We are a few years into hell; let's look back on views, feelings, predictions etc i wrote a while ago and see what's gotten better, what's gotten worse worse, and what's stayed the same.

• hallucinations: they've improved a bit. still not in the sense that you can trust them blindly though.
• lobotomy: not much has changed, except they've managed to lobotomize one of them in the opposite direction (grok) and turn it into mega hitler.
• local models: silver lining, some of them are getting very good (allegedly). now the problem is that you need trillions of GBs of RAM to run the ones that can rival frontier models (which i don't have). also RAM prices are fucked, hopefully not forever.
• contributions to humanity: still mostly slop and scams i'm afraid. everyone is "building" apps and i still haven't found one that is useful or innovative.
• solving problems: i'll admit that i've been impressed by claude solving some tasks, especially work related ones. still nothing in the creative/ideas department though, and i feel like we've hit a hard limit on it, a "secret ingredient" that humans have and could possibly be replicated by machines eventually but can't be solved by scale alone.
• media generation: fine, some models are getting scarily good at generating images, videos, music... but then the question becomes: where are all the interesting books generated by llms? where are the movies? maybe they're not as impressive as they seem.
• ubiquitousness: a lot of people don't like ai and companies know it. even google has started offering a way to turn off gemini features for example. it's hard to tell for sure but i can almost see the line of forced hype begin to curve down. hopefully it becomes something that isn't in your face all the time.

As for what's new:

• other than GPUs and RAMs, the prices of storage is possibly also going to increase in the near future. will locally run hardware become a thing of the past? will gaming die? or will this force developers to design more efficient software? who knows, maybe the bubble will burst and hardware will become crazy cheap again.
• the latest fad is something called openclaw. it's basically a thing where you give a llm control of your computer, phone, emails etc and let it do whatever. i wish them the best even though i think it's overhyped and don't really see its utility other than being a fun (and hazardous) experiment. (also, it's basically a wrapper that makes api calls to cloud models so i have no idea why people are buying mac minis to host it)

So what happens now? Will we be out of a job soon? Will the bubble burst? Will nothing happen? We'll have to wait and see...