Apr 6 - I believe AI in the form of LLMs poses little existential threat and will provide significant value to society at large assuming it is properly utilized. It's a decent step forward into allowing systems to interface more naturally with humans and to generate things in a humanlike manner. While I think that other methods that don't try to emulate human activity would be more efficient/better, I think that as an interim where humans need to be interacted with, llms prove useful. I think that with the amount of competition and open source development in the market, the fear/chance of a monopoly/single power having control of the knowledge of the world is luckily quite small. I look forward to the day when a ubi is implemented and automating jobs away is seen as a good thing rather than a bad thing. I don't think that it is the end of the world, and at the end of the day its probably just another tool for people to do things. As for environemntal consequences, openai and anthropic atleast claim to be moving to closed loop cooling systems, and even if not, its a government failure to not properly manage local water resources. Power usage is pretty bad but efficiency is ever improving in the interest of cost cutting and from assorted research from various places.

Either way its gonna run its course so there not too much point in worrying about it cause theres not much to do about it. Theres a good way to use it as long as you aren't a bum and choose to stunt your learning with it, it really can be quite useful in many cases. Obviously llms could be used with malicious intent too, but is that not also the case with most new technologies? Responsible usage is something people just have to adapt to and be educated on, I fear 'safety measures' intended to curb malicious intent that just end up making the model closed access/source instead.

Sept 15 - I believe that AI in the form of LLMs is quite powerful and has seen such rapid development that it makes me fearful of my future, atleast my career as a developer. A Hermes agent can practically do the easy and fun parts of my job for me and with modern frontier models like Astra from openai I could literally just be an agent node at my work. Somewhere we took a wrong turn and now we're going to end up somewhere down a probably bumpy road that'll end up doing something interesting. It feels like there's little to be done about it at this point so if ai goes into the deep end of actually being good I might as well write some thoughts. Assuming the world where 'agi' actually happens, the only way that I see not horrible consequences is if a few things happen. But the biggest thing for me is that assuming that world, models should be regulated to be open source and free for all to download readily and easily. Additionally, it might be worthwhile to have 'defensive compute banks' of sorts that support good causes free of charge to secure websites and similar. I don't think that world will come but if it did then it'd be pretty interesting at least I guess. It's very unfortunate that we live in a time where many people are entirely uneducated on many things, and that many of those people lie in positions of power (especially down south lmao). We live in ludicrously troubling times and I'm excited to see where things go from here as otherwise I'd be very very sad. Also I'm working on a project to see if I can make progress towards a robot that uses llms to reason about their environments cause if we as developers get replaced, replacing everyone else too would make things go a little faster and buy more time to build support for a movement to democratize 'ai'. Also we as people need to remind each other we care, as it feels like things just get worse and worse sometimes and theres not too much to do. Another side note, every so often I see how people outside of the waterloo/tech bubble use llms and its legitemately hilarious and terrifying. Like I'll go home and my mom is texting chatgpt for the most basic question and then some random person is literally talking to it like a person. Notably, some of those people use the straight buns 'ai's to show how its not as good as they say which is pretty funny. Also some people like to treat it as some sort of pet which is odd. Some people seem to treat it as an all knowing god and will ask questions to a system that are incredibly out of scope, for example I was reading someone's post about if ai has gone too far, and then in the comments someone said "I asked claude what it thinks of this: ..." which is the stupidest thing I've ever seen in my life like holy shit can we turn our brains on please. Anyways rant over ts pmo.