zlacker

[parent] [thread] 24 comments
1. 015a+(OP)[view] [source] 2023-11-19 00:11:46
Bad take. Not "everyone" feels that what they did was wrong. We don't have insight into what's going on internally. Optics matter; the division over their decision means that its definitionally non-obvious what the correct path forward is; or, that there isn't one correct path, but multiple reasonable paths. To admit a mistake of this magnitude is to admit that you're either so unprincipled that your mind can be changed at a whim; or that you didn't think through the decision enough preemptively. These are absolutely signs of weakness in leadership.
replies(4): >>peyton+L2 >>vikram+84 >>tick_t+q5 >>hn_thr+ow
2. peyton+L2[view] [source] 2023-11-19 00:27:22
>>015a+(OP)
Satya is “furious.” What’s reasonable about pissing off a guy who can pull the plug? I don’t think it’s definitionally non-obvious whether to take that risk.
replies(2): >>option+d4 >>no_wiz+z9
3. vikram+84[view] [source] 2023-11-19 00:35:59
>>015a+(OP)
Whether or not you agree with the decision they obviously screwed up the execution something awful. This is humiliating for them and honestly setting altman free like they did was probably the permanent end of AI safety. Just take someone with all the connections and the ability to raise billions of dollars overnight and set them free without any of the shackles of AI ethics people in a way that makes all the people with money want to support him? That's how you get skynet
replies(1): >>015a+HI
◧◩
4. option+d4[view] [source] [discussion] 2023-11-19 00:36:17
>>peyton+L2
Yeah, he can be furious all he wants but he is not getting the OpenAI he used to have back. It’s either Sam + Greg now or Ilya. All 3 are irreplaceable.
5. tick_t+q5[view] [source] 2023-11-19 00:44:53
>>015a+(OP)
> Bad take. Not "everyone" feels that what they did was wrong.

But everyone important does so who cares about the rest?

replies(1): >>no_wiz+X5
◧◩
6. no_wiz+X5[view] [source] [discussion] 2023-11-19 00:48:42
>>tick_t+q5
You mean the “the rest” as in the people who execute on the company vision?

It’s really dismissive toward the rank and file to think that they don’t matter at all.

replies(2): >>threes+m8 >>hn_thr+Gu
◧◩◪
7. threes+m8[view] [source] [discussion] 2023-11-19 01:03:00
>>no_wiz+X5
a) The company vision up until this point included commercial products.

b) Altman personally hired many of the rank and file.

c) OpenAI doesn't exist with customers, investors or partners. And in this one move the board has alienated all three.

replies(1): >>no_wiz+W8
◧◩◪◨
8. no_wiz+W8[view] [source] [discussion] 2023-11-19 01:08:30
>>threes+m8
I seriously doubt customers or (most) partners care about this. I have yet to hear of a single customer or partner leave the service, and I do not believe it to be likely. Simply, unless they shut down their offerings on Monday they will have their customers.

Investors care, but if new management can keep the gravy track, they ultimately won’t care either.

Companies pivot all the time. Who is to say the new vision isn’t favored by the majority of the company?

replies(3): >>threes+of >>wwtrv+ej >>kcb+8n
◧◩
9. no_wiz+z9[view] [source] [discussion] 2023-11-19 01:12:32
>>peyton+L2
Last I checked he only had 49% of the company.

I also feel, that they can patch relationships, Satya may be upset now but will he continue to be upset on Monday?

It needs to play out more before we know, I think. They need to pitch their plan to outside stakeholders now

replies(1): >>discor+we
◧◩◪
10. discor+we[view] [source] [discussion] 2023-11-19 01:51:12
>>no_wiz+z9
Which other company will give them the infra/compute they need when 49% of the profitable part has been eaten up?
replies(2): >>threes+Sf >>no_wiz+ei
◧◩◪◨⬒
11. threes+of[view] [source] [discussion] 2023-11-19 01:57:19
>>no_wiz+W8
The fact that this happened so soon after Developer Day is a clear signal that the board wasn't happy with that direction.

Which is why every developer/partner including Microsoft is going to be watching this situation unfold with trepidation.

And I don't know how you can "keep the gravy track" when you want the company to move away from commercialisation.

◧◩◪◨
12. threes+Sf[view] [source] [discussion] 2023-11-19 01:59:55
>>discor+we
And how will they survive if Microsoft/SamAi ends up building a competitor ?

Microsoft could run the entire business as a loss just to attract developers to Azure.

replies(1): >>no_wiz+Kh
◧◩◪◨⬒
13. no_wiz+Kh[view] [source] [discussion] 2023-11-19 02:10:40
>>threes+Sf
That assumes Altman competitor can outpace and outclass OpenAI and maybe it can. I know Anthropic came about from earlier disagreements and that didn’t slow OpenAIs innovation pace, certainly.

Everything just assumes that without Sam they’re worse off.

But what if, my gosh, they aren’t? What if innovation accelerates?

My point being is it’s useless to speculate that Altman starting a new business competing with OpenAI will be successful inherently. There’s more to it than that

replies(3): >>qwytw+vl >>palebl+Cv >>int_19+cN
◧◩◪◨
14. no_wiz+ei[view] [source] [discussion] 2023-11-19 02:13:38
>>discor+we
First it remains to be seen if Microsoft is going to do something drastic.

I also suspect they could very well secure this kind of agreement from another company that would be happy to play ball for access to OpenAI tech. Perhaps Amazon for instance, who’s AI attempts since Alexa have been lackluster

◧◩◪◨⬒
15. wwtrv+ej[view] [source] [discussion] 2023-11-19 02:18:23
>>no_wiz+W8
> I have yet to hear of a single customer or partner leave the service

Which doesen't mean a lot. Of course they'd wait for this to play out before committing to anything.

> but if new management can keep the gravy track

I got the vague impression that this whole thing was partially about stopping the gravy train? In any case Microsoft won't be too happy about being entirely blindsided (if that was the case) and probably won't really trust the new management.

◧◩◪◨⬒⬓
16. qwytw+vl[view] [source] [discussion] 2023-11-19 02:32:10
>>no_wiz+Kh
> Everything just assumes that without Sam they’re worse off.

But it's not just him is it?

replies(1): >>no_wiz+jn
◧◩◪◨⬒
17. kcb+8n[view] [source] [discussion] 2023-11-19 02:42:01
>>no_wiz+W8
The new management has declared that their primary goal in all this was to stop the gravy track.
replies(1): >>no_wiz+cs
◧◩◪◨⬒⬓⬔
18. no_wiz+jn[view] [source] [discussion] 2023-11-19 02:42:55
>>qwytw+vl
Sure, I suppose not, but they aren’t losing everyone en masse. Simply Altman supporters so far.

I think a wait and see approach is better. I think we had some inner politics spill public because Altman needs to the public pressure to get his job back, if I was speculating

◧◩◪◨⬒⬓
19. no_wiz+cs[view] [source] [discussion] 2023-11-19 03:14:01
>>kcb+8n
I don’t think there has been a formal announcement on the new direction yet
◧◩◪
20. hn_thr+Gu[view] [source] [discussion] 2023-11-19 03:34:10
>>no_wiz+X5
> It’s really dismissive toward the rank and file to think that they don’t matter at all.

I had the exact opposite take. If I were rank and file I'd be totally pissed how this all went down, and the fact that there are really only 2 possible outcomes:

1. Altman and Brockman announce another company (which has kind of already happened), so basically every "rank and file" person is going to have to decide which "War of the Roses" team they want to be on.

2. Altman comes back to OpenAI, which in any case will result in tons of time turmoil and distraction (obviously already has), when most rank and file people just want to do their jobs.

◧◩◪◨⬒⬓
21. palebl+Cv[view] [source] [discussion] 2023-11-19 03:40:03
>>no_wiz+Kh
> Everything just assumes that without Sam they’re worse off. > > But what if, my gosh, they aren’t? What if innovation accelerates?

It reads like they ousted him because they wanted to slow the pace down, so by design and intent it would seem unlikely innovation would accelerate. Which seems doubly bad if they effectively spawned a competitor that is made up by all the other people that wanted to move faster

22. hn_thr+ow[view] [source] 2023-11-19 03:44:25
>>015a+(OP)
> These are absolutely signs of weakness in leadership.

The signs of "weakness in leadership" by the board already happened. There is no turning back from that. The only decision is how much continued fuck-uppery they want to continue with.

Like others have said, regardless of what is the "right" direction for OpenAI, the board executed this so spectacularly poorly that even if you believe everything that has been reported about their intentions (i.e. that Altman was more concerned about commercializing and productization of AI, while Sutskever was worried about the developing AI responsibly with more safeguards), all they've done is fucked over OpenAI.

I mean, given the reports about who has already resigned (not just Altman and Brockman but also other many other folks in top engineering leadership), it's pretty clear that plenty of other people would follow Altman to whatever AI venture he wants to build. If another competitor leap frogs OpenAI, their concerns about "moving too fast" will be irrelevant.

◧◩
23. 015a+HI[view] [source] [discussion] 2023-11-19 05:21:39
>>vikram+84
I tend to think: We, the armchair commentators, do not know what happened internally. I don't know enough to know that the board's execution wasn't the best case scenario to achieve their goal of aligning the entire organization with the non-profit's mission. All I feel comfortable saying with certainty is that: its messy. Anything like this would inevitably be messy.
replies(1): >>vikram+hS
◧◩◪◨⬒⬓
24. int_19+cN[view] [source] [discussion] 2023-11-19 06:07:06
>>no_wiz+Kh
The thing I really want to know is how many of the people who have already quit or have threatened to quit are actual researchers working on the base model, like Sutskever.
◧◩◪
25. vikram+hS[view] [source] [discussion] 2023-11-19 06:58:30
>>015a+HI
Right and thats what I'm saying. It's messy. They screwed up. Messy is bad. If they needed to get rid of him this last minute and make a statement 30 minutes before market close, then the failure happened earlier.
[go to top]