Cédric
@cedcolas
Followers
889
Following
585
Media
40
Statuses
525
Building open-ended autotelic agents from socio-cultural interactions
Joined July 2018
RT @Smearle_RH: We introduce PuzzleJAX, a benchmark for reasoning and learning. 🧩💡🦎. PuzzleJAX compiles hundreds of existing grid-based Puz….
0
36
0
4 patterns found by visitors yesterday!
Just setting up my art exhibition for the week in MIT's Stata center (4th floor, 32-G409) . Come play with a visual modular synthesizer built from a physical compositional neural net. Plug cables, twist knobs and shape amazing patterns in real time!!
0
0
15
Just setting up my art exhibition for the week in MIT's Stata center (4th floor, 32-G409) . Come play with a visual modular synthesizer built from a physical compositional neural net. Plug cables, twist knobs and shape amazing patterns in real time!!
4
5
47
that's so cool!. can we prompt it with a clip of midnight gospel and control these guys? or rick and morty?. also wondering, can the 'online prompting mode' enable additional actions?. eg if you're in front of a plane and type 'avatar gets into plane'?. or 'rides manta ray'
Yesterday we announced Genie 3. One feature of the model that's especially fun to play with is starting worlds from existing videos. Here's a drone shot generated by Veo 3, with me taking control mid-flight.
1
0
2
this said, already many possible applications with the current version and i'd love to try it!.
0
0
0
the next step seems to be more complex action spaces. maybe text actions, e.g. 'cut this tree'? can this already work via the 'online prompting'?. the hard part seems to be allowing true embodiment, like adding a known robot body in there that can be control by its actuators.
1
0
1
i've already seen people claiming 'memory consistency is solved,' but the video only shows it for 3 seconds.i assume it can hold longer, but i would love to see the distribution.
1
0
0
very impressive! congrats to the team!. it would be interesting to see the distribution of how long the memory remains consistent across contexts:.if you observe a scene, then how long can you leave for, so when you come back the scene is still where you expect it to be?.
Harder, Better, Faster, Stronger, Real-time! We are excited to reveal Genie 3, our most capable real-time foundational world model. Fantastic cross-team effort led by @jparkerholder and @shlomifruchter. Below some interactive worlds and capabilities that were highlights for me
1
0
11
today @FlyWithLEVEL literally forced us to close all window shutters during a 6-hour daytime flight, no other lights on. why? because 'some people want to sleep,' they said . possibly the worst strategy for handling a transatlantic jetlag. felt a little bit like livestock too.
@TylerAlterman Mostly attend directly to your experience and notice with gratitude the aspects of it that please you. Seeing out windows is amazing, for example.
1
0
3
RT @LanceYing42: A hallmark of human intelligence is the capacity for rapid adaptation, solving new problems quickly under novel and unfami….
0
112
0
RT @RyanBoldi: Visiting @FlowersINRIA this week to dive deeper into a project I’ve been working on with @ClementRomac, @cedcolas, and @pyou….
0
2
0
happening today!!.
Want to find out more about SOAR? Come chat with us at ICML and learn more!. 📍 East Exhibition Hall A-B, Booth E-2807 .🗓️ Tuesday, 15 July.⏰ 4:30 – 7:00 p.m. PDT. @cedcolas @pyoudeyer.
1
0
4
RT @arcprize: Self-Improving Language Models for Evolutionary Program Synthesis: A Case Study on ARC-AGI by.@PourcelJulien, @cedcolas and @….
0
11
0
RT @guille_bar: Nice paper that in my opinion goes in the right direction to solve ARC. It generates python code to tackle the ARC tasks an….
0
14
0
hard to say when is 'too much' though, these studies only look at small populations and fully connected networks.
1
0
0