Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

Mira Murati leaves OpenAI

twitter.com

Post reply on HN
Print view View original

581–590 of 644 posts

Re: Mira Murati leaves OpenAI

#581

Post by philosopher1234 »

navjack27 wrote: post #514

Okay hear me out. Restructuring for profit right? There will probably be companies spawned off of all of these leaving. If the government ever wants a third party to oversee safety of openAI wouldn't it be convenient if one of those that left the company started a company that focused on safety. Safe Superintelligence Inc. gets the bid because lobbying because whatever I don't even care what the reason is in this mad…

These are some serious mental gymnastics. It depends on:

1. The government providing massive funds for AI safety research. There is no evidence for this. 2. Sam Altman and everyone else knowing this will happen and planning for it. 3. Sam Altman, amongst the richest people in the world, and everyone else involved, not being greedy. (Despite the massive evidence of greed) 4. San altman heroically abandoning his massive profits down the line.

Also, even in your story, Sam Altman profits wildly and is somehow also not motivated by that profit.

On the other hand, a much simpler and more realistic explanation is available: he wants to get rich.

Re: Mira Murati leaves OpenAI

#582

Post by gilmore606 »

bmitc wrote: post #524
lionkor wrote: post #130

Earlier quoted context omitted.

Maybe its marketing and LLMs are the peak of what they are capable of.

I continue to be surprised by the talk of general artifical intelligence when it comes to LLMs. At their core, they are text predictors, and they're often pretty good at that. But anything beyond that, they are decidely unimpressive. I use Copilot on a daily basis, which uses GPT 4 in the backend. It's wrong so often that I only really use it for boilerplate autocomplete, which I still have to review. I've had collea…

LLMs let the massively stupid and incompetent produce something that on the surface looks like a useful output. Most massively stupid incompetent people don't know they are that. You can work out the rest.

Re: Mira Murati leaves OpenAI

#584

Post by deepGem »

dragonelite wrote: post #487
fsndz wrote: post #481

Earlier quoted context omitted.

hypocrisy has to be the core of every corporate or political environment I have observed recently. I can count the occasions or situations where telling the simple truth is helpful. even the people who tell you to tell the truth are often the ones incapable of handling it.

From experience unless the person mention their next "adventure"(within like a couple of months) or gig it usually means a manager or c-suite person got axed and was given the option to gracefully exit.

By the barrage of exits following Mira's resignation, it does look like Sam fired her, the team got the wind of this and are now quitting in droves. This is the thing about lying and being polite. You can't hide the truth for long.

Mira's latest one liner tweet 'OpenAI is nothing without it's people" speaks volumes.

Re: Mira Murati leaves OpenAI

#585

Post by int_19h »

HarHarVeryFunny wrote: post #538
Sunhold wrote: post #529

Earlier quoted context omitted.

Look at the sample chain-of-thought for o1-preview under this blog post, for decoding "oyekaijzdf aaptcg suaokybhai ouow aqht mynznvaatzacdfoulxxz". At this point, I think the "fancy autocomplete" comparisons are getting a little untenable. https://openai.com/index/learning-to-reason-with-llms/

It depends on how well you understand how the fancy autocomplete is working under the hood. You could compare GPT-o1 chain of thought to something like IBM's DeepBlue chess-playing computer, which used MTCS (tree search, same as more modern game engines such as AlphaGo)... at the end of the day it's just using built-in knowledge (pre-training) to predict what move would most likely be made by a winning player. It's n…

Saying "it's just autocomplete" is not really saying anything meaningful since it doesn't specify the complexity of completion. When completion is a correct answer to the question that requires logical reasoning, for example, "just autocomplete" needs to be able to do exactly that if it is to complete anything outside of its training set.

Re: Mira Murati leaves OpenAI

#586

Post by mnky9800n »

snowwrestler wrote: post #572
navjack27 wrote: post #514

Okay hear me out. Restructuring for profit right? There will probably be companies spawned off of all of these leaving. If the government ever wants a third party to oversee safety of openAI wouldn't it be convenient if one of those that left the company started a company that focused on safety. Safe Superintelligence Inc. gets the bid because lobbying because whatever I don't even care what the reason is in this mad…

I think the departures and switch to for-profit model may point in a different direction: that everyone involved is realizing that OpenAI’s current work is not going to lead to AGI, and it’s also not going to change. So the people who want to work on AGI and safety are leaving to do that work elsewhere, and OpenAI is restructuring to instead focus on wringing as much profit as possible out of their current architectu…

I feel like it's unfair to expect growth to remain within your walls. bell and Xerox both drove a lot of growth. That growth just left bell and Xerox to go build things like intel and apple. They didn't keep it for themselves and that's a good thing. Could you imagine if the world was really like those old at&t commercials and at&t was actually the ones bringing it to you? I would not want a monolithic at&t providing all technology.

https://youtu.be/xBJ2KXa9c6A?si=pB67u56Apj7gdiHa

I do agree with you. They are locked into pulling value out of what they got and they probably aren't going to build something new.

Re: Mira Murati leaves OpenAI

#587

Post by spidersouris »

trashtester wrote: post #558
uhtred wrote: post #143

Earlier quoted context omitted.

Artificial General Intelligence requires a bit more than parsing and predicting text I reckon.

That's not quite how o1 was trained, they say. o1 was trained specifically to perform reasoning. Or rather, it was trained to reproduce the patterns within internal monologues that lead to correct answers to problems, particularily STEM problems. While this still uses text at some level, it's no longer regurgitation of human-produced text, but something more akin to AlphaZero's training to become superhuman at games…

> While this still uses text at some level, it's no longer regurgitation of human-produced text, but something more akin to AlphaZero's training to become superhuman at games like Go or Chess.

How did you know that? I've never seen that anywhere. For all we know, it could just be a very elaborate CoT algorithm.

Re: Mira Murati leaves OpenAI

#588

Post by ToucanLoucan »

NoGravitas wrote: post #568
ToucanLoucan wrote: post #537

Earlier quoted context omitted.

I’m not seeing anything convincing here. OpenAI says that it’s models are better at reasoning and asserts they are testing this by comparing how it does solving some problems between o1 and “experts” but it doesn’t show the experts or o1s responses to these questions nor does it even deign to share what the problems are. And, crucially, it doesn’t specify if writings on these subjects were part of training data. Call…

"Any sufficiently advanced technology is indistinguishable from a rigged demo." A corollary of Clarke's Law found in fannish circles, origin unknown.

Especially given this tech's well-documented history of using rigged demos, if OpenAI insists on doing and posting their own testing and absolutely nothing else, a little insight into their methodology should be treated as the bare fucking minimum.

Re: Mira Murati leaves OpenAI

#589

Post by karlzt »

leloctai wrote: post #388
layer8 wrote: post #39

Plain-text version for those who can’t read images: “ Hi all, I have something to share with you. After much reflection, I have made the difficult decision to leave OpenAI. My six-and-a-half years with the OpenAI team have been an extraordinary privilege. While I'll express my gratitude to many individuals in the coming days, I want to start by thanking Sam and Greg for their trust in me to lead the technical organiz…

Doesn't seems like it was written by ChatGPT. I find that amusing somehow.

Perhaps some parts were written by ChatGPT, probably mixed up.

Re: Mira Murati leaves OpenAI

#590

Post by HarHarVeryFunny »

int_19h wrote: post #585
HarHarVeryFunny wrote: post #538

Earlier quoted context omitted.

It depends on how well you understand how the fancy autocomplete is working under the hood. You could compare GPT-o1 chain of thought to something like IBM's DeepBlue chess-playing computer, which used MTCS (tree search, same as more modern game engines such as AlphaGo)... at the end of the day it's just using built-in knowledge (pre-training) to predict what move would most likely be made by a winning player. It's n…

Saying "it's just autocomplete" is not really saying anything meaningful since it doesn't specify the complexity of completion. When completion is a correct answer to the question that requires logical reasoning, for example, "just autocomplete" needs to be able to do exactly that if it is to complete anything outside of its training set.

It's just a shorthand way of referring to how transformer-based LLMs work. It should go without saying that there are hundreds of layers of hierarchical representation, induction heads at work, etc, under the hood. However, with all that understood (and hopefully not needed to be explicitly stated every time anyone wants to talk about LLMs in a technical forum), at the end of the day they are just doing autocomplete - trying to mimic the training sources.

The only caveat to "just autocomplete" (which again hopefully does not need to be repeated every time we discuss them), is that they are very powerful pattern matchers, so all that transformer machinery under the hood is being used to determine what (deep, abstract) training data patterns the input pattern best matches for predictive purposes - exactly what pattern(s) it is that should be completed/predicted.

Post reply on HN