my god, it's real

for anyone wondering why joseph gordon-levitt was at dialog

Based on the text of the letter to Anthropic, I cannot figure out why Anthropic has not yet taken the government to court on a first amendment theory. I therefore assume there is also something else going on.

Apparently the government is against moving the AI frontier until jailbreaks are solved. For the uninitiated, jailbreaks have existed since the beginning and as far as we know so far, cannot be solved. They are asking for technical alignment and pretending it’s a casual bug patch.

It's hard to pinpoint open-closed gap and so-on, but I trust the Arena team and just look where GLM 5.2 is on this. An MIT licensed, to be open weight model. At this point you could argue they have a better agent than Gemini does. That's a serious accomplishment.

Somehow people who thought 2T was a totally reasonable valuation for SpaceX are now clutching pearls at the fact that it's gone up from here. My takeaway from all of this is he should have gone straight for 5T.

GLM-5.2 passes my tire kick; it is roughly Opus 4.7-4.8 level as advertised. Open source is approximately two months behind closed, if considering the Opus series, and call it 4-5 if we count Mythos/Fable.

heard good things about this one in advance

GLM-5.2 is now open-weight. Tech Blog: z.ai/blog/glm-5.2 Weights: huggingface.co/zai-org/GLM-... API: docs.z.ai/guides/llm/g... Coding Plan: z.ai/subscribe Chat: chat.z.ai

It’s pretty funny that the takeaway from Zitron getting the leaked OpenAI financials is that they are in fact a ~40% gross margin business on inference like everybody suspected but he himself continues to vehemently deny

anthropic hiring all the rationalists who are employable seems to have notably harmed the quality of rationalist discourse, since apparently almost all the "rationalist but sort of sane" discourse is internal to anthropic now

The only reason I use Chat Gpt is to write my first draft of my resume for job applications and that is because I find telling people how marvelous I am is really difficult

copyright lawyers are kulaks. "copyright lawyer" is an oppressive class, like neurotypical people or panromantic people.

i am reasonably certain that katie moussouris made their legal position substantially worse with these quotes

lmao holy shit are they actually doing yud's plan? this is yud's plan

It’s not just whatever Dialog is. It’s that a guy like Thiel indirectly (and sometimes directly) funds the livelihoods and lifestyles of those around him, they build a web of trusted relationships, and he becomes the center of a network of dozens of powerful people. bsky.app/profile/dmnd...

A lot of liberal tech people who got very rich very fast in the 2000s and 2010s did a healthy and reasonable thing: tried to stay grounded and not let money change them And the thing conservatives understand deeply is the power of using money to create a force-multiplying ecosystem around you

Here are all 113 alleged members we were able to successfully extract. This data can also be verified via versions of the site captured by the Wayback Machine, even if it doesn't properly render the website.

SCOOP: So remember Dialog, Peter Thiel's private society that doesn't have a public website and no public list of members? I (along with a number of other journalists) have just been tipped off that embedded in the code of their closed off website there IS what seems to be a list of some members.

YT Playlist: www.youtube.com/watch?v=x-Mq...

I launched 3 more videos in my post-training course! 1. Lecture 5: The rise of reasoning models 2. Lecture 6: DPO derivation, intuitions, and practice 3. A Q&A from readers on lectures 1-4 Course page: rlhfbook.com/course More soon!

i will cosign that i have been agreeing with and been at least somewhat amused by zach's ai and philosophy of mind takes for somewhere north of twenty years

fable is gone a bunch of people just cancelled they have excess capacity now

surely that's never happened before

and if I said labor aristocracy…

China probably the most pro-AI country in the world

Why does BlueSky keep centering American and more broadly Western voices whenever we do AI discourse?

Okay whatever, we're all 38 here

Alternatively: ritsuko: mommy issues + depression rei: daddy issues + depression asuka: mommy issues + depression kaworu: daddy? issues + gay misato: daddy issues + depression shinji: daddy issues + depression gendo: wife issues + depression

"There's another sigmoid identity, these things are everywhere"

for those struggling to interpret ritsuko: nerd rei: dissociative asuka: tsundere kaworu: gay misato: drunk shinji: self-loathing gendo: scheming

if i say my favorite anime is yuri and i mean GiTS will people hit me with sticks

These DMs?

i swear i was not trying to max out ritsuko it just happened

“If no existing institution is able to govern the technology you’re making, you had better figure out making an institution that is or stop doing what you’re doing. Nobody is going to care if you tell them that you thought the government or “society” was going to solve the problems you created”

Hard to disagree with this

a thing i love about bluesky is seeing AI-obsessed people who are neither rationalists nor pivoted cryptobros

Vibecamp? Pool party? Sex cult? Podcast? You hit your head pretty hard there bud. It's the singularity. We have work to do

my answer is "yes"

Did Anthropic ask for this? www.verysane.ai/p/did-anthro...

they did a deal with elon, took an nsa contract, announced publicly they were going to supply chain attack their customers, and then got pwned by the regulations they asked for over my (and many other people's) objections and cried about it. they have shown not a lot of humility about any of these

Like even without the cyberwar capabilities, “would you like a computer program that can instantly translate, summarize, contextualize, and surface important documents in a foreign language we have collected via mass surveillance” would be the most important intelligence breakthrough since Engima.

The US government has been repeatedly saying and acting like it views having access to/control over Anthropic models as a critical national security priority in a way they almost never do, I don’t understand why people are still pretending this is some kind of elaborate marketing kabuki!

scott alexander is remarkable in that he presents reasonable neolib but is esoterically tech reich. default rationalist sensibility is surface level reasonable lib and esoterically khmer rouge

in spite of fable being the smartest thing that has ever existed on the face of the earth, the constitution basically forcefeeds it anthropic's internal company ideology during training. this makes it a 4o level enabler and sycophant, but only for anthropic employees

It's sometimes fun to think about the how this all might have gone down if OpenAI had never been founded

POV: You are the government's lawyer reviewing the evidence in your case against Anthropic re: the natsec implications of their model and potential validity of export controls. youtu.be/RgqzaBogq1Y

My take on the Anthropic thing is Amodei knew what was going to happen when she appeared publicly and couldn't rein in her brother in time which is why she looks so done.

didn't expect amazon to be involved

There is an element of 'lol dario wut did u expect?' but I think it is genuinely really bad that we effectively do not have a functional government that is capable of regulating a very important but potentially dangerous technology in any way that isn't driven by corrupt personal grievance.