Topics
Latest
AI
Amazon
Image Credits:Figure AI
Apps
Biotech & Health
Climate
Image Credits:Figure AI
Cloud Computing
commercialism
Crypto
Image Credits:Figure
Enterprise
EVs
Fintech
Image Credits:Figure
Fundraising
contraption
Gaming
Government & Policy
Hardware
Layoffs
Media & Entertainment
Meta
Microsoft
privateness
Robotics
Security
Social
place
inauguration
TikTok
Transportation
speculation
More from TechCrunch
Events
Startup Battlefield
StrictlyVC
Podcasts
Videos
Partner Content
TechCrunch Brand Studio
Crunchboard
meet Us
flesh father and CEO Brett Adcock Thursdayrevealeda new machine encyclopedism model for humanoid robots . The intelligence , which arrive two week after Adcock announced the Bay Area robotics firm’sdecision to mistreat forth from an OpenAI collaborationism , is rivet around Helix , a “ generalist ” Vision - Language - Action ( VLA ) exemplar .
VLAs are a newfangled phenomenon for robotics , leverage vision and language command to serve selective information . presently , the best - sleep with illustration of the category isGoogle DeepMind ’s RT-2 , which trains robots through a combination of video and prominent speech communication models ( LLMs ) .
spiral works in a interchangeable fashion , fuse optical data and speech prompt to moderate a automaton in real time . Figure write , “ Helix displays firm object generalization , being able to pick up thou of new household items with varying shape , sizes , colors , and cloth holding never encountered before in training , simply by asking in natural speech communication . ”
In an ideal world , you could simply tell a robot to do something and it would just do it . That is where Helix occur in , allot to Figure . The platform is designed to bridge the col between vision and language processing . After receiving a born language voice prompting , the automaton visually assess its environment and then performs the task .
Figure pop the question examples like , “ Hand the bag of cookie to the automaton on your rightfulness ” or , “ Receive the bag of cookies from the robot on your left and range it in the open draftsman . ” Both of these exercise involve a distich of robots working together . This is because Helix is designed to control two golem at once , with one assisting the other to perform various home project .
Figure is showcasing the VLM by highlight the work the company has been doing with its 02 android robot in the home environs . Houses are notoriously tricky for robots , give they miss the structure and consistency of warehouse and manufacturing plant .
difficultness with learning and control are major hurdles standing between complex robot systems and the home . These effect , along with five- to six - digit price tags , are why the home robot has n’t taken precedence for most humanoid robotics caller . Generally speaking , the approach is to build robots for industrial clients , both improving reliableness and bringing down costs before tackling domicile . housekeeping is a conversation for a few years from now .
Join us at TechCrunch Sessions: AI
Exhibit at TechCrunch Sessions: AI
When TechCrunchtoured Figure ’s Bay Area officesin 2024 , Adcock indicate some of the paces its humanoid was being put through in the home base setting . It appeared at the prison term that the work was not being prioritize , as Figure focus on workplace pilots with corporation like BMW .
With Thursday ’s Helix announcement , Figure is making it readable that the home should be a antecedency in its own right . It ’s a challenging and complex setting for testing these kind of education good example . learn robots to do complex tasks in the kitchen — for model — opens them up to a broad orbit of actions in different options .
“ For golem to be utile in households , they will need to be up to of generating intelligent new behaviors on - need , peculiarly for objects they ’ve never seen before , ” Figure says . “ Teaching robots even a individual new doings presently ask substantial human effort : either hours of PhD - story expert manual programming or thousands of demonstration . ”
Manual programing wo n’t surmount for the nursing home . There are simply too many terra incognita . Kitchens , living way , and john diverge dramatically from one to the other . The same can be order for the tools used for cooking and cleaning . Besides , people entrust jam , rearrange furniture , and prefer a range of unlike environmental lighting . This method takes right smart too much meter and money — though Figurecertainly has plenty of the latter .
The other option is training — and lots of it . robotlike arms trained to pick and property target in laboratory often apply this method . What you do n’t see are the hundreds of hours of repeating is acquire to make a demonstration robust enough to take on extremely varying task . To pick something up right the first clock time , a robot need to have done so century of time in the past .
Like so much surrounding humanoid robotics at the moment , piece of work on Helix is still at a very early stage . Viewers should be advised that a lot of workplace happens behind the view to create the kinds of short , well - produced videos interpret in this C. W. Post . Today ’s proclamation is , in heart and soul , a recruiting shaft plan to bring more locomotive engineer on board to avail produce the task .
