Say you want to draw a circle with radius R in location X,Y on your screen. To do this in the CPU you would need to do all the math to figure out which pixels should light up and which should not in order to make the circle, and then send that pixel data to the monitor. This takes a lot of CPU resources and will generally be very slow.

What a GPU does is simplify common operations like this for the CPU. Instead of doing all the math, the CPU can send a command to the GPU to draw the circle with the same parameters, and the GPU will handle it automatically. This will free up the CPU to do other stuff. Additionally the GPU will have dedicated circuits in the chip that can do all the math and draw to the screen much faster than the CPU.

So technically yes the CPU could send direct to the display, but a separate GPU will accelerate a lot of the common math needed to draw the images to the screen.