How Baidu and Samsung Foundry’s New AI Chip Knows What I’m Thinking Even Before I Do.
My love-hate relationship with the snooze button was back in full swing, and on this first day back after CES I was running late. In Tasmanian devil style, I raced against the clock rushing through my routine. As I bolted into the kitchen, the sight of my dog waiting patiently by the coffee machine with his leash in his mouth stopped me in my tracks. I paused, and smiled. Tired of waiting, he had predicted my next move and met me there. Not surprising that after all these years he knows me and my routine so well.
I guess it shouldn’t surprise me when my smart assistant or search engine is predicting my next move. As we enter the Age of Experience with broad expectations of robot personal assistants and self-driving cars in the not so distant future, consumers may not realize that artificial intelligence (AI) is already hard at work behind the scenes, making our lives easier and services more accessible. Perhaps you’ve seen the clever commercial with the young girl asking “How does it know I like sprinkles on my donut?” It’s an excellent question that’s worth a deeper look at how technology companies like Samsung are feverishly innovating semiconductor chips that can process information in a way that appears like magic.
Reading between the lines
Technology industry predictions regarding increased opportunities for edge-based AI inferencing vs. training are coming to life. Samsung Foundry and Baidu recently announced the development of an AI chip that’s expected to revolutionize Baidu’s search ranking, speech recognition, image processing, natural language processing, autonomous driving, and deep learning tools. Developed on Samsung Foundry’s 14nm process technology, the KUNLUN chip is the magic that enables things like Ernie, a pre-training model for natural language processing, to inference at three times faster than that of conventional GPU/FPGA-accelerating model.
What does this mean to the everyday consumer? Every time we ask our voice assistant a question, or give it a command, we depend on speech recognition and natural language processing. Have you ever wondered how search engines know the results you’re looking for even when you don’t type the words exactly right? Or when the ads for a service you never imagined turn out to be exactly what you need? That’s the magic of this new compilation of the most advanced semiconductor technologies simply reading between the lines.
Predicting my next move
As our Foundry capabilities expand further into datacenter applications by developing and mass-producing AI chips – it’s exciting to know we’re bringing AI learning to the edge. Faster inference means a smarter assistant and more efficient search. Although that’s amazing, it’s so much more than that, the technology we’re immersed in can accurately learn to surmise our needs in a fraction of a second. Compare that to the years it takes family, friends and pets to figure us out. Sounds like something I’ve been dreaming up in-between snoozes, until I wake up to my smart assistant’s greeting. She suggests a new restaurant and correctly orders my breakfast in response to my barely decipherable morning mumblings. While enjoying my new favorite meal, I ponder what new experiences AI has discovered for me that I won’t want to live without. Suddenly my wandering imagination is brought to a halt by a nudge from my dog. The walk he’s been waiting for is the next experience he can’t live without!