41 comments

[ 3.3 ms ] story [ 98.2 ms ] thread
> During our interview, Altman repeatedly declined to answer the main question on everyones’ minds: exactly why he was fired to begin with.
During the interview, he also explained frankly why he declined. As the CEO and face of OpenAI, you don't want to have a position that is inconsistent with the organization's position, which is coming later after the review.
Lol.

Not to worry. He did not take any position on anything during this interview.

It was hilariously devoid of positions or information.

This article could have been a one-liner: "Had an interview with Altman, but he wouldn't say anything he hadn't already said on Twitter."

The credibility of everyone who was involved in this imbroglio is shot for me until I get some receipts for this. I personally doubt RLHF can scale anyway.

Altman recently gave an interview at Oxford (right before all of this drama) of all places where, while answering an unrelated question about compute costs, he mentioned that he “seriously thinks” we’ll achieve fusion in the next few years. The recklessness and silliness of that comment really struck me. It made me wonder how much he’s applying this kind of lazy, undisciplined cognition to his own work.

I haven’t seen hype, drama, and speculation like this since crypto. Someone, please, bring this back down to earth. Lately I’m beginning to feel like we’ve crossed over into some weird Twilight Zone where HBO’s Silicon Valley has moved from entertaining parody to cutting social commentary.

You see the same kind of thinking here on HN all the time. There's a few generations of us that were brought up on a kind of teleological notion of technological progress, inspired by the insanely massive late 20th century advances in aerospace, then semiconductors, and computer science, and being brought up in a triumphalist west where the G7 had buried its economic and political foes etc. In that context, with a heavy dose of science fiction and a heap of ideology, it somewhat makes sense that there's people who see these kinds of things (fusion in 5 years! AGI right around the corner!) as not only posisble, but inevitable.

But the magic of Moore's law in transistor technology was/is accomplished through insanely hard nuts and bolts engineering and science by dedicated smart people. It's not investors and techbros and writers and pundits that make this stuff happen. It's just grit and hard work and continuous incremental improvements.

There's no magic that makes technology grow in exponential terms. Moore's law doesn't apply to anything but transistors. "Progress" is not inevitable. Technology doesn't change the world -- people do. Practical fusion energy doesn't happen just because Altman's friends sprinkle cash all over it and talk it up.

Anyways, yeah, just ranting to agree with you. I am getting so sick of this crap.

Where are the promised cities on the Moon?

ChatGPT may be the Moon-landing event: highly inspirational, highly expensive, modest (relative to the hype) actual practical value long term.

Man, the popular science stuff I remember ingesting as a young child in the late 70s and early 80s... we were supposed to have the whole solar system colonized by now...
Realistically, this is an interview that should never have happened. Giving non answers non-stop isn't that interesting. It's understandable when there are contractual obligations for press availability such as in sports but for a CEO of a company who can just say no (he could have, no?) it makes no sense.
> Realistically, this is an interview that should never have happened. Giving non answers non-stop isn't that interesting.

But it is telling.

Honestly I'm confused why even post an interview on the topic if the most obvious question still hasn't been answered.
Because the audience doesn't know that until after ads have been displayed.
> Q: The reports about the Q* model breakthrough that you all recently made, what’s going on there?

> A: Altman: No particular comment on that unfortunate leak.

Interesting that he calls it an unfortunate leak.

Sounds like a slimy way to manipulating current situation and distancing people from some real problem.
Isn't this tacit verification that Q* is real?
Maybe it's Q as in Quantum not Q-Learning...
No, because basic common sense says that to have quantum breakthrough with AI they'll need massive quantum compute which... well doesn't exist yet.

Their research is AI built on conventional neural networks (the way they've been since literally the 1960s), not quantum physics.

:"You can always hit a wall, but we expect that progress will continue to be significant. "

Wasn't there interview ~2 months ago when speaking about government regulations that he said that they were expecting a plateau in abilities? To not worry, they were hitting the limits with LLM's?

Maybe they're working on more than LLMs (or at least, the Transformer architecture)?
making tons of valuable progress on lucrative prospects, but absolutely no progress on anything worth regulating.

progress is contextual.

That has to be the most words that add up to “no comment” ever written. There’s no reason to even click “publish” on that article beyond getting clicks to sell ad impressions.
Hey now. The article demonstrates Altmans sincere and heartfelt commitment to "no comment".
Thank you for saving my time.
Finding out that people actually don’t comment is also journalism.
To be honest I extracted only one useful piece of information here.

"No particular comment on that unfortunate leak"

The word unfortunate for me sounds significant. If it was random noise as so much we heard recently, such a word wouldn't be used and so much emphasis wouldn't be given to this paragraph, right?

Gee I wonder who ordered the leak ;)

Just more fuel for the hype machine to maintain the valuation

That "unfortunate" probably has been deliberately put there by Sam, possibly for the reason you said.
He probably wants you to extract exactly that. Which means you can’t really deduce anything useful from it.
The fact that he called back to share that the team has been doing very well without him is great leadership!
It's flattery of the team, but it also undermines all the _stum und drang_ over his firing 'destroying' OA, if the leadership team was so wonderful that his absence would make no difference. If that is the case, it wasn't his firing that would destroy OA, but his reaction to it...
Question: The reports about the Q* model breakthrough that you all recently made, what’s going on there?

Altman: "...No particular comment on that unfortunate leak. But what we have been saying — two weeks ago, what we are saying today, what we’ve been saying a year ago, what we were saying earlier on — is that we expect progress in this technology to continue to be rapid, and also that we expect to continue to work very hard to figure out how to make it safe and beneficial..."

It’s almost as if this is all a distraction and sandbag for a breakthrough that couldn’t be achieved.
the semantic content was null.
Q* is a strange thing, youtube is full of videos explaining what q* is, but nobody really knows. All those ideas are combinations of reinforcement learning or path finding with language models. It doesn’t sound like something that would be a new idea, but much of the progress in AI has been just based on compute growing.

If there has been a big leap, the thing that worries me most is the new board of open AI and what that means about how they intend to use the power they now hold. We had a board that felt so strongly about something that they acted swiftly to remove Altman and now he is back. The previous board seemed to hold social responsibility and safety as the priority. The new board will have a different outlook based on its members past statements.

If I had to guess, Q* would involve solving problems quantum-style, by breaking them down along dimensions (not step-by-step!!), then solving the subtasks in parallel and recombining.

Basically, most of the LLM intelligence is at the word-to-word selection level, so friction needs to be created there instead of at the task or question level.

For all the negativity here, I enjoyed the article. There are also some details here, personal or about the timeline.