Seonglae ChoVoxstral
- realtime interaction social
- OCR + Agent tool
- robotics
- competition arena leaderboard
- investment
- on device onnx
- cool diffusion gradient design
- gamification
Battle arena game
- give agent tools to move kick punch etc
- 3d humanoid model
- gta like realtime openworld
- humanoid vla arena with hp attack game pvp
idea
Car accident prediction. leaderboard targetting
stack: convex, next, vercel etc
Leveraging human's prediction capability to solve early detection of the car accdient
- reference ../conference-arena
- but style should be differnt. for this product, noise based diffusion gradient style like fully loaded https://originator.inc/ or old version of https://www.inceptionlabs.ai/.
- battle is important
- enter the battle (enter your name modal)
- watiitng for opponents max 5 secs
- if noone which also finding a match, compete against mistral ai agent
- which oppoent, three people complete at the same time
- generate a tile and all car trajecotries for 10 seconds
- all 3d simulated are rendered by react three fiber
- show user and ai agent a 3rd person view of the created map with a lot or roads and car trucks etc
- define several 3d block types vehicle types and let ai to generate them
- basically ai agents have an access to image, and datas. grep current state tool etc. curernt position of specific vehicle etc several tool.
- the button should be point a specific pint
- Click the point whwat you expected to be an accident point.
- How close the corrdinate and early means you are correct. combine two metric carefully to make user to avoid random predictoin
- after that, update rating for temporary named user by identifying unique browser.
copy claude code from there. commit frequently etc.
Slides
my mo
인간한테 어려운데 ai 한테 쉬운 문제들은 상대적으로 많이 풀렸다. 다만 for ai is hard but for human is really easy 는 사회에 적용 안된게 많음. 이런 경우는 데이터 형태에 원인을 두거나 사람의 large overview insight 가 필요한경우다.
next token prediction, technically world moㄸrdel is next state prediction.
문화가 이기는 이유
prediction market 하고 simile case 소개
prediction is important and fundamental motivation of AI
- extendable
- convert real world probelm to solvable situation
- dataset genreation platform
- make ml model online learning and particiatepe a sa jusdge
- add sidebar agent decision show trace
- this is our cool webpage 파사드
findings
- they are making a local group
- similar position tend to have a same belief eventually
- locality
- contagious
- even investing
어디갔어