Assorted thoughts on AIv1

AI is a lot of things

The intersection of personal experiences gives me unique and conflicting perspectives on AI (or LLMs more typically). I’ll hone in on three of them here.

On one hand, as a developer, I see great value in pushing the software development industry beyond creating and tending to boilerplate code.

On another hand, as someone who enjoys works of human expression and creates a good deal himself, AI represents a potentially dark force in human creativity.

On a third hand, as a person who lives, breathes, and respects everyone’s journey in doing the same, AI is a mixed bag that can give and take in equal measure based on how it’s used.

As a developer…

AI is in equal parts exciting and horrifying. I’ve seen firsthand how much more quickly I can get projects from ideation to completion and that’s exciting. I’ll grant I do miss the puzzle of software development but I get that in my free time with game development anyways anytime I want. There are definitely some concerns about skill atrophy on my end as there are for every other dev who now feels more akin to an open source maintainer than a developer but there are ways to mitigate that by being mindful.

Security though, is going to be wild. Not just because AI is quite good at breaking into systems, but because AI developed code will likely fall into making somewhat repeatable mistakes that leave doors open in more predictable ways. I’ll grant handwritten code isn’t necessarily secure against an antagonistic AI assault, but I suspect the variability in designs across individuals and teams likely does result in some degree of “biodiversity” among codebases which should be healthy.

As an artist…

By art I’m referring to anything designed with the chief intention to convey emotion or feeling. That of course includes drawings, but also music, games, performances, and so on.

AI is a net negative to artistic expression in my eyes. Personally I think anything can be art, so even if someone broadly prompts [insert agent here] and looks at output they barely envisioned and thinks, “I made this”, that’s still art.

Now that said, not all art is equally worthy of merit. To me what makes art truly great is a high degree of intentionality in design (yes, this is incredibly biased). This doesn’t necessitate a clear vision of the final outcome from the onset but it does necessitate a clear mindset of where things need to be by the end. AI art generally lacks intentionality as, to many, the homogenized default output from AI is “close enough”.

Now I don’t mind one bit if more bad art is thrust upon the world. Nearly all great artists start their journey leaving a wake of questionable art. But AI I worry will displace many a budding artist’s early start given:

• People who want art but don’t know how to make it turn to AI instead of learning art themselves
• People who need art but don’t want to pay for it turn to AI instead of artists, leaving art as an amateur passion
• In the deluge of custom art out in the world, people become atrophied to art altogether
• The unique styles artists cultivate are copied and replicated wholesale

I won’t fully close the door on some potential merits to AI art as I do think there are some interesting ideas in using AI to procedurally generate things that would prove cumbersome otherwise (like generating playable cities with functioning interior spaces). But do consider me highly skeptical of AI being a force for good here.

As a person…

AI can certainly be a force for good as much as it can be a force for evil. Ultimately LLMs, as they are and will be for a while, are tools; and tools aren’t inherently biased (granted the sources AI is trained with likely are so it isn’t that simple). While I’m not on the “those who control AI are evil” bandwagon, unfortunately I do think those who control AI are at the mercy of financial markets, which isn’t much better. Big companies just need to make next quarter bigger and better (or at least make it seem like it will be) and in that journey anything that isn’t the bottom line can be collateral, including us everyday people.

But ultimately I do worry about AI brain drain the most. When people stop needing to think and can stumble their way into workable solutions to their problems without really understanding where they are and how they got there, that presents an issue. Even ignoring more extreme cases such as AI psychosis, having a society where everyone is aimlessly marching around to the beat of the same drum is quite scary. A personal belief of mine is that one should never let anyone do their thinking for them, yet people seem especially quick to with AI.

I also find myself still questioning the legal underpinnings of AI usage given I doubt it’d be “fair use” under conventional legal interpretations of copyright in most instances, but it seems like we’re all just meant to ignore that so… I feel silly dwelling too deeply on this here.

I’m optimistic with some uses of AI such as in medicine. I do think there will be some fantastic quality-of-life improvements for us there, but unfortunately AI auto-denials from insurers seem to be quite a bit easier than genuine medical progress. Still, I figure we get noteworthy major medical advances in spades before too long.

So…

Does this mean I think AI will be a net positive or negative? Yes, it will be one of those. Anyone who claims they know with certainty probably has a more narrow outlook on things (or at the very least prioritizes some things quite heavily over others).

Where does that leave me? As a developer, I’ll keep using AI; it’s just the direction the trade is going and the productivity benefits can’t be ignored. Even if all the frontier model makers die out when the bubble pops, local models will still offer quite a good development speedup. But elsewhere in life I’ll try to steer away from being too reliant on AI as convenience should never come at the expense of personal development.