3 ms·
A simple solution might be to devise a test with non-expensive results for failure. For example, we could teach them to construct and transmit a jpeg using the
by ohyes 15y ago
A simple solution might be to devise a test with non-expensive results for failure.
For example, we could teach them to construct and transmit a jpeg using their words for left and right. If the jpeg comes through flipped, we got it wrong, and switch the words. If it comes through correctly, we got it right.
- exit 15y agofor that to work you would have to have already communicated that the x-axis component in a 2-dimentional image increases rightwards..
- ohyes 15y agoExactly, that we have sent that instructions correctly is what we are testing. Then they take a picture of something that we both can see (like a star), and if the image is actually 'mirror flipped' of what they would see, we know the instructions were wrong. You are testing the instructions, not the picture itself.
- exit 15y agothey would have to send the image back to us also encoded. if they misinterpret the instructions, why would we notice when they use the same encoding scheme to send us back a picture?
- pavel_lishin 15y agoI think one of the constraints is that you can't use illustrations.