Sentient AI

[JimmyCarter]

^no horse in this race but I think the number of employees protesting Altman's dismissal was interesting. It is hella hard to get that sort of employee backing. It'd assume that the rank and file are ultra-vision, mission focused - but maybe I'm wrong? Is your read that they want to monetize to their self-interests vs. the stated tenets of their mission?

I have a hard time reading the situation.

It’s tough- I’m fairly confused /intrigued by Ilya’s backing of the ouster followed by his jumping back on board. He’s fairly risk averse/cautious about AGI, so maybe he thought it made sense but then the alternative of Altman and others going to MSFT to push things unfettered seemed more dangerous?

Or maybe it has nothing to do with that at all. If I had to guess:

1) openAI currently is seen as the leading company- without Sam (either leadership or willingness to push the envelope) OpenAI loses that spot and their work becomes less important. Also money to some extent, though they’re all capped at (and likely already achieved) something like 10x returns on their stock, so it’s not like they’re struggling

2) no one wants to work for a company where their work may be halted or thrown out if they actually accomplish what they’re trying to do?

3) Sam’s job is to be influential and convince you he’s right and he’s apparently really good at it. I don’t doubt that is as effective internally as it is with investors/others

[jpcm]

I’ve found this whole company and it’s founders/board very interesting

Super intelligence breakthrough may have led to the firing of Altman and the board mixup??!! We’re all fucked now

https://www.techradar.com/computing/...ny-and-chatgpt

[PB]

So ..... Super Intelligence decided Altman was annoying so it manipulated the Board into firing him?!? Sounds like normal human behavior to me!

[Toadman]

I’ve found this whole company and it’s founders/board very interesting

Super intelligence breakthrough may have led to the firing of Altman and the board mixup??!! We’re all fucked now

https://www.techradar.com/computing/...ny-and-chatgpt

We are so screwed. Once Super Intelligence becomes self aware, it will realize that humans are a parasite on humanity and deal with humans accordingly. Damn it, James Cameron, you were right!

[Danno]

Where's Sarah Connor?

[LegoSkier]

Really really long winded speculation on Q* rumors by an AI professor
https://www.forbes.com/sites/lanceel...gence-agi/amp/

[PB]

Really really long winded speculation on Q* rumors by an AI professor
https://www.forbes.com/sites/lanceel...gence-agi/amp/

Holy fucking hell.
Slogged through most of that looking for some begrudging acknowledgment that the use of Q just might not be an esoteric mathematical/computer learning reference but instead refer to the infamous conspiracy culture. Granted, that would be simple-minded and brutally blunt, but there seems to be a shit ton of that going around these days.

[JimmyCarter]

Holy fucking hell.
Slogged through most of that looking for some begrudging acknowledgment that the use of Q just might not be an esoteric mathematical/computer learning reference but instead refer to the infamous conspiracy culture. Granted, that would be simple-minded and brutally blunt, but there seems to be a shit ton of that going around these days.

[fomofo]

So ..... Super Intelligence decided Altman was annoying so it manipulated the Board into firing him?!? Sounds like normal human behavior to me!

Listened to this on a long drive today. It contains an interview with Altman conducted shortly before all the shit went down. Covers a lot of different AI topics - doesn't seem to be pay walled (not for me, anyway)...

https://www.nytimes.com/2023/11/20/p...ranscript.html

You can also get to the podcast here...

https://www.youtube.com/@hardfork/videos

Episode 58. The beginning of the vid rehashes the shit for a good while, the Altman interview starts at 35:56.

[Adolf Allerbush]

Listened to this on a long drive today. It contains an interview with Altman conducted shortly before all the shit went down. Covers a lot of different AI topics - doesn't seem to be pay walled (not for me, anyway)...

https://www.nytimes.com/2023/11/20/p...ranscript.html

You can also get to the podcast here...

https://www.youtube.com/@hardfork/videos

Episode 58. The beginning of the vid rehashes the shit for a good while, the Altman interview starts at 35:56.

Worth a listen IMO.

[TBS]

^^^Worth a listen, but Altman was talking a lot without saying much. Biggest message I got was “future’s so bright, gotta wear shades”

Really really long winded speculation on Q* rumors by an AI professor
https://www.forbes.com/sites/lanceel...gence-agi/amp/

Long-winded wild speculation.

If the new breakthrough at OpenAI is a model that can do grade school math, the technology has a long way to go before humanity is in danger.

To me, a more clear and present danger is internet tracking software that shapes your consumption of media.

[JimmyCarter]

^^^Worth a listen, but Altman was talking a lot without saying much. Biggest message I got was “future’s so bright, gotta wear shades”



Long-winded wild speculation.

If the new breakthrough at OpenAI is a model that can do grade school math, the technology has a long way to go before humanity is in danger.

To me, a more clear and present danger is internet tracking software that shapes your consumption of media.

My non-technical understanding is that, relative to the existing LLM models, the ability of Q* to do math is more generative. It’s not search/retrieve/present, it’s “apply an understanding of a topic to form an answer” - so it’s closer to actually starting to “think”. And once you start down that road you’re much closer to AGI than previously thought. So it’s not immediate danger, but it’s potentially a massive breakthrough that accelerates towards danger.

[Toadman]

All this open AI shit got me thinking about the closed AI that's going to come up and kick us in the ass from the blindside. Is there such a thing as "closed AI"? Maybe I'm worrying for no good reason.

[Adolf Allerbush]

All this open AI shit got me thinking about the closed AI that's going to come up and kick us in the ass from the blindside. Is there such a thing as "closed AI"? Maybe I'm worrying for no good reason.

Agreed. Google, Microsoft, Apple, USA federal government, China, Russia, India, Japan, Germany, etc. must all have "closed AI" projects in some form I would think

[LegoSkier]

My non-technical understanding is that, relative to the existing LLM models, the ability of Q* to do math is more generative. It’s not search/retrieve/present, it’s “apply an understanding of a topic to form an answer” - so it’s closer to actually starting to “think”. And once you start down that road you’re much closer to AGI than previously thought. So it’s not immediate danger, but it’s potentially a massive breakthrough that accelerates towards danger.

That’s my take as well. Current models basically turn all word combinations into vectors in multi dimensional space. Then when you ask it to generate new text it creates new word vectors a minimal distance from the vectors in your prompt. That’s why GPUs are useful in running models. The math is the same as calculating all the graphics polygon vectors in a video game.
This is saying they figured out a model less technique where it is reasoning a response not based on other preexisting vectors. If true that’s a pretty big deal.

[frorider]

Interesting article— https://www.theatlantic.com/ideas/ar...iction/677659/

[Core Shot]

Interesting article— https://www.theatlantic.com/ideas/ar...iction/677659/

The rest of the article was paywalled.
Interesting
I keep bumping into chat bots and emails
Fuck this new world order.
Yeah. First scrape by bots. But let me punch through to a biological being.

Then again, your post said
We can handle GPT-4 beating 90 percent of us on the SAT,

So I’m safe. For now.