Cristiano Amon, president and CEO of Qualcomm, speaks throughout the Milken Institute World Convention on Might 2, 2022, in Beverly Hills, Calif.
Patrick T. Fallon | AFP | Getty Pictures
Qualcomm and Meta will allow the social networking firm’s new giant language mannequin, Llama 2, to run on Qualcomm chips on telephones and PCs beginning in 2024, the businesses introduced in the present day.
To date, LLMs have primarily run in giant server farms, on Nvidia graphics processors, because of the expertise’s huge wants for computational energy and knowledge, boosting Nvidia inventory, which is up greater than 220% this 12 months. However the AI growth has largely missed the businesses that make vanguard processors for telephones and PCs, like Qualcomm. Its inventory is up about 10% thus far in 2023, trailing the NASDAQ’s achieve of 36%.
The announcement on Tuesday means that Qualcomm needs to place its processors as well-suited for A.I. however “on the sting,” or on a tool, as a substitute of “within the cloud.” If giant language fashions can run on telephones as a substitute of in giant knowledge facilities, it may push down the numerous value of operating A.I. fashions, and will result in higher and sooner voice assistants and different apps.
Qualcomm will make Meta’s open-source Llama 2 fashions obtainable on Qualcomm units, which it believes will allow purposes like clever digital assistants. Meta’s Llama 2 can do most of the similar issues as ChatGPT, however it may be packaged in a smaller program, which permits it to run on a cellphone.
Qualcomm’s chips embrace a “tensor processor unit,” or TPU, that’s well-suited for the sorts of calculations that A.I. fashions require. Nonetheless, the quantity of processing energy that’s obtainable on a cell gadget pales compared to a knowledge middle stocked with cutting-edge GPUs.
Meta’s Llama is notable as a result of Meta printed its “weights,” a set of numbers that helps govern how a specific AI mannequin works. Doing it will permit researchers and finally business enterprises to make use of the AI fashions on their very own computer systems withotu asking permission or paying. Different notable LLMs, like OpenAI’s GPT-4, or Google’s Bard, are closed-source, and their weights are carefully held secrets and techniques.
Qualcomm has labored with Meta carefully previously, notably on chips for its Quest digital actuality units. It has additionally demoed some A.I. fashions operating slowly on its chips, such because the open supply picture generator Steady Diffusion.
