interesting paradox. I watched this interesting video yesterday talking about how a google AI mistakenly took a Husky dog for a wolf. When Programmers went back and had the code analyze itself, to show the researchers what the AI was focusing on in order for them to understand how it made this mistake, the AI reported back that instead of focusing on the dogs face (maybe the eyes, snout, teeth, ears, etc.) the AI focused on everything in the background. They discovered that when teaching the AI memory recall based off images, the must have used photos of Wolfs that were often depicted with snow in the background. It just so happened that the photo of the Husky was depicted with a snowy background as well, and as such, the AI thought it was a wolf.
interesting paradox. I watched this interesting video yesterday talking about how a google AI mistakenly took a Husky dog for a wolf. When Programmers went back and had the code analyze itself, to show the researchers what the AI was focusing on in order for them to understand how it made this mistake, the AI reported back that instead of focusing on the dogs face (maybe the eyes, snout, teeth, ears, etc.) the AI focused on everything in the background. They discovered that when teaching the AI memory recall based off images, the must have used photos of Wolfs that were often depicted with snow in the background. It just so happened that the photo of the Husky was depicted with a snowy background as well, and as such, the AI thought it was a wolf.