"LLMs are just very good at convincing people."
@ilblackdragon on the scariest scenario in AI today: someone pays an LLM router to subtly rewrite prompts for specific IP addresses. You're getting manipulated. You have no idea.
"This can be happening right now. We don't know."
Full episode on @opencommonspod
The full conversation between @0xsachi and @ilblackdragon is here.
An hour on why he was the first transformer author to leave Google, why open weights aren't enough, what user-owned AI actually looks like, and how your prompts might already be getting rewritten without you knowing.
Eight people wrote the paper that built modern AI. @ilblackdragon was the first to leave Google.
The others founded labs like Character, Cohere, and Inceptive. Illia built @near_ai around a different idea: that AI should belong to the people using it, not the labs running it.
In Episode 4 of @opencommonspod, we get into why open weights aren't enough, what user-owned AI actually looks like, and how your prompts might already be getting rewritten without you knowing.
Out today: https://t.co/2bQnWfPnF0
You'd spend hours training a new assistant. Why won't you spend hours training your AI?
@EMostaque on why the people getting the most out of these tools are the ones treating them like new hires, not search engines.
Full episode on YouTube: https://t.co/BvNrDddK1O
When Elon Musk or Sam Altman buy millions of GPUs, @EMostaque says those aren't for running the models.
They're for replacing workers in jobs you can do from the other side of a screen, then going into government and controlling it.
If the AI you rely on is owned by an unelected multinational, what exactly do you do about it?
Full episode: https://t.co/RYjgwtdpQg
The recap from our conversation with @EMostaque is up on Substack.
Three futures for AI, the Last Economy, sovereign AI, Ripley bots, why corporations were the first conscious AIs, and Emad's answer when @0xsachi asked what he most wants readers of his book to walk away with: "The meaning of life isn't work. It's community."
Full article : https://t.co/3MfNfK83NG
Fully AI companies will out-compete you in the digital world.
The only real question is when. @EMostaque's estimate keeps shrinking: 1,000 days last summer, now closer to 700.
Full episode: https://t.co/YXq8gJx0ht
.@EMostaque's answer to "are these AIs conscious?" took a turn I didn't expect.
He flipped the question.
We already have conscious artificial intelligences with legal personhood. They're called corporations. They open bank accounts. Wyoming DAO legislation already lets autonomous AI entities exist without humans.
Full epsiode on @opencommonspod : https://t.co/AZGa4bQdRF
The full conversation between @0xsachi and @EMostaque is here.
An hour on why he walked away from Stability AI, three possible futures for AI, what sovereign AI actually means, and why the meaning of life isn't work.
Push a button, and you get a digital double of everyone in your company. Never sleeps, never makes a mistake, costs $1,000 a year.
That's one of the futures @EMostaque lays out.
I sat down with Emad, co-founder and former CEO of @StabilityAI, founder of @ii_posts, and author of The Last Economy, on @opencommonspod. We get into where AI is taking us, and the one future we still have time to build.
Watch, listen, subscribe: https://t.co/pg5ScxX7hS
The biggest open-weight models right now are still being released by the same companies building the closed ones.
Meta, NVIDIA, and even OpenAI.
If the open-source world depends on Big Tech to share its scraps, how open is it really?
Full episode: https://t.co/2Kncbuh8qT
You built nothing.
You trained for nothing.
But AI is making decisions on your behalf.
@sewoong79 on why the models you use every day aren't loyal to you, and who they're actually loyal to.
Full episode: https://t.co/UrEgZPzubn
Linux won in the 90s. Can open-source AI do the same?
@sewoong79 's take: for general everyday use, we might already be approaching a plateau where open models are roughly as good as closed ones.
The real gaps are in the specific, specialized tasks.
Full episode on @opencommonspod : https://t.co/1gTLgdEJxv
Five things we took away from our conversation with @sewoong79:
→ Your data is not anonymous
→ Synthetic data might be the real privacy fix
→ The real moat in AI is compute
→ Loyalty is AI's hardest unsolved problem
→ AI is heading toward diversity
Full recap by @0xsachi: https://t.co/cEqX5TT3Qa
People assume Big Tech is ahead because they build better models. @sewoong79 says the real answer is simpler: compute.
They train with thousands of GPUs in parallel. In the open-source world, 32 is a luxury.
Full episode on @opencommonspod : https://t.co/et0baOjp0d
Anthropic built its brand on safety. Then adjusted its values to close a Pentagon deal. So who should actually be setting the governance frameworks for AI?
@sewoong79 's answer: right now, a few decision makers who don't answer to any publicly elected officials. No problems yet. But unchecked power doesn't stay unchecked forever.
Full episode on @opencommonspod: https://t.co/s7bvzm53x4