zlacker

Ilya should just go to Anthropic AI at this point. They have better momentum at this point after all this, and share his ideals. But it would be funny because they broke off of OpenAI because of their Microsoft ventures already in 2019, haha. He'd be welcomed with a big "We told you so!"

>>jug+(OP)
That would actually be a good idea imo. Ilya should make OpenAI a PBC, then merge it with Anthropic and Amazon's compute power. Meanwhile Altman can become the next real life Nelson Bigetti or sth.

>>jug+(OP)
Considering Anthropic and them joining up, with Ilya and Dario that would be a technical powerhouse. As Amodei already showed that such a key person can ramp up quality real fast out of nothing. The two back together would be fantastic. Between Altman and Brockman there's nothing to write home about tech-wise.

>>jug+(OP)
I don't consider Anthropic's approach to safety fantastic. They train the model to lie, play cat and mouse with jailbreakers, run moderation on generations with delay etc. This makes the model appear safer, as it's harder to jailbreak, but this approach solves nothing fundamentally.

If Ilya is concerned about safety and alignment, he probably has a better chance to get there with OpenAI, now the he has more control over it.

>>Athari+w6
I haven't paid a lot of attention to Anthropic. Are you able to summarize, or link anything about, those events for those who missed it? Particularly the "training to lie" bit

>>Keyfra+r6
I've heard the opposite about Brockman. What makes you so confident about this tech abilities?

>>Athari+w6
Anthropic safety is overboard. I tried the classic question of "how many holes does a straw have?" And it refused to talk about the topic. I'm assuming because it thought holes was sexual.

>>dalore+do
Given what AIs "know" about humanity, I think it's safe to assume that they "think" every word is sexual. For example straw could be short for strawman, which is a man, which is sexual. Or it can be innuendo for... you know.

As for your actual question, it seems to me that a straw is topologically equivalent to a torus, so it has 1 hole, right?

>>MattHe+zj
There are interviews with all three at Dwarkesh Patel youtube channel. One is definitely not like the other two, but that might just be my impression based on those interviews. edit: Brockman might've been on Lex only.

>>dalore+do
When did you last try that? I checked right now and it says

> A straw has one hole that runs through its entire length.

>>didntc+0h
David Shapiro complained about Anthropic's approach to alignment. In his video https://www.youtube.com/watch?v=PgwpqjiKkoY he discusses ableism, moralism, lying.

As to cat-and-mouse with jailbreakers, I don't remember any thorough articles or videos. It's mostly based on discussions on LLM forums. Claude is widely regarded as one of the best models for NSFW roleplay, which completely invalidates Antropic's claims about safety and alignment being "solved."

>>JBiser+Hq
> it seems to me that a straw is topologically equivalent to a torus, so it has 1 hole, right?

For a mathematician, yes. For everyone else, it obviously has two, because when you plug one end, only then it has one.

>>dalore+do
that sentence makes no sense to me, what is a straw here?

>>visarg+kv
Now follow up with: how many holes do trousers have?