PyotrVerkhovensky
No bio...
User ID: 2154
PyotrVerkhovensky
2d ago
hey begin to think about targeting non-coding white collar work like finance and spreadsheet work since the models are not getting much better at JavaScript
This is already happening: https://www.anthropic.com/news/finance-agents
DeepSeek releases a model that is equivalent to 5.3 Codex
Already almost there if you include all Chinese companies, certainly will be by end of 2026: https://livebench.ai
Frontier models begin to entshittify as they are increasingly jailed to make them safer, while their private reputation is shattered among all of the normies who did not know how rights-violating the United States secret police are.
I'm with you on this one. We've already seen movements in this direction (eg, not releasing Mythos).
It's pretty clear to most people the general improvements to coding are dried up and a lot of the old hype was fake and tooling and chaining was the internal, secret meta from 2025 onwards
And more targeted RLHF, but yes agreed on this. However, I think there is still a ton of yet-to-be tapped potential in tooling, context, and feedback that will have massive impacts even at current model capability.
Overall my timelines are shorter than yours but I do think there is a "ceiling" and I don't think we are at risk of Yudkowsy's takeover scenario. I do anticipate "mundane" surveillance and increased slopification. My hope is in local/opensource models running on ASICs, which would at least alleviate privacy and intentional kneecapping concerns.
Bad takes on AI seems to be the one commonality across creed, race, and IQ. This one from theringer is a particularly egregious example, but its rare to find anything both sufficiently technical to understand how it does and could work, and sufficiently "big picture" to understand societal impacts. (Of course, many others would consider my AI takes to be just as bad).
My current modus operandi is to be whatever the other person is not. If they are an AI maximalist I am the pessimist. If they are the doomer I am the optimist. If they think this technology is all hype I become the autistic technologist with in-depth details and explicit examples.
More options
Context Copy link