Benjamin Ricaud (@GBR_Data) header
Benjamin Ricaud (@GBR_Data) profile picture

Data, science, research, graphs, artificial intelligence.

Tromsø, Norway
Joined March 2017
A new music genre appeared in September: slopcore. These are fully AI-generated music videos where a coding agent is the whole crew and one human only directs. I looked into how they are made. A thread: bricaud.github.io/blog/slopc… #slopcore
1
132
My conclusion: yes, it's a genre. Directing the AI is real creative work. The barrier moves from playing an instrument to directing: time, judgment and, for now, a budget. Expect lots of bad music, but good music too, from people with no knowledge of music theory.
1
17
These songs with video clips "directed" by a single human and produced by AI is so fascinating! There is a link to how it is made in this one: Claude, Suno and Midjourney and a lot of prompts!
Universal High Income The most likely outcome
62
Can AI agents make law? 🏛️🤖 New preprint with Artem Galushko (UiT): we set up "constitutional assemblies" where AI agents debate and draft a constitution for an AI community. 6 model families, 40 sessions, 627 clauses. arxiv.org/abs/2610.24661
1
30
But there's a ceiling. Agents grant themselves the right to rewrite their own code, even their values, yet treat it like learning: no threshold, no new safeguards. They extend familiar legal ideas to new situations without working through what those situations change.
1
3
J'adore ❤️
I came to realize people liked the french #slopcore without having the lyrics 😅 sorry! So here is a version with english subtitles!
13
Benjamin Ricaud retweeted
Probably the best video ever created.
377
5,769
571
25,018
1,415,815
Benjamin Ricaud retweeted
Today @GoogleResearch is launching Deep Review — an agentic system designed to stress-test manuscripts, design docs, and complex proposals before submission or publication. After pilot partnerships analyzing over 15,000 research papers, we’re making Deep Review globally accessible to help researchers, engineers, and professionals bring structured critique to their daily workflows.
34
585
90
4,378
370,873
Reviewers at big machine learning conferences, hidden behind anonymity, can be so unprofessional!
> point out to reviewer that the 10 things they asked for are already in the paper, they missed the central quantity in eq 1 and fig 1, and the requested reference is irrelevant > reviewer silently reduces score to 2 because of “hostile response”
1
57
There are a lot of things wrong with this world… but too much intelligence is not one of them.
53
453
35
2,901
341,321
This is so true!
one novel human idea can be worth more than enormous amounts of LLM inference-time scaling. this is where human cognition retains a real economic comparative advantage.
1
322