Server Repository. control terminal supercomputer cloud storage.

Unbabel among the first AI startups to win millions of GPU training hours on EU supercomputers

Server Repository. control terminal supercomputer cloud storage.

Image Credits: Vladimir_Timofeev / Getty Images

The European Union has announced the winners of a “Large AI Grand Challenge” it kicked off earlier this year in a bid to accelerate the pace of homegrown innovation by large-scale AI model makers.

Four startups will share €1 million in prize money and, perhaps more importantly, 8 million GPU hours to train their models on a couple of the bloc’s high-performance computing (HPC) supercomputers over the next 12 months. The Commission reckons this will enable them to shrink model training times “from years to weeks,” as its PR puts it.

The winning four startups are in alphabetical order: French fintech Lingua Custodia, which does financial document processing using natural language processing (NLP); Belgian startup Textgain, which also uses NLP for text processing but focuses on analysis of unstructured data, such as monitoring social media chatter for hate speech; Latvian startup Tilde, another language specialist that’s focuses on Balto-Slavic languages, offering machine translation and AI-powered chatbots in the target tongues; and Portugal’s Unbabel, which has historically blended machine translation with the expertise of native human speakers and applying AI for customer service and productivity use cases for enterprise customers.

The Commission said the AI Challenge received a total of 94 proposals.

Unbabel likely has the highest profile of the four winners. The Y Combinator-backed translation business has been around for the best part of a decade and raised close to $100 million over its run, per Crunchbase.

Whether Unbabel needs an extra quarter million euros or even 2 million freebie GPU training hours is up for debate, but even veteran AI startups may feel every little bit helps, given the fast-paced developments in generative AI over the past 1.5 years or so.

At the end of the training period, the EU expects all the winners to release their developed models under an open source license for noncommercial use or publish their research findings. 

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 state of the union address last fall, saying at the time that it wanted “ethical and responsible AI startups” to be first in line to tap computational support.

The European High Performance Computing Joint Undertaking (aka EuroHPC JU) — to give the bloc’s supercomputer initiative its full name — currently has eight operational (nine procured) supercomputers, and two of which will be providing the allocation of 8 million GPU hours to the four winners: namely, Finland-based Lumi and Italy-based Leonardo (which are both pre-exascale HPC supercomputers).

A fifth startup, Spain-based Multiverse Computing, which is focusing on trying to improve the energy efficiency and speed of large language models using “quantum-inspired tensor networks,” just missed out on any prize money. But there’s a consolation: It will be allocated 800,000 computational hours on another of the supercomputers, Spain’s (pre-exascale) MareNostrum 5.

This handful of European startups building large-scale AI models won’t be the first to get a taste of what HPC hardware can do. French general purpose AI model maker Mistral was a participant in an early pilot phase of the supercomputing provision last summer, using Leonardo to “run a few small experiments,” as co-founder and CEO Arthur Mensch told TechCrunch back in December — though he said it had not been used for model training at that point.

The EuroHPC JU has also historically provided some capacity to commercial players. However demand for the supercomputers typically far outstrips supply, so the AI startups are essentially getting bumped to the front of the queue.

EU policymakers have also recognized there’s a need to reconfigure and retool the HPC infrastructure for the generative AI age. Hence why, in January, the Commission announced a package of “AI innovation” measures that included proposals for upgrading the supercomputers and building out a support layer to improve accessibility so that AI startups can more easily tap the infrastructure.

EU to expand support for AI startups to tap its supercomputers for model training

EU to let ‘responsible’ AI startups train models on its supercomputers

Unbabel among the first AI startups to win millions of GPU training hours on EU supercomputers

Server Repository. control terminal supercomputer cloud storage.

Image Credits: Vladimir_Timofeev / Getty Images

The European Union has announced the winners of a “Large AI Grand Challenge” it kicked off earlier this year in a bid to accelerate the pace of homegrown innovation by large-scale AI model makers.

Four startups will share €1 million in prize money and, perhaps more importantly, 8 million GPU hours to train their models on a couple of the bloc’s high-performance computing (HPC) supercomputers over the next 12 months. The Commission reckons this will enable them to shrink model training times “from years to weeks,” as its PR puts it.

The winning four startups are in alphabetical order: French fintech Lingua Custodia, which does financial document processing using natural language processing (NLP); Belgian startup Textgain, which also uses NLP for text processing but focuses on analysis of unstructured data, such as monitoring social media chatter for hate speech; Latvian startup Tilde, another language specialist that’s focuses on Balto-Slavic languages, offering machine translation and AI-powered chatbots in the target tongues; and Portugal’s Unbabel, which has historically blended machine translation with the expertise of native human speakers and applying AI for customer service and productivity use cases for enterprise customers.

The Commission said the AI Challenge received a total of 94 proposals.

Unbabel likely has the highest profile of the four winners. The Y Combinator-backed translation business has been around for the best part of a decade and raised close to $100 million over its run, per Crunchbase.

Whether Unbabel needs an extra quarter million euros or even 2 million freebie GPU training hours is up for debate, but even veteran AI startups may feel every little bit helps, given the fast-paced developments in generative AI over the past 1.5 years or so.

At the end of the training period, the EU expects all the winners to release their developed models under an open source license for noncommercial use or publish their research findings. 

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 state of the union address last fall, saying at the time that it wanted “ethical and responsible AI startups” to be first in line to tap computational support.

The European High Performance Computing Joint Undertaking (aka EuroHPC JU) — to give the bloc’s supercomputer initiative its full name — currently has eight operational (nine procured) supercomputers, and two of which will be providing the allocation of 8 million GPU hours to the four winners: namely, Finland-based Lumi and Italy-based Leonardo (which are both pre-exascale HPC supercomputers).

A fifth startup, Spain-based Multiverse Computing, which is focusing on trying to improve the energy efficiency and speed of large language models using “quantum-inspired tensor networks,” just missed out on any prize money. But there’s a consolation: It will be allocated 800,000 computational hours on another of the supercomputers, Spain’s (pre-exascale) MareNostrum 5.

This handful of European startups building large-scale AI models won’t be the first to get a taste of what HPC hardware can do. French general purpose AI model maker Mistral was a participant in an early pilot phase of the supercomputing provision last summer, using Leonardo to “run a few small experiments,” as co-founder and CEO Arthur Mensch told TechCrunch back in December — though he said it had not been used for model training at that point.

The EuroHPC JU has also historically provided some capacity to commercial players. However demand for the supercomputers typically far outstrips supply, so the AI startups are essentially getting bumped to the front of the queue.

EU policymakers have also recognized there’s a need to reconfigure and retool the HPC infrastructure for the generative AI age. Hence why, in January, the Commission announced a package of “AI innovation” measures that included proposals for upgrading the supercomputers and building out a support layer to improve accessibility so that AI startups can more easily tap the infrastructure.

EU to expand support for AI startups to tap its supercomputers for model training

EU to let ‘responsible’ AI startups train models on its supercomputers

Smartphone and judges gavel on black background

Google dubs Epic's demands from its antitrust win 'unnecessary' and 'far beyond the scope' of the verdict

Smartphone and judges gavel on black background

Image Credits: SomeMeans (opens in a new window) / Getty Images

In a new filing, Google is pushing back against Fortnite maker Epic Games’ numerous proposed remedies after a court determined Google engaged in anticompetitive practices on its Play Store. Following the jury’s decision late last year, the two sides pled their cases about how Google should have to change its behavior in light of the ruling. For its part, Epic Games issued a wild laundry list of demands, which included it gaining access to the Play Store catalog of app and game titles for six years, the ability to distribute its own app store on Google Play with no fees and much more. It also wanted to put an end to all agreements, incentives and deals, as well as penalties that would allow the Play Store or Google Play Billing to gain the upper hand against rivals.

The tech giant’s surprising and swift defeat was a historic ruling, especially since Epic Games largely lost a similar antitrust case with Apple, which had not been tried by jury. In the Epic-Apple lawsuit, the court decided Apple was not a monopolist, but did agree that developers should be able to steer their customers to alternative means to pay via the web. The case was appealed up to the Supreme Court, which declined to hear it, allowing the lower court’s ruling to stand.

While the jury in Google’s case was convinced that the tech giant leveraged its market power in illegal ways, it didn’t get to decide the next steps — that’s up to the judge. The new filing, along with Epic’s proposal, will help to inform Judge James Donato in a hearing scheduled on May 23 about what actions to take next to put Google’s power in check.

Epic Games in April had detailed its demands in a proposed injunction, found here. At a high level, Epic wants Google to allow users to download apps from any app store or the web, depending on their preference. It doesn’t want Google to be able to block or coerce OEMs or carriers to favor Google Play. And it doesn’t want Google to be able to impose additional fees for routing around the Play Store, which Epic Games believes is also an anticompetitive practice.

The Fortnite maker additionally asked the court to enforce other changes, including giving Epic access to the Play Store catalog so it can perform users’ app updates, without warning screens or extra fees. Plus, Epic wants developers to be able to tell its users how to pay for their apps and services elsewhere, and how much they could save by doing so. It wants to eliminate the requirement to use Google’s “User Choice Billing,” which offers only a small discount to developers who process payment transactions themselves, and much more.

Google, of course, disagrees about how the court should proceed.

In a statement, Google Vice President of Government Affairs & Public Policy Wilson White referred to Epic’s demands as overreaching and unnecessary.

“Epic’s demands would harm the privacy, security, and overall experience of consumers, developers, and device manufacturers,” he said. “Not only does their proposal go far beyond the scope of the recent U.S. trial verdict — which we will be challenging — it’s also unnecessary due to the settlement we reached last year with state attorneys general from every state and multiple territories. We will continue to vigorously defend our right to a sustainable business model that enables us to keep people safe, partner with developers to innovate and grow their businesses, and maintain a thriving Android ecosystem for everyone.”

In the injunction filed Thursday in a U.S. District Court in California, Google argues that Epic’s demands put users’ security and privacy at risk as it would remove its ability to implement trust and safety measures about the use of the third-party app stores. (Apple has used a similar strategy to fight against regulations around opening up its App Store to competition, too, saying it’s responsible for users’ privacy and security.)

Plus, Google says that it would be required to tell all third-party app stores, without user consent, which apps a user has installed. This would expose personal apps’ usage, including around sensitive areas like religion, politics or health, without rules about how that data could be used.

The company also said that Epic is asking it to remove safeguards around sideloading apps.

And in case those arguments fail, in another tactic, Google points out that the remedies Epic proposed are not needed because it already settled with state attorneys general to no longer sign wide-ranging exclusivity agreements with developers. Epic’s proposal would additionally prevent Google from working with developers to provide exclusive content through Play Store apps, which it says is an important opportunity for developers.

Finally, the state AG settlement would allow any app store to compete for placement on Android devices, Google said, but Epic’s proposal would cut it out of that process, reducing competition. Without Google involved, rival app stores would underbid, impacting OEM margins, it said.

The judge’s coming decision over the remedy in this case will be an interesting one to watch as it will set the stage for how app stores deemed monopolists will have to make concessions to allow more competition. Though Epic lost its fight with Apple, the Justice Department’s case against the iPhone maker is still underway as is its lawsuit with Google over its alleged search monopoly. The outcome of these cases will determine to what extent tech giants’ power will continue to remain unchecked, given the glaring lack of legislation in the U.S. to reign in the tech monopolies.