SOBRE ESTE EPISÓDIO
This episode has been a long time coming. I think I’ve been teasing it since I relaunched the podcast back in July. Outside of some private conversations, for the most part I’ve been very quiet on the topic of machine learning technology. Part of that is nerves. I mean, who in the world really wants to be slapped in the face by all of the internet? Not me, that’s for sure.
However, I think these things need to be said. The conversation around machine learning, AI, and large language models has reached a fever pitch, particularly in creative communities. And, by and large, I think a lot of people (most people?) have it wrong. The things that most people are worried about (jobs, theft, slop) are a lot of the same things we’ve been seeing for years; just repackaged. Likewise, all the stuff that most people seem excited about (“democratized” creativity, fast output, the future of work) also misses the mark.
I say this in the episode, but I suspect it’s worth repeating here more explicitly. I’m not a pro-AI guy. But I’m not opposed to machine learning tools just on principle. I’m not anti-AI. I’ve tested AI tools over the last few years since they started gaining popularity, but I don’t really use them for anything… not knowingly, at least. I don’t like using AI for what I do because it’s the stuff I enjoy doing. And for the time being, I think I’m better at it. That said, if a machine learning tool comes along that meets very specific criteria, I would consider it.
And yes, a big chunk of that criteria involves open source and running locally on my own hardware.
This is a long episode and I cover a bunch of stuff in it. I spend a good chunk of it describing the lens through which I see all of this and then explaining why I’m not particularly interested in talking about those things I listed above. The shortest bit, ironically, covers stuff that I am actually interested in. Things like the environmental impact of cloud-based data processing, the death of the creative introvert, concerns about digital cloning, and the opportunities available with small language models, open source, running locally, and self hosting.
And I end the episode referencing a point brought up by Wes Burke, co-founder of Autotroph, related to the candle-making industry. You can catch that reference in his BCON Austin talk.
