Posted in The Switch 2 thread
MPH - you must be delighted to hear you might actually be right.
User since | Last active | Started 21 topics | Posted 1551 times
MPH - you must be delighted to hear you might actually be right.
Yes, I'm about 2/3rds of the way through it and started playing it again. I'm past the obvious spoiler tho.
Who were the Death Stranding Truthers on here? cav and Felt? I think if you liked DS1 you should like this quite a lot. It's not gonna' convert many people IMO (it all depends) but it just keeps getting better and better.
I thought DS1 was astonishing (and I like a bit of Kojima bullshit - MGS4 being my personal highlight). I just didn't enjoy the core gameplay loop much. DS2 is pretty interesting, I might dip when it hits PC.
I think extending this into a megathread might make sense, but won't it drive traffic even lower? I appreciate no forum is getting new members anywhere (rllmuk is in a slow death spiral too) but fresh threads are the lifeblood of any forum.
If not a monthly PWB - maybe there need to be a few more repeatable threads that generate traffic. Or maybe we need a Play, a Want and a Bin thread.
I'd disagree with that. "still no use case for it that people are willing to pay for at a price that covers its costs"
It's absolutely transforming virtually every middle management, coding and procedural task. I pay the GPT fee because I can't afford not to now. If they crank it ten fold I'll still continue to pay it. It's going to wipe out entire professions. A lot of professions are entirely tractable LLM problems. The object permanence is completely solvable, and there are plenty of methods to do so. You can already see this happening in extended video content on social media (with repeating characters). Yes, its all prompt engineering and tech stacks, but I don't buy that the complexity of what it's trying to replicate is intractable. GPT 4.x's deep learning research capabilities are astounding. I have it triaging and finding cancer targets, prioritising them, assisting in molecular design and then analysing the results. I then do lab validation of a lot of its output. Years of work can be bypassed in an afternoon. And work that contains novel insights in new disease types, where no previously published literature exists. So it's not (just) stealing shit, it's reasoning (or giving the impression of reasoning) based on data and context. It can make mistakes, but there is an eerie progression in its capabilities, even in the last 3-6 months. We're firmly into the "what is human intelligence anyway?" question. Maybe it's just LLMs all the way down.
The market is overheated and overvalued yeah, but it's not a crypto bubble. This is like a rush to stake the claim to be the google of the next 30 years. Few winners, enormous rewards for those who make it, blasted wasteland for everyone else.
Microsoft have committed to investing something like $65 billion into the death cult of LLM infrastructure; that money needs to come from somewhere and Satya Nadella (may his toes never be un-stubbed) doesn't care where, so long as it's not his compensation package.
AI has largely done for music now, the current gen of generative AI music is largely indistinguishable from the real thing. As people have said, how long until Spotify simply replaces all the real artists with AI generated music you'll barely be able to tell the difference from? Why bother paying royalties. Film isn't far behind. Games not far behind that. In, what, 3 years or so, you will probably be able to generate a Gears or Forza at home. Maybe even sooner. Entire creative industries are going to be picked off, one by one, by tech bros who don't give two shits.
And sure, you can be a carpenter or electrician, but if you have total economic collapse who are you serving and how are they buying your services?
We are galactically fucked. It's UBI, live events/connection, rethinking capitalism and a completely new way of doing human civilization, or a massive war.