Details emerge of surprise board coup that ousted CEO Sam Altman at OpenAI

Post by sicariusnoctis » Sun, Nov 19, 2023, 8:45 AM UTC

devbent wrote: post #653

true_religion wrote: post #575

Earlier quoted context omitted.

What if AI self improvement is not exponential? We assume a self improving AI will lead to some runaway intelligence improvement but if it grows at 1% per year or even per month that’s something we can adapt to.

Assume the AGI has access to a credit card and it goes ahead and reserves itself every GPU cycle in existence so it's 1 month is turned into a day, and now we're back to being fucked. Maybe an ongoing GPU shortage is the only thing that'll save us!

Post by fakedang » Sun, Nov 19, 2023, 9:28 AM UTC

peyton wrote: post #659

fakedang wrote: post #657

Earlier quoted context omitted.

That's true, but Sutskever and Co still have the head start. On the models, the training data, the GPT4 licenses, etc. Their Achilles heel is the compute which Microsoft will pull out. Khosla Ventures and Sequoia may sell their Open AI stakes at a discount, but I'm sure either Google or Amazon will snap it up. All Sam and Greg really have is the promise of building a successful competitor, with a big backing from Mic…

It doesn’t sound like Sutskever is running anything. OpenAI reportedly put out a memo saying they’re trying to get Sam and Greg back: https://www.theinformation.com/articles/openai-optimistic-it...

Post by limpanz » Sun, Nov 19, 2023, 9:41 AM UTC

Chamix wrote: post #202

cratermoon wrote: post #68

Earlier quoted context omitted.

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’…

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.

Post by quadcore » Sun, Nov 19, 2023, 10:36 AM UTC

vkou wrote: post #491

quadcore wrote: post #385

Earlier quoted context omitted.

Not at the same level of quality. For example, their game, ultima underworld if my memory doesnt fault me, didnt have sub-pixel precision for texturing. Their texturing was a lot uglier and unpolished compared to Wolf and especially Doom. I remember I checked, they were behind. And their game crashed. Never saw Doom crash, not even once.

> Not at the same level of quality. Engine quality? No. In terms of systems? Design? Storytelling? LGS games were way ahead of their time, and have vastly more relevance than anything post-Romero ID made.

Post by jacquesm » Sun, Nov 19, 2023, 10:52 AM UTC

chris_wot wrote: post #648

jacquesm wrote: post #634

Earlier quoted context omitted.

I had this exact thing happen a few weeks ago in a company that I have invested in. That didn't quite pan out in the way the would-be coup party likely intended. To put it mildly.

Dear god that sounds interesting and yet terrifying.