PC やスマートフォンなどの表示画面の画像処理用のチップとして用いられているほか, AI における膨大な計算処理にも利用されているものはどれか。
令和2年度 ITパスポート試験 問65
This question's explanation and your answers so far.
問 65
Incorrectア
AR stands for Augmented Reality, the name of a technology that overlays information onto real-world images. It is not a chip (a semiconductor component), so it applies neither to an image-processing chip nor to AI computation.
Incorrectイ
DVI stands for Digital Visual Interface, a standard (connector/interface) for connecting a PC to a display to transmit video signals. It is a standard that sends video, not a chip that performs image processing or AI computation.
- ウ
正解 The correct answer is choice ウ, GPU (Graphics Processing Unit). A GPU is originally a chip specialized in image processing (rendering) for the display screens of PCs and smartphones, and it has a structure that performs parallel computation with many arithmetic cores so as to handle a large number of pixels at the same time. This characteristic of carrying out the same computation massively in parallel fits well with the huge matrix and vector operations of AI (deep learning), so it is also widely used for AI computation (using a GPU for general-purpose computation other than image processing is called GPGPU). Only the GPU satisfies both conditions in the stem: being a chip for image processing and also being used for the massive computation in AI.
Incorrectエ
MPEG is the name of the group that established the encoding methods (standards) for compressing video and audio, as well as the general term for those compression formats. It is a data format/standard, not a chip that carries out computation, so it does not apply.
- A GPU is a parallel-computing chip that handles image rendering (image processing for the display screen), and its high parallel-computing power is also used for the massive computation of AI and deep learning (using it for general-purpose computation other than images is GPGPU).
- AR (a technology name), DVI (a video-transmission standard), and MPEG (a video-compression method) are all distinguished from the GPU in that none of them is a chip that performs computation.