Really surprised by the results. I probably have a skew in my followers and/or people have not updated their beliefs after recent events (Pentagon contracts and AI principles anyone? 😅)

Personally, I'd look at trustworthiness first to decide whether to believe public statements and assurances by an actor, and then look at those together with their known actions given the historical record to draw an overall conclusion

For the former, it should be sufficient to observe any inconsistencies in words vs actions to update on the effective trustworthiness of actors regardless of their perceived virtues and prestige.

E.g. they might appear the most virtuous but if they race and treat safety as an afterthought right now, they might not actually live up to their own standards right now

And if they are not trustworthy, you cannot trust they will act according to espoused values when the stakes are even higher (safe AGI and later ASI)

That is, one should ask oneself: is Google trustworthy given how they say they are against mass surveillance and autonomous weapons in PR statements and aspirational contract language while the reported operational reality seems different? (And obviously there is a lot more to look at historically, e.g. read the Project Mario chapter in The Infinity Machine, etc)

And why are Google so silent on this topic in general? (One could assume they would have corrected the record by now otherwise btw)

Eric Schmidt once said: “If you have something that you don’t want anyone to know, maybe you shouldn’t be doing it in the first place.”

I'm not sure what that means here but it seems important

Similarly: is Anthropic trustworthy given that they have lived what they say and fight for it in regards to their interactions with the Pentagon (regardless of whether you agree with it all)? (Not sure about other details of the historical record here)

Repeat for xAI and OpenAI

This should give a good indication for overall trustworthiness

One could say that trustworthiness is a low bar but then it still seems too high for some 😬

(xAI is interesting as I have no expectations that they care about safety.)

Thus, without judgement, to decide whether Google, Anthropic, OpenAI, xAI are to be trusted to deliver safe AGI, I'd look at Anthropic's policies and values, but I'd discount Google and OpenAI's statements, values, and published frameworks to varying degrees, and look at their actions so far instead (and acknowledge that xAI was an option to catch trolls 😇)

My own stance for this exercise is that sometimes it's good to take a step back from one's day-to-day and think about what one's work is actually enabling and whether it does that for the right actor because it would be a shame to have regrets later on