Today’s most attractive and hot AI tools (e.g. OpenAI’s ChatGPT, Microsoft’s Bing, Google’s Bard, Adobe’s Photoshop) run in data centers packed with powerful and expensive servers. But on Monday, Intel revealed details of its upcoming Meteor Lake PC processors that could help laptops play a bigger role in the artificial intelligence revolution.
Meteor Lake, due in computers later this year, includes circuitry to speed up some AI tasks that can drain your battery. For example, it could improve AI for blurring or replacing backgrounds during video conferencing, said John Layfield, lead of client AI work at Intel.
AI models use methods inspired by the human brain to recognize patterns in complex real-world data. By running AI on the processor of your laptop or mobile phone instead of in the cloud, you get benefits such as better privacy and security as well as more nimble responses as there is no network latency. .
What’s unclear is how much AI work will actually move from the cloud to the PC. Some software, such as Adobe Photoshop and Lightroom, make extensive use of AI for finding people, skies, and other subjects in photos, as well as many other image editing tasks. The app can recognize your voice and transcribe it to text. Microsoft has built an AI chatbot called Windows Copilot directly into its operating system. However, most computing work today uses the more traditional parts of the processor: central processing unit (CPU) and graphics processing unit (GPU) cores.
If you build it, it might come true. Adding AI acceleration directly to the chip, as is already done with smartphone processors and Apple M-series Mac processors, could encourage developers to create more software that leverages his AI capabilities. There is a nature.
However, GPUs are already very good at accelerating AI, and developers don’t have to wait for millions of people to upgrade their Windows PCs to take advantage of it. GPUs offer the best AI performance on his PC, but new AI-dedicated accelerators are better suited for lower power consumption, Layfield said. Both can be used simultaneously for best performance.
Meteor Lake, Intel’s flagship chip
There are other reasons why Meteor Lake is important. It’s designed for low-power operation, which is perhaps its single biggest competitive weakness compared to the Apple M-series processors. Graphics acceleration has been upgraded, which is important for games and even for some AI tasks.
The processor is also key to Intel’s years of restructuring efforts. It is the first large chip built on Intel 4, a new manufacturing process essential to catch up with chip manufacturing leaders Taiwan Semiconductor Manufacturing (TSMC) and Samsung. It also employs a new advanced manufacturing technology called Foveros that allows Intel to stack multiple “chiplets” onto a single, more powerful processor more flexibly and economically.
Chipmakers are scrambling to capitalize on the AI revolution, but few have been as successful as Nvidia, which reported a quarterly blowout in early May as demand for its top-end AI chips exploded. Intel also sells AI chips for data centers, but it emphasizes economy over performance.
Intel calls its AI accelerators Vision Processing Units (VPUs) in its PC processors. It’s a product family and name derived from his acquisition of AI chip maker Movidius in 2016.
These days, a variation called generative AI can create realistic images and text with human-like speech. Meteor Lake can run his one such image generator, Stable Diffusion, but a large AI language model like ChatGPT doesn’t fit on a laptop.
However, it takes a lot of work to change this. Both Facebook’s LLaMA and Google’s PaLM 2 are large language models designed to scale down to small “client” devices such as PCs and mobile phones with much less memory.
“AI in the cloud has latency, privacy, security challenges and is fundamentally expensive,” said Layfield. “Over time, more will be migrated to the client as computing efficiency increases.”
