Topics
in style
AI
Amazon
Image Credits:Vladimir_Timofeev / Getty Images
Apps
Biotech & Health
Climate
Image Credits:Vladimir_Timofeev / Getty Images
Cloud Computing
Commerce
Crypto
endeavor
EVs
Fintech
Fundraising
gizmo
gage
Government & Policy
ironware
Layoffs
Media & Entertainment
Meta
Microsoft
Privacy
Robotics
security system
societal
outer space
Startups
TikTok
transport
speculation
More from TechCrunch
event
Startup Battlefield
StrictlyVC
Podcasts
Videos
Partner Content
TechCrunch Brand Studio
Crunchboard
Contact Us
The European Union hasannouncedthe winners of a “ Large AI Grand Challenge ” it kicked off sooner this twelvemonth in a bid to accelerate the pace of homegrown excogitation by large - scale AI model manufacturer .
Four startups will share € 1 million in prize money and , perhaps more significantly , 8 million GPU hours to civilize their model on a couple of the bloc ’s high - public presentation computing ( HPC ) supercomputer over the next 12 months . The Commission view this will enable them to funk model training time “ from years to weeks , ” as its PR pose it .
The winning four startups are in alphabetical order : Gallic fintechLingua Custodia , which does financial text file processing using rude language processing ( NLP ) ; Belgian startupTextgain , which also use natural language processing for text processing but focuses on analytic thinking of amorphous data , such as monitoring societal media chatter for hate speech ; Latvian startupTilde , another terminology specializer that ’s focuses on Balto - Slavic languages , offer car interlingual rendition and AI - power chatbots in the target tongues ; and Portugal’sUnbabel , which has historically blended machine version with the expertise of aboriginal human speakers and applying AI for customer inspection and repair and productiveness utilize cases for enterprise customer .
The Commission enounce the AI Challenge received a amount of 94 proposals .
Unbabel in all likelihood has the highest visibility of the four success . The Y Combinator - back up transformation business has been around forthe upright part of a decadeandraised close to $ 100 millionover its run , per Crunchbase .
Whether Unbabel needs an extra after part million euro or even 2 million freebie GPU education hours is up for debate , but even veteran AI startups may feel every minuscule bit help , given the fast - paced developments in procreative AI over the past 1.5 age or so .
🎉 Unbabel is one of four AI startup from Europe to win the Large AI Grand Challenge , divvy up € 1 M and 8 M computational hours , advance Europe ’s AI leadership ! 🚀 vast thanks to our amazing squad who made this accomplishment possible ! 👏 understand more 👉 https://t.co/anjf71P7pYpic.twitter.com/ZIodZFODYz
At the ending of the training period of time , the EU expects all the winners to release their grow models under an open source license for noncommercial use or publish their research findings .
Join us at TechCrunch Sessions: AI
Exhibit at TechCrunch Sessions: AI
EU supercomputers to support AI startups
The EU unveiled a plan to expand startup access to the bloc ’s supercomputing hardware in President Ursula von der Leyen ’s commonwealth of the union addresslast nightfall , saying at the time that it desire “ ethical and responsible AI inauguration ” to be first in line to beg computational support .
The European High Performance Computing Joint Undertaking ( aka EuroHPC JU ) — to give the axis ’s supercomputer initiative its full name — currently haseight operational ( nine pimp ) supercomputer , and two of which will be providing the apportionment of 8 million GPU hours to the four winners : namely , Finland - based Lumi and Italy - ground Leonardo ( which are both pre - exascale HPC supercomputer ) .
A 5th startup , Spain - basedMultiverse Computing , which isfocusing on stress to improve the energy efficiency and speed of large language modelsusing “ quantum - barrack tensor networks , ” just neglect out on any prize money . But there ’s a consolation : It will be apportion 800,000 computational hours on another of the supercomputers , Spain ’s ( pre - exascale ) MareNostrum 5 .
This handful of European startups building great - scale AI model wo n’t be the first to get a appreciation of what HPC hardware can do . Gallic general purpose AI example Godhead Mistralwas a participant in an former pilot burner form of the supercomputing supply last summer , using Leonardo to “ run a few modest experimentation , ” as co - laminitis and CEO Arthur Mensch narrate TechCrunchback in December — though he read it had not been used for model education at that point .
The EuroHPC JU has also historically provided some content to commercial player . However demand for the supercomputers typically far outstrips supply , so the AI startups are fundamentally getting bump to the front of the queue .
EU policymakers have also distinguish there ’s a motivation to reconfigure and retool the HPC infrastructure for the generative AI age . Hence why , in January , the Commission announced a bundle of “ AI introduction ” measures that include proposals for upgrade the supercomputers and building out a support level to improve handiness so that AI inauguration can more well exploit the base .
