zlacker

I see literally no reason for Sam to stay without a full board resignation and return to CEO. All other options are just downsides when he can walk, start Newco, and take everyone with him. He'd lose the restrictive governance model and gain full-control.

I think Murati is actually on team Altman, but that just makes me think that he should walk even more. Take Murati and start Newco with the exact same org chart.

>>Shank+(OP)
Lmao start from scratch when they have gpt4 already?

>>Shank+(OP)
> take everyone with him

That's always the assumption people are going with, but is it true?

>>Shank+(OP)
> Newco

NextAI

>>wnoise+v
Ilya is the irreplaceable one there though …

I know it is to much to wish for, but I hope Sam and Ilya reconcile their differences. They are the most obvious example of 1+1>2.

>>wnoise+v
It must be true or they wouldn’t be considering bringing him back so quickly.

>>m3kw9+c
My gut has the same feeling, but to put numbers on it, for a 1.76 trillion parameter model...

- It takes 10s of millions of dollars in GPU time for training?

- Curation of data to train on

- Maybe 10s of thousands of man hours for reinforcement?

- How many lines of code are written for the nets and data pipelines?

Does anyone have any insight on these numbers?

>>wnoise+v
I won't say everyone, but 1/3 of people with Sam will be a disastrous blow to existing OpenAI as a company.

>>option+P
Not irreplaceable. There are many other very good, very qualified machine learning researchers, and Ilya relies on the people under him just as much.

>>consum+Z
100M for GPT-4 I think. They would be in a strong position to negotiate a discount on that GPU bill though.

>>option+P
Is he, though? Genuine question - I am sure the Chief Scientist (or equivalent) of Xerox Parc was a brilliant person, but if we take our engineer blinders off, a successful birthing of products that millions benefit from is from more than just the genius of the tech/product person.

>>Shank+(OP)
> He'd lose the restrictive governance model and gain full-control.

He'd lose the dataset, which is by far the most valuable thing they have. The genie is out of the bottle and making such a dataset again is not going to be easy or cheap (or maybe even legal).

>>option+P
Ilya is done. This is a clear instance of the chef who made the secret sauce breaking down after the restaurant expanded. Except now, regardless of his contributions, there's enough people making similar sauce and not poisoning it that it'd be better if he went back to some idealistic mom&pop so that the business could bring in line cooks.

>>Shank+(OP)
Everyone says Sam can start a newco so easily, but he'd lose the data, codebase and deployed infrastructure, and a lot of knowledge. If people think it's devastating for OpenAI to lose 1/3 of its employees, Sam would basically be starting with the equivalent of an OpenAI that lost 2/3, along with all the other problems. Of course, with enough time and money he can recover, but I don't think it would be that easy. And to Sam, such a slowdown when he had so much momentum before may be the hardest thing to accept.

>>Eji170+v1
Or even possible. They won't have twitter api access that they did until recently. And they certainly won't have the gigantic dataset ChatGPT is collecting.

>>m3kw9+c
Yes. GPT-4 is surely a drop in the hat compared to the endgame. Each generation of hardware becomes more and more capable, and hardware is what matters in the end. If anything starting fresh would likely pay off in an enormous way for him, as cash would be easy to raise, and he would be able to maintain his position indefinitely without risk of the board losing their minds.

Get NVidia, AMD, or Apple to help fund the new entity and/or get some chip designers on board to push things further than OpenAI can without reaching into Microsoft’s pocket. A pocket I’m sure will be much tighter after the recent chicanery.

Capital would NOT a problem at this point as it’s beyond proof of concept… A normal start up, trying to prove itself, sure, but at this point Altman has proven the idea and himself at the helm. I’d also argue the dataset they used to train it, is not that relevant long term. As the data itself was agglomerated from the internet and can be had again. Even better data perhaps because the copyright holders can become investors. You really just need the capacity to deal with it, from ingestion to legal, which is a capital problem.

>>ah765+h2
Are gpus even available? Espically in the scale this hypothetical new company would need.

>>Eji170+v1
We're still in the early days of LLM training- saying that datasets are tapped is like saying we hit peak oil in 1880. This is still early days for the field, and it's not clear how efficient new training methods might become, or how small the smallest viable training set can be. There's more scrutiny now sure, but Altman was one of the people pushing for that scrutiny. He likely is capable of navigating the pit traps that would be there for competitors.

>>adastr+M
Y-TwitchAI