AI developer platform, Hugging Face has settled a lawsuit with FrendliAI, a Korean AI startup which accused it of patent infringement.
According to a Friday filing in the US District Court for the Northern District of Delaware, FriendliAI entered into a “confidential agreement” with Hugging Face on January 8. The company has agreed to dismiss its lawsuit “with prejudice,” which means the case cannot be brought back to court.
The settlement comes almost two years after FriendliAI filed a lawsuit against Hugging Face over the infringement of one of its patents. The Korean AI startup however reached a settlement on Friday.
“FriendliAI and Hugging Face hereby stipulate to dismissal … without costs and fees to any party of this case,” reads the filing.
“The court shall retain jurisdiction over the parties’ settlement agreement.”
Court papers.
In its lawsuit, which was filed nearly two years ago, the Korean AI startup accused Hugging face of violating its patent on “batching with iteration-level scheduling.” FriendliAI also claims that the patented technology described in an application to the US Patent and Trademark Office improves the process of batching data fed into the AI system to allow the system to handle multiple requests at once.
According to Tech Crunch, the patent also covers AI systems that enable finished requests from within a batch to be sent to a user and new requests to be added to the batch before the full batch has been processed.
According to a complaint filed on July 28, 2023, in the US District Court for the District of Delaware, Hugging Face’s “Text Generation Inference” server for its large language model AI used the Korean AI startup’s patented method to batch incoming requests to increase the tool’s efficiency.
As such, FriendliAI wanted compensation from Hugging Face, enhanced damages for willfulness, in addition to a court order preventing further infringement or awarding a licensing fee and attorney’s fees and costs.
According to a Bloomberg report, FriendliAI’s method which the startup refers to as “batching with iteration level scheduling” and markets as Periflow or Orca, enhanced the process of batching inputs to allow the AI to process multiple requests.
FriendliAI also has an office in Redwood City, California. The VC-backed startup was founded in 2021 and primarily develops infrastructure solutions for the fast-growing AI tech.
According to TC, Hugging Face has raised a total of $235 million from backers, including Google, Amazon, Nvidia, Intel, and Salesforce. It is one of the largest repositories of AI tools and models in the world.
Apart from hosting models and tools, the Brooklyn-based Hugging Face develops its own, and it offers consulting services to help enterprises fine-tune, customize, and deploy AI.
Meanwhile, tech giant Microsoft has officially released its latest language model, Phi-4, on Hugging Face. According to AINews, the model is available under the MIT licence, which allows for broad usage by developers, researchers, and businesses alike.
The move has been described as a significant step toward democratizing AI innovations. Phi-4 was unveiled in December last year.
A Step-By-Step System To Launching Your Web3 Career and Landing High-Paying Crypto Jobs in 90 Days.