Discussion in 'Education & Personal Growth' started by Amulet, Nov 25, 2018.
Why wait for anyone to ask? Here it is.
what things were you asked ? easier to doodle a golf club than a car or a truck. If our languages deteriorate to the point of tiny drawings (emoji's), you'd have trouble in spelling bee's.
I was asked to draw a goatee, a feather, a mushroom, a chandelier, a finger and a map.
If it eases your gutted ego, here's my performance narrative.
First word: school bus.
I spent 10 seconds to identify the distinguishing factor between school bus and ordinary bus. How do I indicate to the learning bot that it was a *school bus*. 5 more seconds gone. Then I started to draw a line. (prompt guess at the bottom: Tooth brush?). I was distracted for another 3 seconds on that thought prompt at the bottom of the screen. Then I drew another line (prompt guess at the bottom: window). Time up!
Second word: garden
Why such ill-luck! Again I spent 10 seconds on how to subtlize it as a garden from greenhouse and jungle and rainforest. (tick tick) I drew a rough tree. Time up!
Third word: lobster
Huh! But there are two types of lobsters. Which one did you mean, bot? Clawed lobster or spiny lobster.
Sketching all objects wrong means that we are dedicated and righteous about teaching the neural pupil the precise visuals of the object. I could not have taught it that a clump of plants is a garden and a segmented crust is a lobster. Botty, I will teach you only artfully to perceive with exactitude.
Map and chandelier are challenging!
Actually, map is more challenging to insinuate a representation out of an unshapely outline.
I didn't think of your last paragraph and it helped my self-confidence.
Two lessons out of this experience:
1) I have a lot to learn and it is worth leading the rest of my life learning.
2) Creativity and time are in the opposite spectrum.
I got the other three right.
(dolphin, crayon, wristwatch)
But, when a human would not have recognized my oddly-shaped dolphin, how did the bot do it? Unless everyone drew it sloppy as an inflated blob. Ask me, my dolphin sketch resembles a rogue leaf more than any living organism.
I have my son right in the next room and I see the visual of his goatee everyday. So much for my observation capacity. Feather should have made me think of Lord Krishna and I would have drawn it right in 10 seconds. I get mushrooms every now and then in my yard. I have two chandeliers at home. In order to draw a finger, I need to look at my own finger typing it in the keyboard. I don't know why I thought of India for a map.
I guess, I need to be as observant as @Srama to be good in doodles and live in the present moment. If I am able to relate all the objects around me, I would have succeeded in my attempt.
The AI imposes the scanned outline onto a filtering shape and not onto the common features.
In the below game result, nose, necklace, cactus, belt, windmill were matched.
But boomerang was booted out because the preferential representations of boomerang were flipped (next page examples). My boomerang even matched a hot dog
In the below array, a human would have instantly recognized only that boomerang.
What you have drawn is nowhere closer to hot dog and it is close to boomerang. I identified following issues in my skill-set:
1) Learning how to move the cursor properly & quickly to draw a sketch (provide for a fraction of a second time delay in the intended sketch)
2) Observe the question longer before clicking "draw" (time saving)
3) Don't read the messages from AI when drawing a sketch (distraction)