Algorithm World

Through high school and even early college, I would’ve considered myself a tech-optimist (though probably not with those words). I looked forward to not only the yearly E3 videogame conference, but Apple announcement events; even new CPU and GPU releases. I’m not quite old enough to remember a time before the internet, per-say, but I do remember its gradual roll-out. I remember a time before YouTube, where videos were distributed across the web if at all. I remember phones before the smartphone revolution. I remember the net seeping into our lives, wrapping tight around our hearts like roots in the earth.

I guess what happened for me was the promises started to out-pace the results. I remember watching the announcement for the Apple watch, and the disappointment I felt when seeing its square face, the minimal list of features. Is that it? Of course, hype existed long before then, and will continue on as long as people are there to buy it. Since then, the promises have continued far ahead of the results. Uber is convenient, but is it an essential infrastructure?

Before you might think I’m advocating a RETVRN to oral storytelling and an agrarian lifestyle, I still remain in awe of a lot of technology. When I hear a word in Japanese I don’t know, within 20 seconds not only can I look it up and get several dictionary definitions, I’m able to make a new flashcard so I don’t forget it. Another app can tell me a bird’s species with a single photo. I wear a ring that tracks my sleep cycles and daily activity, monitors trends in my health, makes suggestions that have a noticeable impact on my quality of life. Technology is good, except when it’s not.

And here we get to AI. I am an AI-reality hater, but an AI-promises fan. I’ve been using Arc as my internet browser, and they recently decided to refocus the company towards “doing your busywork for you.” In practice, that means AI. They released an Android version of their browser, with one of their attempts towards this end. The entire mobile browser is focused around an AI search engine, designed to “read” relevant web pages for you, and return the information you’re looking for without all the messy middle steps.

Normally I wouldn’t be interested in something like this, but I felt that The Browser Company had earned some goodwill through their earnest attempts to innovate and compete with the likes of Google and Apple. So I applied to the beta program. When I was granted access to it, TBC asked users to use Arc Search as their primary browser. I didn’t go that far, but I gave it a good shake. And what I realized is that those messy middle steps are the most important part. If you’ve ever tried to make something, you should know that trying things out is creation itself. I think it got one (1) inquiry of mine correct in the week-ish I used it. More often than not, it gave me a summary of outdated bullet points, devoid of context or the ability to delve deeper without posing another inquiry. But, that’s all AI as we know it is capable of.

Put me in the camp of those who think LLMs are fundamentally flawed. I’m not worried about “AI” becoming smarter than us and destroying the world, more than I’m worried about my Instagram account destroying the world. I am worried about these programs making the internet as we know it unusable. But actually using any of these apps has been only disappointing thus far.

I think there’s obvious cause for concern with a lot of this stuff, but also reasons for hope, or at least points you might think of as off-ramps on the horizon. Flat out, these systems might simply be unsustainable for the cost, both monetarily and in energy usage compared to the economic benefit they provide (I’m no expert, but this seems minimal aside from the influence its had on certain stocks). But if things were to continue, we’re hurtling towards an internet that’s at best fully obfuscated by algorithms, if not choked up with AI generated, dubious data. I know few people who go to use a social media app and are thrilled by the sorting or serving of content based on algorithmic preference, so it’s a little disorienting to see many of those same people champion LLMs. Getting a computer to do what you want it to is far from an easy task at times, but I struggle to see the utility when this is abstracted beyond iteration.

I’ve said this before, and I’ll say it again: I like the idea of a digital companion, one that knows my interests and aims, and can help streamline and that I can discuss topics with. I just fail to see that as the direction we’re moving in. Technology often improves, but what if the approach is wrong? At the very least, I don’t want the distance between myself and the computer to continue to expand, every the more unclear what’s happening behind the curtain.