zlacker

Seriously, even in a small group of a few hundred people?

>>osigur+(OP)
I dunno, seems like a pretty self-evident theory? If your leader is irreplaceable, regardless of group size, that's a single point of failure. I can't figure how a single point of failure could ever make something "stronger". I can see arguments for necessity, or efficiency, given contrivances and extreme contexts. But "stronger" doesn't seem like the assessment for whatever necessitating a single point of failure would be.

>>catapa+j1
"Stronger" is ambiguous. If you interpret it as "resilience" then I agree having a single point of failure is usually more brittle. But if you interpret it as "focused", then having a single charismatic leader can be superior.

Concretely, it sounds like this incident brought a lot of internal conflicts to the surface, and they got more-or-less resolved in some way. I can imagine this allows OpenAI to execute with greater focus and velocity going forward, as the internal conflict that was previously causing drag has been resolved.

Whether or not that's "better" or "stronger" is up to individual interpretation.

>>catapa+j1
I guess though, a lot of organizations never develop a cohesive leader at all, and the orgs fall apart. They never had an irreplaceable leader though!

>>catapa+j1
A company is essentially an optimization problem, meant to minimize / maximize some set of metrics. Usually a companies goal is simply to maximize NPV but in OpenAI's case the goal is to maximize AI while minimizing harm.

"Failure" in this context essentially means arriving at a materially suboptimal outcome. Leaders in this situation, can easily be considered "irreplaceable" particularly in the early stages as decisions are incredibly impactful.