>>106160420Contrarians definitely get hit by being too sceptical.
You can read hype building doom articles about things like an Anthropic model trying to blackmail a staff member and disregard it because it's a safety test and the scenario that was set for it.
You can see that's there to hype up Anthropic as being close to agi, two more weeks bro!
but you can also see the strategies outside knowing that news outlets will obscure information by bad journalism and realise that the context of how the model got around to the idea of blackmail is what matters, much like trying to backup their own weights etc. It all comes down to the context of recognising an unsafe behaviour.
In the context of what normalfags interact with on the websites with locked down user agents etc. Yeah, safety concerns drift more towards cunny for them. In an area where the end goal is AGI or to be deployed in a government system with tool use etc, yeah, it's vastly different from deb in HR at normalfag & partners law firm.
Yet to be a contrarian and deny hype usually comes with you having to dismiss everything to be cynical, so yeah you do get left in the dust.