Folks in privacy engineering and related fields, it's PEPR time again -- submit talk proposals about topics related to designing, building, and understanding products and systems which foster privacy and respect. I'm looking forward to seeing your talks!
www.usenix.org/conference/p...
It was incredibly sweet to see Hinton teach @weeklyshowpodcast.bsky.social some deep learning!
youtu.be/jrK3PsD3APk?...
Yes, and also, this isn't real "news" :-)
It is good content though.
My latest @theconversation.com article on Google's new "AI mode" and the issues with vertical integration 👇🏼
theconversation.com/google-is-go...
Some great reporting work here!
Very cool. Looking forward to your demo at Tinkerers! :-)
"But, so many ways to exit... Gee, I wonder if I am doing the right thing here as a beginner..."
Don't get me wrong -- absolutely ❤️ vim.
You guys, look, we are headed towards the trough of disillusionment [1]
[1] en.wikipedia.org/wiki/Gartner...
I'll hazard a summary: we seem to be on some kind of a downward spiral of bad leadership -> bad outcomes for everyone (especially working class) -> bad leadership.
Historically, democracies tend to self-correct. So, if you have a hopeful outlook (like mine), you say: eh, this too shall pass.
Gary Marcus' analysis is the opposite of level-headed. It is self-indulgent and biased in its own way.
o3 is not AGI.
But, it does represents a substantial improvement on one important benchmark and open labs will replicate it and we will learn a lot more soon.
I get that you want to fight the hype (I do too!)
But, is getting into a rapid-response mode really the right approach here?
Let more data come in. We will learn what is what in some time. Give open labs time to replicate findings.
Let the scientific inquiry take its course.
Peace!
This is a "war"... Really?
You made one off-hand comment from a journalist into a blog post...
And your post is all confused... Somehow, you are equating human intelligence to AGI, which... I don't even know where to begin with that.
And you have made it all about some "influencers" and yourself.
IMO, the best moment of Ilya Sutskever’s insightful talk at NeurIPS was his answer to this question towards the end:
youtu.be/1yvBqasHLZs?...
@reckless.bsky.social did a revealing interview of Mustafa Suleyman for Decoder:
1. His vagueness about the MS<–>OAI partnership hangs ominously, like a Chekhov’s gun.
2. His answer on how he makes decisions makes me think that he has never given the idea of "how to decide" any thought, ever.
Yeah. It is weird. The solution is open alternatives! so happy to see you hanging here! :)
Also, are you folks thinking about using the AT protocol for The Verge. If you do, I will subscribe!
Promise. :-)
I am not an expert, but:
chatgpt.com/share/674f22...
and
stackoverflow.com/questions/25...
Happy building! :)
Would love to learn more about your experience as you build it.
All the best!
Glad that we are talking about AI power concentration!
Agree but not it's not just rhetoric from well-funded companies training models...
Solutions for less concentration might be both political (inequality?) and technical (R&D for efficient models) as well.
Right now. "they" own the rails.
Has anyone tried one of their "boxes"?
www.youtube.com/watch?v=_aqM...
I am thankful that you decided to execute towards your vision!
Happy Thanksgiving to you and the team at Bluesky!
Essay suggests LLM training is hitting a scaling wall —diminishing returns from bigger models or more data.
Remedy: train models to reason step-by-step, trading training compute for inference compute.
But the remedy might not be all that is promised:
aidanmclaughlin.notion.site/reasoners-pr...
Already reposted. :-)
Nice to meet you, Max! I am building techne.app
Created a Build in Public starter pack. 🎉
If you're an indie hacker building in public, feel free to join:
👉 Share your project below
🧡Like this post for the algo push
🔄 Repost this for visibility
go.bsky.app/PvitcMe
Thanks for the feedback!
Eventually, monthly subscriptions for extra features that are under construction.
For now, the idea is to put it out there so people can try it and give early feedback. :-)
Did you know there’s a site to search for starter packs of people to follow? Built by @mubashariqbal.com
blueskydirectory.com/starter-pack...
Yes. It's that extension that helps people make sense of Hacker News! (for now)
Thanks for adding me to the SP! Looking forward to sharing progress and learn what other folks in the community here are up to.
Chicagoland.
