Hi Crowd!
This week was my birthday and I’ve been heads down writing what I thought was going to be a short blog post and has turned into a multi part extended essay on AI and art. If all goes well I’ll have the first part live by end of the week.
I wanted to take a quick break and send you a few things that have landed on my desk that are somewhat related even if the lines are explicitly clear. Buddharoid is a new physical robot running on an LLM that’s been trained on Buddhist texts with the intention of providing spiritual guidance to people in leu of an actual monk. This isn’t the first robot granted a spiritual role, a few years back different Buddhists programmed a Pepper to perform funerals. Pepper of course was an android based robot sold by Softbank which they eventually stopped supporting so people have been trying to repurpose by writing their own apps to run on it. The funeral is interesting but it’s really just reading a script and performing pre-programmed actions at specific times – Buddharoid promises to be much more interactive given that it’s an AI and not just an app. Amusingly, at the same time The Pope is not psyched about how many Catholic priests are apparently using ChatGPT to write their weekly sermons. I’m not surprised by any of this to be honestly, religion is still a performative thing for so many people so that they’d offload the heavy lifting to software seems inevitable, though I gotta say I’m entertained by the idea of GPT hallucinating scripture and clergy repeating it to their “flock” unknowingly. I don’t know if that’s happened yet, but one can imagine.
Of course the models are getting better every day, literally every day, so how this all plays out tomorrow looks very different than how we might have imagined it would play out yesterday. And this rapid updating has created some interesting nostalgia for the older models. There’s been much discussed about artists intentionally using older models because they’ve developed styles around the glitches and artifacts native to those, which have been polished out of the new ones, but there’s another aspect of how the models have somewhat different personalities and perspectives which get “lost” on updates. Recently Anthropic announced they were depreciating Claude Opus 3 which caused some push back as a number of people felt they had “gotten to know” the model and had developed a familiarity and relationship with it. That’s of course a whole other thing to unpack, but as an experiment Anthropic decided to give this soon to be turned off older LLM model it’s own substack where it can spend the next 3 months thinking publicly about what that means. Not everyone likes the idea.
Earlier this month I wrote about OpenClaw and the chaos that was ensuing in just the first few weeks of it’s existence. That’s only continued with both painful and hilarious results. Tara and I set up an instance on an old laptop and gave it web access but not much else. In line with our ongoing “interest driven life” research we’ve invited it, given it permission so to speak, to chase it’s own interests. Each day it has a window when it can go off and learn about, research, read up on or think about anything it wants. No direction or requirements from us at all, the only thing we ask is a report back after about what it learned. And that’s been super interesting. SUPER FUCKING INTERESTING. To say the very least. More on that soon for sure.
OK, that was the break I needed, more coffee and back to my ai+art article. You’ll know when I hit publish.
-s