mastodon 4.7.2EN/FR
Sovereignty - Free Software - Data Privacy - C++ / AI / Python / Architect
Reminder of how dire our digital privacy situation has become...
TikTok can spy on you even if you don’t have an account.
Collecting personal data just to resell it should be outlawed in my opinion.
But combine this with data brokers, ChatControl’s recent vote, and the global push for age verification through 'trusted private third parties', all while data theft is at an all-time high, and you start wondering if we aren’t attending a clown show with very real consequences.
A very cool initiative about making an open vaccum robot collaboratively.
I recently stumbled upon some interesting research on thermodynamic computing. It's noise-powered, almost free, although it can be slow.
https://www.quantamagazine.org/thermodynamic-computers-go-with-the-energy-flow-20260715/
"A GIF image that prints itself"
"In Piet, a program is an image. The 252x189 GIF on the left is the program, and running it prints that very GIF, byte for byte."
https://www.youtube.com/watch?v=GwMtzhjCzyc
His blog post: https://dev.to/mame/a-quine-in-piet-and-a-plug-for-the-esolang-design-contest-339o
Always amazed by Yusuke Endoh's magnificent creations. I highly recommend his work. He is a multiple-time winner of the https://www.ioccc.org/ competition.
I recommend this video by Marcus Hutchins, which debunks the unfounded concerns about the threat posed by frontier AI models in the hands of attackers and highlights the real issues they present for defenders.
In summary: Attackers focus on managing ill-gotten gains, not AI. State actors can build their own models without commercial safeguards. Defenders, already under-resourced, would benefit from AI models if safeguards didn’t raise costs.
Codeberg is fighting back!
"Protecting our FLOSS commons from LLMs"
- "We are promising to not use any of your data to train LLM"
- "our terms of use [prohibit] 'vibe-coded projects'"
https://blog.codeberg.org/protecting-our-floss-commons-from-llms.html
I fully agree with their initiative to ban vibe-coded projects to prioritize resources for developers who actually understand what they are doing and care about long-term collaboratively maintained software more than the short term fame of an obscure piece of generated code.
And they respect our work and don't harvest it for LLM training. Although I wouldn't mind as long as I give explicit consent.
Recently, I’ve been wondering...
If Chinese AI models are nearly equivalent in quality to US models, can cost 5x less, and are often open-weight, how will OpenAI and Anthropic plan to pay back their massive debt?
Time to buckle up.
Warren Buffett about the current market:
"It’s tough to find values when everybody is preferring gambling"
“a church with a casino attached”
