Hacker News Reader

Comment by lacker | Hacker News Reader

It's easy to have missed this part of the story in all the chaos, but from the NYTimes in March:

Ms. Murati wrote a private memo to Mr. Altman raising questions about his management and also shared her concerns with the board. That move helped to propel the board’s decision to force him out.

https://www.nytimes.com/2024/03/07/technology/openai-executi...

It should be no surprise if Sam Altman wants executives who opposed his leadership, like Mira and Ilya, out of the company. When you're firing a high-level executive in a polite way, it's common to let them announce their own departure and frame it the way they want.

View on HN

Greg Brockman, OpenAI President and co-founder is also on extended leave of absence.

And John Schulman, and Peter Deng are out already. Yet the company is still shipping, like no other. Recent multimodal integrations and benchmarks of o1 are outstanding.

> Yet the company is still shipping, like no other

If executives / high level architects / researchers are working on this quarter's features something is very wrong. The higher you get the more ahead you need to be working, C-level departures should only have an impact about a year down the line, at a company of this size.

Funny, at every corporation I've worked for, every department was still working on last quarter's features. FAANG included.

That’s exactly what they were saying. The department are operating behind the executives.

This is a good point. I had not thought of it this way before.

C-level employees are about setting the company's culture. Clearing out and replacing the C-level employees ultimately results in a shift in company culture, a year or two down the line.

You may find that this is true in many companies.

> the company is still shipping, like no other

Meta, Anthropic, Google, and others all are shipping state of the art models.

I'm not trying to be dismissive of OpenAI's work, but they are absolutely not the only company shipping very large foundation models.

Indeed Anthropic is just as good, if not better in my sample size of one. Which is great because OpenAI as an org gives shady vibes - maybe it's just Altman, but he is running the show.

Claude is pretty brilliant.

Perhaps you havent tried o1-preview or advanced voice if you call all the rest SOTA.

If only they’d release the advanced voice thing as an API. Their TTS is already pretty good, but ai wouldn’t say no to an improvement.

VP Research Barret Zoph and Chief Research Officer Bob McGrew also announced their departures this evening.

Greg’s wife is pretty sick. For all we know this is unrelated to the drama.

Sorry to hear that, all the best wishes to them.

Context (I think): https://x.com/gdb/status/1744446603962765669

Big fan of Greg, and I think the motivation behind AGI is sound here. Even what we have now is a fantastic tool, if people decide to use it.

> Yet the company is still shipping, like no other.

I don't see it for OpenAI, I do see it for the competition. They have shipped incremental improvements, however, they are watering down their current models (my guess is they are trying to save on compute?). Copilot has turned into garbage and for coding related stuff, Claude is now better than gpt-4.

Honestly, their outlook is bleak.

Yeah, I have the same feeling. It seems like operating GPT-4 is too expensive, so they decided to call it "legacy" and get rid of it soon, and instead focus on cheaper/faster 4o, and also chain its prompts to call it a new model.

I understand why they are doing it, but honestly if they cancel GPT-4, many people will just cancel their subscription.

Past efforts leds to today's products. We need to wait to see the real imapct on the ability to ship.

In my humble opinion you're wrong, Sora and 4o voice are months old and no signs they're not vaporware, and they still haven't shipped a text model on par with 3.5 sonnet!

> like no other

Really? Anthropic seems to be popping off right now.

Kagi isn’t exactly in the AI space, but they ship features pretty frequently.

OpenAI is shipping incremental improvements to its chatgpt product.

"popping off" means what?

Modern colloquialism generally meaning Moving/advancing/growing/gaining popularity very fast

Are they? In my recent experience, ChatGPT seems to have gotten better than Claude again. Plus their free limit is more strict, so this experience is on the free account.

Its just tribalism. People tend to find a team to root for when there is a competition. Which one is better is subjective at this point imo.

Well, I use both of them all of the time, so I'm in a better position to compare them. Sometimes, I tend to prefer one, like ChatGPT after Claude decreased the free quota, but I still choose it sometimes, because I still find value in its answers.

The features shipped by Anthropic in the past month are far more practical and provide clear value for builders than o1's chain of thought improvements.

- Prompt Cache, 90% savings on large system prompts for 5 mins of calls. This is amazing

- Contexual RAG, while not ground breaking idea, is important thinking and method for better vector retrieval

Quite interesting that this comment is downvoted when the content is factually correct and pertinent.

It's a very relevant fact that Greg Brockman recently left on his own volition.

Greg was aligned with Sam during the coup. So, the fact that Greg left lends more credence to the idea that Murati is leaving on her own volition.

> It's a very relevant fact that Greg Brockman recently left on his own volition.

Except that isn’t true. He has not resigned from OpenAI. He’s on extended leave until the end of the year.

That could become an official resignation later, and I agree that that seems more likely than not. But stating that he’s left for good as of right now is misleading.

> Quite interesting that this comment is downvoted when the content is factually correct and pertinent.

>> Yet the company is still shipping, like no other.

this is factually wrong. Just today Meta (which I despise) shipped more than openAI in a long time.

> When you're firing a high-level executive in a polite way, it's common to let them announce their own departure and frame it the way they want.

You also give them some distance in time from the drama so the two appear unconnected under cursory inspection.

To be fair she was also one of the employees who signed the letter to the board demanding that Altman be reinstated or she would leave the company.

Does that actually mean anything? Didn't 95% of the company sign that letter, and soon afterwards many employees stated that they felt pressured by a vocal minority of peers and supervisors to sign the letter? E.g. if most executives on her level already signed the letter, it would have been political suicide not to sign it

She was second-in-command of the company. Who else is there on her level to pressure her to sign such a thing, besides Sam himself?

Isn’t that even worse? You write to the board, they take action on your complaints, and then you change your mind?

It means when she was opting for the reinstating of Altman, she didn't have all the information needed to make a decsion

Now that she's seen exactly what prompted the previous board to fire Altman, she fires herself because she understands their decision now.

Exactly, Sam Altman wants group think, no opposition, no diversity of thought. That's what petty dictators demand. This spells the end of OpenAI IMO. Huge amount of money will keep it going until it doesn't

Reply on news.ycombinator.com