@VilleKuosmanen

gentleman scientist 🤖 @armnet_dev building the world's largest open arm farm 🇫🇮🇬🇧

 London, UK
villekuosmanen.com/
Joined May 2020
  • Tweets2.6K
  • Following1.1K
  • Followers3.6K
483Media
A month ago @pravsels and I set out to reproduce @physical_int’s RL Token paper. Today, I am sharing our research notes from the journey we’ve been on. xitter.cf/VilleKuosmanen/status/…
Ville🤖
@VilleKuosmanenMar 24
"RL Token" looks like a great and surprisingly simple post-training methodology for optimising robot models for dexterous tasks in the real world! Over the next few weeks, me and @pravsels will be attempting to reproduce the results (& open source the code) Stay tuned 👀
1
12
2
129
22,765
personal and company AI agent startups - ship your agent with its own personal phone number and ability to make and receive phone calls and I will pay you most annoying thing in any business is having to call every vendor to enquire for details / price
3
1
248
I don't want the agent to answer my phone, I want it to have its own phone number I can fill in the enquiry form so the vendor calls the agent instead of me, then the agent having context to understand what it should ask
46
Another world-class robotics lab coming out of stealth, exciting results in in-context learning! Ed is one of the leading UK robotics researchers, I remember seeing him demo Instant Policy at a tiny room in his Imperial lab early last year, amazing to see the progress since!
Edward Johns
@Ed__Johns9h
Today we’re out of stealth, and we’ve raised a $40m seed round co-led by @creandum and @usv, to bring Instant Robot Learning to life. For the past few months, we‘ve been busy building a new company. Machenta. This follows results from a new kind of robot foundation model we’ve built, which I’m so excited to share with you today! Usually, it takes hours/days of data collection and training to teach a robot. We’ve built a model that learns tasks instantly: one demonstration, no further training. A model trained entirely in simulation, without a single pixel of real-world data. A model where teaching robots is so quick and easy, it feels like physical vibe coding. A model for robots that learn like humans do, from someone like you. Welcome to the era of Instant Robot Learning. Thank you @Kathi_Wilhelm, @rebeccakaden, @hkanji, @jonno_evans and all of our angels for your support! @Machenta_ Learn more: machenta.com/blog/instant-ro… See below for what you can do with Instant Robot Learning!
1
3
68
3,378
🤯🤯
Andrew Seddon
@seddonandrewOct 26
PCBs made in America. Twice as fast as Shenzhen. Today @CircuitHub is unveiling The Grid, our factory-scale robot in Massachusetts that builds circuit boards on demand. No changeovers, no per-design setup. Prototypes and production batches, fully automated.
1
29
3,988
why does every pre-seed robotics startup's website read "the X layer for physical AI" if we added them all together we'd have like 20 separate layers all trying to sell to each other instead of end customers 🫠
11
4
3
85
5,806
US robotaxi deployments by state and provider the deployment of AVs has barely started, and it's interesting how each of the big 3 lead at one or more states (Tesla in TX, Zoox in NV) Physical AI will be far more geographically fragmented than digital AI or SaaS ever was
1
1
5
464
data from robotaxitracker.com/

Robotaxi Tracker — Fleet Data & Service Maps

Community tracking for Tesla, Waymo, Zoox, and Avride: vehicle directories, sourced service maps, booking information, ride statistics, and available government data.

robotaxitracker.com
74
a single popular robot like the Microduck will generate demand for ~1m actuators dear investors, why aren't you funding more startups in the robotics hardware layer?
𝐀𝐆
@AGkorthosOct 26
Pollen Robotics reportedly received 15,000 additional Microduck orders following GTC Berlin. According to Korean reporting, it also signed an agreement with ROBOTIS covering demand for 750,000 actuators across 50,000 planned robots.
6
3
26
1,559
started seeing YouTube ads for torque sensors 😅
2
5
508
Wonderful to see the data from ArmnetBench v0.1 being used for interesting applications like this! Great work @josefchen !!
Josef Chen
@josefchenOct 23
RoboJev: people are watching thousands of videos to label task success. Now, you can grade robot data faster than you make it. 48x cheaper and 28x faster than 6 Astra. 27B parameters, open weights.
1
1
14
839
Ville🤖 retweeted
Humanoid
@TheHumanoidAIOct 23
Reinforcement learning doesn't work ... until it does!
3
11
1
117
11,345
If you like robots, you should listen in on @zaringleb 's Robot Learning Collective paper club, or come in person in London🇬🇧 if you're around! I will be presenting the FACTR and FACTR 2 papers along with a live demo 🤖 discord.gg/DKqU4ynJe

Join the Robot Learning Collective Discord Server!

Check out the Robot Learning Collective community on Discord – hang out with 622 other members and enjoy free voice and text chat.

discord.com
Ville🤖
@VilleKuosmanenOct 23
Replying to @Caro_Nahkriin
At least I managed to sneak some time to build a cool robot demo for a Paper Club run by the Robot Learning Collective this evening 👀 after making the slides of course! discord.gg/DKqU4ynJe

Join the Robot Learning Collective Discord Server!

Check out the Robot Learning Collective community on Discord – hang out with 622 other members and enjoy free voice and text chat.

discord.com
1
1
7
663
not entirely sure when this happened but somehow my job has become making slide decks and proposals i think I worked on a different slide deck or proposal every day this week 🫠
3
23
1,115
"just make Claude do it" in order to avoid the decks looking like, and the proposals reading like slop, I have to tightly script them to the point where the AI saves maybe 50% of the time. the complex part is figuring out how to present an idea, not doing the visuals
1
1
154
Exciting new training recipe, if you train robots Paul is someone whose work you should follow :)
Paul Zhou@zhiyuan_zhou_Oct 20
Last year I trained the advantage-conditioned VLA for π*0.6 during my internship @physical_int. Now, I have a much improved recipe. Introducing ✨Advantage Guidance✨, a performant RL method with stable advantage-conditioned supervised training and contrastive guidance sampling
2
1
33
3,018
Ville🤖 retweeted
will o’brien
@WillobOct 20
Another takeaway after three years of building a company is that the returns to just "being a good person", and not acting transactionally, are higher than I expected. There are so many transactional, boring, and at times immoral people in the tech industry that if you act with sustained warmth, authenticity, humor, and integrity, great gifts will be laid before you. White pill.
79
95
36
1,612
100,732
awesome home robot from a world-class team!
Ros Shinkle
@make_robot_goOct 20
Introducing Voxel, a droid butler made by Tuesday Labs.
17
1,629
Always love being welcomed home by the beautiful English weather 🏴󠁧󠁢󠁥󠁮󠁧󠁿😍 California was too hot anyway
12
575
SF is awesome but I will make a hot take claim I think we might have passed peak SF? Last summer it felt like the only place you could build a successful AI company was SF. Accelerators rushed to relocate their portcos here But is SF the right place for the age of deployment?
6
26
2,320
Is this how @sundayrobotics transport the robots? Or is Memo driving the van? @chichengcc can you confirm 😄
3
6
127
9,237
Load more