Hacker News Reader

Comment by kylecazar | Hacker News Reader

Here's what I don't understand.

There clearly were tensions between the for and not-for growth factions, but the Dev Day is being cited as a 'last straw'. It was a product launch.

Ilya, and the board, should have been well aware of what was being released on that day for months. They should have at the very least been privy to the plan, if not outright sanctioned it. Seems like before launch would have been the time to draw a line in the sand.

Did they have a 'look at themselves in the mirror' moment after the announcements or something?

View on HN

> They should have at the very least been privy to the plan, if not outright sanctioned it.

Never assume this. After all, their communication specifically cited that Sam deceived them in some way, and Greg was also impacted. Ilya is the only board member that might have known naturally, given his day-to-day work with OAI, but since ~July he has worked in the area of superalignment, which could reasonably be a different department (it shouldn't be). The Board may have also found out about these projects, maybe from a third party/Ilya, told Sam they're moving too fast, and Sam ignored them and launched anyway. We really don't know.

>Ilya, and the board, should have been well aware of what was being released on that day for months

Not necessarily, and that may speak to the part of the Board's announcement that Sam was not candid

I can’t imagine an organization where this wouldn’t have come up on some roadmap or prioritization meeting, etc. How could leadership not know what the org is working on?! They’re not that big.

Board is not exactly leadership. They meet infrequently and get updates directly from management, they don't go around asking employees what they're working on

True. So the CTO knew what was happening, wasn’t happy, and then coordinated with the board, is that what appears to have happened?

CTO who is now acting CEO.

Not making any accusations but that was an odd decision given that there is an OpenAI COO.

They do typically have views into strategic plans, roadmaps and product plans.

Going into detail in a talk and discussing AGI may have provided crucial context that wasn't obvious from a PowerPoint bullet point, which is all the board may have seen earlier.

More supervision than leadership...

Surely Ilya Sutskever must have known what was being worked on as Chief Scientist?

I can't imagine an organization that would fire their celebrity CEO like this either. So maybe that's how we arrived here.

Could be many things, like Sam not informing them of the GPTs store launch, or saying he won't launch and then launching.

It sucks for openAi, but there's too many hungry hungry competitors salivating at replacing OpenAI so I don't think this will have big king term consequences in the field.

I'm curious what sorts of oversight and recourse all the investors (or are they donors?) Have. I imagine there's a lot of people with a lot of money that are quite angry today.

They don’t have investors, it’s a non profit.

The “won’t anyone think of the needs of the elite wealthy investor class” that has run through the 11 threads on this topic is pretty baffling I have to admit.

They do have investors in the for-profit subsidiary, including Microsoft and the employees. Check out the diagram in the linked article.

That’s right. Which isn’t the company that just fired Sam Altman.

I take your point, but still, I don’t think it’s correct to imply that investors in the for-profit company have no sway or influence over the future of OpenAI.

I sure as shit wouldn’t wanna be on Microsoft’s bad side, regardless of my tax status.

It’s a nonprofit that controls a for-profit company, which has other investors in addition to the non-profit.

> They don’t have investors

OpenAI has investors [0].

[0] https://openai.com/our-structure

OpenAI (the nonprofit whose board makes decisions) has no investors.

the subordinate holding company and even more subordinate OpenAI Global LLC have investors, but those investors are explicitly warned that the charitable purpose of the nonprofit and not returning profits to investors is the paramount function of the organization, over which the nonprofit has full governance control.

Thanks for clarifying.

Then what did Microsoft pay for?

Privileged access to technology, which has paid off quite well for them already.

They didn't pay a fee

These people are humans, and there’s a big difference between kinda knowing the keynote was coming up, and then actually watching it happen and receive absolutely rave coverage from everyone in tech.

I could very much see it as a “look in the mirror” moment, yeah.

Let's look closer at the Ilya Sutskever vs Sam Altman tensions, and think of the product/profit as a cover.

Ilya Sutskever is a True Believer in LLMs being AGI, in that respect aligned with Geoff Hinton, his academic advisor at University of Toronto. Hinton has said "So by training something to be really good at predicting the next word, you’re actually forcing it to understand. Yes, it’s ‘autocomplete’—but you didn’t think through what it means to have a really good autocomplete"[1].

Meanwhile, Altman has decided that LLMs aren't the way.[2]

So Altman was pushing to turn the LLM into a for-profit product, to get what value it has, while the Sutskever-aligned faction thinks it is AGI, and want to keep it not-for-profit.

There's also some difference about whether or not AGI poses an "existential risk" or if the risks of current efforts at AI are along the lines of algorithmic bias, socioeconomic inequality, mis/disinformation, and techno-solutionism.

1. https://www.newyorker.com/magazine/2023/11/20/geoffrey-hinto...

2. https://www.thestreet.com/technology/openai-ceo-sam-altman-s...

You are conflating Illya's belief in the transformer architecture (with tweaks/compute optimizations) being sufficient for AGI with that of LLMs being sufficient to express human-like intelligence. Multi-modality (and the swath of new training data it unlocks) is clearly a key component of creating AGI if we watch Sutskever's interviews from the past year.

Yes, I read "Attention Is All You Need", and I understand that the multi-head generative pre-trained model talks about "tokens" rather than language specifically. So in this case, I'm using "LLM" as shorthand for what OpenAI is doing with GPTs. I'll try to be more precise in the future.

That still leaves disagreement between Altman and Sutskever over whether or not the current technology will lead to AGI or "superintelligence", with Altman clearly turning towards skepticism.

Fair enough, shame "Large Tokenized Models" etc never entered the nomenclature.

Some terms I've seen used for the technology:

Big-Data Statistical Models

Stochastic Parrots or parrot-tech

plausible sentence generators

glorified auto-complete

cleverbot

"a Blurry JPEG of the Web" <https://www.newyorker.com/tech/annals-of-technology/chatgpt-...>

and just plain ol' "machine learning"

Do you have a link to one of these talks?

They “should have” but if the board was wildly surprised by what was presented, that sounds like a really good reason to call out the CEO for lack of candor.

they could have been beefing non-publicly for a long time, and might have had many private conversations, probably not very useful to speculate here

Not useful at all.. but it sure is fun! This is gonna be my whole dang weekend.

Probably dang's whole weekend as well.

What if Enterprises get access to a much better version of AI compared to the GPT+ subscription customer?

They always were because it was going to be customised for their needs.

Reply on news.ycombinator.com