pull down to refresh
21 sats \ 0 replies \ @optimism OP 3h \ parent \ on: Don't Overthink It: A Survey of Efficient R1-style LRMs AI
Isn't that part of the mirage needed to get those fiat billions? The message underneath of "you'll never have to do unpleasant work again" is what's being sold. And that's kind of a mirage on its own, because we'll just find something else unpleasant that needs to be done. Like correcting a dumb af chatbot that keeps on generating the same mistake over and over. I'd much rather spend my time on helping a human that retains skills. Emulation of reasoning from a static base of linguistic weights is just another gadget, a hack. It doesn't enhance actual intelligence. A sneaky sales pitch to show progress towards a questionable goal through hacks.
So we're dealing with layer upon layer of questionable narratives that are ridden by a few companies. For chatbots in particular I expect to, in 10-15 years, be amazed that there will be people that still use them, like I am today by people that still use Facebook or whatsapp, or manage their photo's through Google.
I feel that this is our failure - we being people that understand (even if just partially or conceptually) how these systems work - to educate the world. We don't have a good framework to counter distractions like chatgpt, just like we didn't have that for fb. In fact, we don't even have a simple, proven strategy against shitcoins and other scams!
So there's much to learn, but even more wisdom from what we've learned to share with others. I'm hoping to find a way to scale that.