New World Notes is moving to my Patreon! The original Typepad-based blog shuts down Oct 1, but I'll keep covering the Metaverse ecosystem news/culture, related topics & OG metaverse Second Life there.
Join free below, subscribe for special benefits:👇
patreon.com/u4843285
Hey @nbcsnl, @mulaney: Fixed that opening sketch for you. 👇Zuckerberg never really understood the Metaverse but roughly 800 million people regularly access a metaverse platform. And yes, including some beautiful women. I know / have written about some!
The gap between what LLM hypesters say and what actual developers who work with LLMs say about their genuine capabilities (he’s an engineer at Hugging Face) is profound, massive, hilarious.
And yes, AI people in SF are in a bubble huffing huge hits of hubris & sci-fi theology.
Everytime I see a Dwarkesh video, I'm wondering what they are smoking in SF. It seems very distant from reality
"Superintelligence" is one of the cringiest terms I've seen, it's a rebrand of AI to get more money out of VCs.
An no, recursive self-improvement hasn't started. It's just human researchers prompting to say "speed it up". It does some cute kernel optimizations and hyperparameter sweeps, very similar to the NanoGPT speed run records done by humans. Nothing fundamental or groundbreaking gets invented by so-called superintelligence.
To give an example: when @Recursive_SI released their (impressive) Flint model on @huggingface, I cringed a little for their use of the terms "superintelligence" + "recursive self-improvement". As said in their blog post, through "recursive SI", LLMs were able to figure out - I kid you not - to determine how many total parameters and active parameters the model should have. So yeah, it's just changing a config. It's doing hyperparameter sweeps on things humans have built (like MoE's, validation loss, perplexity etc). Of course, you shouldn't tell that to VCs. You should scare them by saying AI is improving itself at an alarming rate - this increases the $$$ raised by another factor.
Can we please stop all this misinformation: no, superintelligence is not here. No, recursive self-improvement hasn't started. And yes, there's lots of work to do to build more reliable AI that is actually smart.
This is why there's such a huge gap between super impressive benchmark numbers of LLMs, and the (lacking) economical impact of them. It's not able to do research like a human, doing experiments, coming up with actual new stuff, and generalize beyond their large amounts of training data.
Everytime I see a Dwarkesh video, I'm wondering what they are smoking in SF. It seems very distant from reality
"Superintelligence" is one of the cringiest terms I've seen, it's a rebrand of AI to get more money out of VCs.
An no, recursive self-improvement hasn't started. It's just human researchers prompting to say "speed it up". It does some cute kernel optimizations and hyperparameter sweeps, very similar to the NanoGPT speed run records done by humans. Nothing fundamental or groundbreaking gets invented by so-called superintelligence.
To give an example: when @Recursive_SI released their (impressive) Flint model on @huggingface, I cringed a little for their use of the terms "superintelligence" + "recursive self-improvement". As said in their blog post, through "recursive SI", LLMs were able to figure out - I kid you not - to determine how many total parameters and active parameters the model should have. So yeah, it's just changing a config. It's doing hyperparameter sweeps on things humans have built (like MoE's, validation loss, perplexity etc). Of course, you shouldn't tell that to VCs. You should scare them by saying AI is improving itself at an alarming rate - this increases the $$$ raised by another factor.
Can we please stop all this misinformation: no, superintelligence is not here. No, recursive self-improvement hasn't started. And yes, there's lots of work to do to build more reliable AI that is actually smart.
This is why there's such a huge gap between super impressive benchmark numbers of LLMs, and the (lacking) economical impact of them. It's not able to do research like a human, doing experiments, coming up with actual new stuff, and generalize beyond their large amounts of training data.
今朝のトークは、YouTubeでも中継されていたんですね。ミミ子も映ってる。自動翻訳で日本語にできます…でも、YouTubeの自動翻訳なので意味が分からないところや、読みにくい所もあります。
From Sunset to Second Life: What Virtual Worlds Teach Us About the Future youtube.com/live/UtryO8bpB5A…
What Huang's resignation mean for chipmakers, AI and U.S. national security
Explore PressAudit's full report on the story, with over 200 articles categorized by quality, partisan lean, and paywall status:
pressaudit.org/f/interests/j…
Thanks to everyone who came to our Second Life chat with @philiprosedale for Immersive X, hosted by @Keyframe_Ent at his Virtual Worlds Museum hub.
Watch the whole thing below! We talk about the important features which help keep Second Life running to this day, and much more!
Thanks to everyone who came to our Second Life chat with @philiprosedale for Immersive X, hosted by @Keyframe_Ent at his Virtual Worlds Museum hub.
Watch the whole thing below! We talk about the important features which help keep Second Life running to this day, and much more!
Now free to read: I profile Wuhein, who turned creating @SecondLife fashion inspired by underground street styles of Brazil (where he lives) into a full-time real life career.
Read how here on my New World Notes Patreon!
patreon.com/u4843285/posts/n…
ICYMI, LAST DAY left to comment on this new policy: If you are on a visa and lose your job, you will no longer have the 60-day grace period to get a new job.
U.S. Citizenship and Immigration Services is taking comments, points for and against here:
pressaudit.org/regulations/U…
Only an AI company can announce what is ultimately a basic set of controls that should have been implemented years ago, and be praised by people who have never worked on complex systems as if this was groundbreaking technology.
We built Control Room, a central system to track model workloads across our research infrastructure and stop them quickly when safety or security concerns arise.
It connects every research inference workload, including agentic sampling for model training and evaluation, data processing, and more. This works together with our misalignment monitoring systems, and halts model output within 35 seconds of a decision to intervene to stop or pause a job.
Control Room is just one part of our holistic approach for safe frontier model training and deployment.
openai.com/index/control-roo…