Can someone make predictions on the rate of progress of Sam’s newco.
Will it exist?
My take: 95%+
Will it surpass OpenAI for SoTA model progress?
My take: 5% if timelines are very short, 65-70% otherwise
Corporate structure:
Probably similar to anthropic on the scale of non profit to for profit.
Safety adherence:
More than Perplexity/Inflection, less than OpenAI, Anthropic, DeepMind
Some justifications for the above:
Sam seems to want to make it a standard growth focused tech company but obviously believes in X-risk pretty seriously.
A couple of researchers seem to have resigned, if the bod can’t come up with a compelling reason for their decision, I expect a lot of other researchers to follow, especially outside of the superalignment team.
I’m super excited to have you join as CEO of this new group, Sam, setting a new pace for innovation. We’ve learned a lot over the years about how to give founders and innovators space to build independent identities and cultures within Microsoft, including GitHub, Mojang Studios, and LinkedIn, and I’m looking forward to having you do the same.
Of course, a feel of being within a big corp can easily kill any creativity whatsoever, but they do a good job shielding people from that, they can move very fast even conditional on short timelines.
Of course, if they actually start competing for SOTA, safety will become a big issue, there is a danger of them being less careful than OpenAI has been so far, so what has happened is not necessarily a win for safety.
Can someone make predictions on the rate of progress of Sam’s newco.
Will it exist? My take: 95%+
Will it surpass OpenAI for SoTA model progress? My take: 5% if timelines are very short, 65-70% otherwise
Corporate structure:
Probably similar to anthropic on the scale of non profit to for profit.
Safety adherence:
More than Perplexity/Inflection, less than OpenAI, Anthropic, DeepMind
Some justifications for the above:
Sam seems to want to make it a standard growth focused tech company but obviously believes in X-risk pretty seriously.
A couple of researchers seem to have resigned, if the bod can’t come up with a compelling reason for their decision, I expect a lot of other researchers to follow, especially outside of the superalignment team.
Edit: 1. was wrong. I’m shocked
Not quite wrong, just a “post-acquisition corporate structure” already ;-)
Now, a lot depends of how close to reality this tweet will be: https://twitter.com/satyanadella/status/1726516824597258569
Of course, a feel of being within a big corp can easily kill any creativity whatsoever, but they do a good job shielding people from that, they can move very fast even conditional on short timelines.
Of course, if they actually start competing for SOTA, safety will become a big issue, there is a danger of them being less careful than OpenAI has been so far, so what has happened is not necessarily a win for safety.