[HN Gopher] Are You Still Using Real Data to Train Your AI?
       ___________________________________________________________________
        
       Are You Still Using Real Data to Train Your AI?
        
       Author : headalgorithm
       Score  : 22 points
       Date   : 2022-02-18 10:54 UTC (12 hours ago)
        
 (HTM) web link (spectrum.ieee.org)
 (TXT) w3m dump (spectrum.ieee.org)
        
       | WalterBright wrote:
       | If a computer program is generating the training data, aren't you
       | just training the AI to do the same thing as the already existing
       | computer program does?
        
         | ynfnehf wrote:
         | You can train an AI to do the inverse of the existing program
         | (as is the case for the self-driving described in the article.)
         | Take some input, generate output using the existing program,
         | and then train the AI with the input/output reversed.
        
         | tomp wrote:
         | No. We had realistic 3D graphics 20 years ago.
         | 
         | I'm not aware of _any_ true 3D computer vision system that
         | could reliably play those games (from just vision).
        
         | throwawaynay wrote:
         | Generating a realistic city and a self driving AI are two
         | wildly different tasks
        
           | SomewhatLikely wrote:
           | Even a simpler task like image classification such as: does
           | the picture contain a lion. Imagine you have 3d model of a
           | lion. You can render it from lots of different angles,
           | lighting conditions, backgrounds, stretched out, curled up,
           | etc. You know the ground truth classification on all
           | renderings is that the picture contains a lion, but being
           | able to generate images of lions is a very different task
           | from recognizing lions in images.
        
         | ausbah wrote:
         | at least in the case of reinforcement learning, no. just
         | because you can simulate the problem doesn't mean you know how
         | to optimally solve it - ex driving a car
        
         | btdmaster wrote:
         | Not necessarily:
         | https://en.wikipedia.org/wiki/Data_augmentation
        
       | SomewhatLikely wrote:
       | Pretty clickbaity. Lots of "some argue", "some say", "has
       | estimated", and "striving to", but not much substance about
       | actual successes. I believe both Tesla and Cruise are working in
       | this direction but there are serious issues to be worked out. I
       | also vaguely remember some work on pose estimation being helped
       | by generating renderings. Going over real successes would make
       | for a more convincing article.
        
         | sockpuppet69 wrote:
        
       ___________________________________________________________________
       (page generated 2022-02-18 23:00 UTC)