Intel reveals low-power Clover Falls AI companion chip for EVO-certified laptops

[Redaktion]

The new Intel Visual Sensing Controller codenamed Clover Falls is coming to Evo-certified laptops and should help preserve battery life by performing some low-power semi-passive system monitoring tasks normally executed by the CPU. It is also supposed to enable PC systems to sense and adapt to their surroundings by connecting to webcam sensors.


https://www.notebookcheck.net/Intel-reveals-low-power-Clover-Falls-AI-companion-chip-for-EVO-certified-laptops.510555.0.html

[Furnace lake]

I must confess
Secretly I use AI features on my ultrabook every day

11th gen is the thing I was waiting for. All my friends might mock me but there is no better way to sort my desktop icons then machine learning. 11th gen speeds up my daily routine by about 1111%

I beg intel to continue upgrading AI. There is no going back after using it for the first time

[kek]

I'm glad I waited. Seems like TigerLake vPro edition might get this AI thing, too.

So far, I'm leaning towards Tiger Lake for my next laptop.

[AI]

AMD will include Computer Vision and Machine Learning (CVML) features with "Van Gogh" and "Dragon Crest" low power APUs, "Rembrandt" APUs, and possibly "Raphael" Zen 4 desktop CPUs.

[vertigo]

This basically sounds like a BIG.little setup, except just one smart little core that uses "AI" (unclear if real or buzzword in this case) to determine what tasks to handle vs relying on the OS for instructions. While that could potentially be useful, I wonder how much better it really is, since software could do the same thing, which would provide the functionality to any device, regardless of specialty AI chips, so long as they have a BIG.little design (which laptop processors desperately need). In fact, depending on a separate AI chip instead of having it coordinated by software seems like it would be at a slight disadvantage, since it's farther away from the CPU, but I suspect that would be negligible, especially since most tasks it's used for wouldn't likely even need to communicate much with the main CPU nor would they, by definition, need high performance, and hardware implementations are always faster than software.

Another thing to keep in mind is that if you care at all about security and privacy, you'll have the webcam closed except when in use, which would make much of what this does useless. If they have any intelligence, they wouldn't tie it in to the webcam, at least not exclusively, and would have it use something else, perhaps a depth sensor, as well as an accelerometer and perhaps proximity sensors in the keyboard/palm rest.

[Fraudtel]

"Intel artificial intelligence" - no one else has it, presumably. The AI won't care how you brand it, Intel.

[_MT_]

This basically sounds like a BIG.little setup, except just one smart little core that uses "AI" (unclear if real or buzzword in this case) to determine what tasks to handle vs relying on the OS for instructions.

To me, it doesn't sound like it at all. It sounds like it has two basic functions. 1) System monitoring, where it's all about efficiency. It might be programmable, but it won't be running software (I imagine it won't be x86 at all). As I said, it's all about efficiency. It might be very slow as monitoring probably doesn't take much computational power (but you waste a lot of energy by waking up the CPU). Who knows. Depends on what exactly it is they're trying to offload. 2) AI, probably focused on reducing power draw. I don't know what you consider proper AI. We're probably talking about computer vision, recognizing people, perhaps hardware acceleration for face recognition (since they mention security as well). Hopefully, it will work with IR cameras used for Hello as well so that you can still use a shutter on the webcam.

What they're trying to do is difficult enough with a webcam. Trying to do it without it is even sillier. Have you never stood further away, off to the side from a laptop but still watched the screen? Even with a camera, who knows how it will work in such a scenario (what is the field of vision, etc.). It's quite annoying when displays automatically dim when you're trying to watch them.

[vertigo]

This basically sounds like a BIG.little setup, except just one smart little core that uses "AI" (unclear if real or buzzword in this case) to determine what tasks to handle vs relying on the OS for instructions.

To me, it doesn't sound like it at all. It sounds like it has two basic functions. 1) System monitoring, where it's all about efficiency. It might be programmable, but it won't be running software (I imagine it won't be x86 at all). As I said, it's all about efficiency. It might be very slow as monitoring probably doesn't take much computational power (but you waste a lot of energy by waking up the CPU). Who knows. Depends on what exactly it is they're trying to offload. 2) AI, probably focused on reducing power draw. I don't know what you consider proper AI. We're probably talking about computer vision, recognizing people, perhaps hardware acceleration for face recognition (since they mention security as well). Hopefully, it will work with IR cameras used for Hello as well so that you can still use a shutter on the webcam.

What they're trying to do is difficult enough with a webcam. Trying to do it without it is even sillier. Have you never stood further away, off to the side from a laptop but still watched the screen? Even with a camera, who knows how it will work in such a scenario (what is the field of vision, etc.). It's quite annoying when displays automatically dim when you're trying to watch them.

That's basically the point of BIG.little, though, to use slow, efficient cores when possible to avoid having to use the more power-hungry ones. Not exactly the same, but similar concept. At least that's how it seems to me, but I'm sure there's more to it.

And yes, certain things will likely not be possible without the webcam and others will be more difficult, but what I'm saying is that, as much as possible, they should try to support the functions with other sensors as well, instead of just relying on the webcam, to allow them to work when the shutter is closed as opposed to simply losing all those functions and making the chip worthless. I guess I've never had the problem you describe, as when I'm using a laptop, I'm using it, and I'm not sure why the screen would dim since I set it to a brightness and it stays there (I assume you're talking about a power saving setting that dims it when not in use), but my point is that a depth sensor should be able to do much of what a camera would do for such tasks. For example, instead of a camera being used to see that you're in front of the computer, even if several feet away, a depth sensor could do the same, thereby keeping the screen brightness up. There would obviously be some limitations, but it would still work far better than relying solely on the camera and having the shutter closed.