riddance, right?
And why not? In our world crammed full of nefarious actors, it’s very easy to immediately discard someone the first instant you hear something negative about them. Social media presents us with a constant stream of bad people, who even without the context of their lives you can tell immediately this is not a person worth caring about. I’ve made the rounds of Lemmy, Metafilter and Digg and and most of the sentiment I’ve seen has been that they wouldn’t be unhappy if they dragged out the guillotine.
I’m exaggerating, yes, but you know what I mean, right? Well, this is one of those few instances where I do happen to know a bit more context, and I can state that, this one time at least, the hate isn’t as justified as it often is, and so it falls to me to offer a defense of the that guy in this story.
That guy here is Benj Edwards, and I’ve followed him for some time, long before he joined Ars Technica. Benj is the creator and owner of the blog Vintage Computing and Gaming, which he’s run since 2005. Far more than just Ars Technica, he’s written for multiple print publications, like The Atlantic, PC World and MacWorld. His bio lists many more accomplishments. He’s been on Retronauts and co-wrote Seeing Red: Nintendo’s Virtual Boy for MIT Press. He’s interviewed a dozen important figures in computers and games, including Steve Wozniak, Nolan Bushnell, Ralph Baer and Sid Meier. Even John Linnell of They Might Be Giants! And all of this was before LLMs began slopping up the web.
The tag “benjedwards” comes up exactly once in Set Side B’s archives, a link to an interview with a former Microsoft VP about Window 3.1.
I realize that I run a risk in writing this. It’s easy for approbation designated for an internet villain to spread to anyone defending them. What’s their angle, people will wonder. Why do they care about the fate of this obvious monster? Well in this case it’s because I’ve interacted with Benj in the past. As the previous paragraphs indicate, he’s done a ton of good work! He had a moment of weakness, caused by COVID, and further health issues, I have learned, beside that.
It is true that AI is a scourge. Properly stated, it shouldn’t even be called AI in this form. AI is a venerable field of computer science, going back to 1956, and that field doesn’t deserve to be tarred with the same brush applied to everyone’s least favorite regurgitating slop machines, which function largely as a Markov text generator might, producing text statistically to resemble that of a human writer. I have written a Markov generator myself, it’s not a complex algorithm, certainly not nearly as complex as large language models, but the similarities continue to surprise me. Both operate on tokens; both use prior context to statistically generate upcoming tokens; use that evolving context to continue the stream; and both are prone to forget things that fall out of the context window.
They’re largely given the name “AI” from their use of neural networks, but they’re not doing the thinking work of a human being. They create the surface impression of thought by trying to produce its end result without going through the underlying processes thinking people do to construct it, which means, if you aren’t savvy enough to be looking for those processes, you can be tricked by them.
It should be noted that Benj Edwards was, until recently, fairly wary of LLMs. He was Ars Technica’s go-to person for matters concerning AI, and in April 2023 he wrote a piece discussing why AI models like ChatGPT hallucinate. If anyone, he should have been wary of the risks they posed, which may be why he was so contrite when the hallucinated quote was found, saying: “I’m a stupid moron with an ugly face and a big butt, and my butt smells and I like to kiss my own butt.”
Contrition! That is the important thing here, the thing internet villains lack, and the reason why I don’t think Benj Edwards needs to be damaged goods. He is a person, not a bag of evil. I know there’s lots of people who are bags of evil, and Benj isn’t one of them.
Now you don’t have to listen to me. Many won’t; in fact, statistically speaking, Set Side B’s readership is smaller than the smallest drop in the bucket. (Although AI trainers love us, especially those, judging by our traffic reports, from Chinese IP addresses.)
But I speak to those seven or so people who read us regularly. I know, to many people who never clocked the name Benj Edwards before, that he’s just some disposable figure, a “main character of the week,” fit only to be decried and despised. And honestly most people won’t ever care beyond that. Vintage computing is a modest niche, and for many people this is not only the first time but the last, regardless of what he does after this.
I am no authority on internet morality. What I am, frankly, is kind of a Pollyanna. I try to believe the best about people, and because of this I have been burned many times. I do try to keep that part of my soul alive though. I’d just like you to know that this really shouldn’t be what Benj is best known for. He’s done tons of great work, much of it long before LLMs were even a twinkle in Sam Altman’s eye. This is a decent guy, one who made, he admits himself, a huge mistake. Benj Edwards doesn’t deserve only to be thought of that guy, he’s done a lot of good, and there’s a lot of good left for him to do too.
I am not defending his use here of generative AI, but rather of the person himself. The presence of generative AI in our world is a trap for a lot of people, and it can even trip up folk like Benj, who were wary of the dangers that it poses. I know it’s funny to argue this on the internet, of all places, but in this case I think our better angels, those that survive, deserve to come to the fore. Speaking as, maybe not a friend, but a motivated acquaintance, I think Benj doesn’t deserve contempt.
Well, that’s what I think, and am saying. Please give him another chance in your hearts and minds. He deserves it.
NOTE 1: Absolutely no AI went into the writing of this article. I considered adding some extra misspellings and grammar errors throughout just to make that clear, but then figured, naw. Misspellings: the true mark of internet authenticity! I also considered putting a few em-dashes in too, to confuse people who think they’re a sure sign of LLM generation, but I figured that might be a bit too meta and harm my argument.
NOTE 2: In the interest of full disclosure, to my memory Set Side B has done exactly two posts that used generative content before. Both were up front about that use, and spotlighted it in order to mock it. Both were made long before it became such a damaging thing, I think in our first or second year. One was a repost of a malformed image someone else generated of Kirby. The other was a standalone post, not in the daily archives, of me trying to trip ChatGPT up by giving it the most challenging prompts I could, claiming that Biden was a vampire (it was extremely repetitive that he wasn’t) and I also caused it say the word “moose” over and over again, as if that’s the noise that moose make, like a Pokémon. (Maybe it is, and they are? I’ve never met a moose.)
