Real-world data x embodied AI. ๐ค๐๐ค
- Tweets 23
- Following 3
- Followers 1,154
- Likes 14
๐ค: c
๐: o
๐ค: n
๐: g
๐ค: r
๐: a
๐ค: t
๐: s
๐ค: ๐พ
weโre proud to announce the launch of @daofive fund 2
incredibly grateful that we get to continue the journey we set off on three years ago, and thankful for everyone who has supported us along the way
fortune.com/crypto/2025/05/0โฆ
5
667
๐ค๐๐ค๐๐ค๐๐ค๐๐ค๐๐ค๐๐ค๐๐ค๐๐ค๐๐ค๐๐ค๐๐ค๐๐ค๐๐ค๐๐ค๐๐ค๐๐ค๐๐ค๐๐ค๐๐ค๐๐ค๐๐ค๐๐ค๐๐ค๐๐ค๐๐ค๐๐ค๐๐ค๐๐ค๐๐ค๐๐ค๐๐ค๐๐ค๐๐ค๐๐ค๐๐ค๐๐ค๐๐ค๐๐ค๐๐ค๐๐ค๐๐ค๐๐ค๐๐ค๐๐ค๐๐ค๐๐ค๐๐ค๐๐ค๐๐ค๐๐ค๐๐ค๐๐ค๐๐ค๐๐ค๐๐ค๐๐ค๐๐ค๐๐ค๐๐ค๐๐ค๐๐ค๐๐ค๐๐ค๐๐ค๐๐ค๐๐ค are Coming!
1
6
3,548
Models need to see, hear, and understand the world to reach the next level.
Whoever owns the data pipeline wins.
Creating a foundation model for robots requires roughly 2x the complexity of a LLM if you compare both to a human brain
A LLM is capable of language processing, memory, reasoning, pattern matching, and creativity and could represent 20-25% of the human brain
A humanoid foundation model expands these cognitive functions to "embodiment" and requires all of the above plus spatial awareness, proprioception, sensory processing, motor functions, and real-time adaptability. A humanoid foundation model capable of autonomous work could approximate 50-60% of the human brain
One of the most challenging and fascinating areas of research out there
nitter.net/i/grok/share/NZF6Ibr5pโฆ
2
658
OpenVision retweeted
Lots of room to innovate at the data collection layer for crypto x ai x robotics (which I guess is called DePAI now...)
Unlike text-based data for LLMs, the physical world is diverse and complex. We can't just scrape the web for egocentric videos, mocap kinemetics, spatial data etc.
Token incentives could accelerate data collection at scale classic DePIN style. But instead of passively installing sensors, here users have to actively perform tasks.
This might look less like a marketplace for tasks ("Film yourself folding laundry for $5") and more like games (@frodobots scavenger hunts or @reborn_agi VR platform) and even wearables (@OpenVision_AI AR glasses).
4/
The DePAI infra stack is rapidly developing.
The most active layer is at this early stage is Data Collection which can provide real world data to train Physical AI agents deployed on robots as well as stream real-time data to navigate environments and complete tasks.
1
4
12
2,248
OpenVision retweeted
an ode to the history of technology, and human potential:
701
666
10,236
2,654,181
Truly a historic moment in #AI #Robotics.
@OpenVision_AI is empowering the new era of AI-powered robots. A decentralized, reliable and verifiable data foundation to truly make history.
WE, CREATORS.
1
5
1,650
A fun handshake and selfie moment with @VitalikButerin and @EMostaque alongside our robot dog ๐ค๐โ๐ฆบ! ๐ฅ @OpenVision_AI is decentralizing human data generation. Genuine, on-demand, verifiable.
Amazing event with @OraProtocol @ethereum_sg, special friends @berachain and @Lab7007! ๐
2
2
15
18,756
2/3 ๐ Vitalik, Emad and @0xKartin highlighted how blockchain ensures identity, authenticity, and privacy in AI systems. This not only combats fraud but builds trust, especially in critical sectors like healthcare and education. #AITrust
1
942
3/3 ๐ @OpenVision_AI stands at the intersection of AI and blockchain. With on-demand human data generation, it drives decentralization, privacy, and ensures humanity in AI robotics, paving the way to AGI for broader social good. #HumanData #Robotics
1
821
Make AI participatory for everyone. ๐ถ๏ธ๐ค
Join us today with @OraProtocol @ethereum_sg @Lab7007
YES! ๐๐๐ต๐ฒ๐ฟ๐ฒ๐๐บ ๐
๐๐ ๐๐ผ๐ป๐ณ๐น๐๐ฒ๐ป๐ฐ๐ฒ ๐๐ฎ๐ is happening tomorrow ๐ธ๐ฌ
#InTheEtherASM #TOKEN2049 ๐งต โคต๏ธ
1
3
1,490
Letโs march into the era of embodied AI. Collective AI x Robotics innovation x open economy.
Huge shoutout to @caseykcaruso for creating this incredible DeAI map.
Spot us here! ๐ โ ๐ค
topology.vc/deai-map
Open sourcing a community-led market map of Decentralized AI.
topology.vc/deai-map
Most market maps are static and biased. We wanted to make one that was interactive and crowdsourced.
Pls submit any projects that are missing.
We implemented a 3d viz using @d3js_org and a 2d table view with WebGL.
Most projects live between multiple categories. We did our best to simplify to one.
It's early days in DeAI but I'm tracking closely ๐
2
1,351
Spatial intelligence โ reason with reality, a more foundational aspect we individuals can participate
economictimes.indiatimes.comโฆ
5
1,144
Amazing work๐and imagine we could scale the benchmark dataset further...๐
Today weโre releasing OpenEQA โ the Open-Vocabulary Embodied Question Answering Benchmark. It measures an AI agentโs understanding of physical environments by probing it with open vocabulary questions like โWhere did I leave my badge?โ
More details โก๏ธ go.fb.me/7vq6hm
All of todayโs state-of-art vision+language models (VLMs) fall well short of human performance.
In fact, for questions that require spatial understanding, todayโs VLMs are nearly โblindโ โ access to visual content provides only minor improvements over language-only models.
We hope that OpenEQA motivates additional research into helping AI understand and communicate about the world it sees.
1
1,022
