Well you have to run the AI, which is significantly more costly than just running A*
Also, the AI has a finite size, which means it has to be scaled up for bigger graphs. I doubt it would work at all for non-euclidian 2d graphs.
The exciting thing is that they made an AI do a toy problem, not that it was efficient.
But people are dumb and tech companies have brain control over them. "AI good and solve all problems"
It reminds me of the "slime mold solves mazes" thing. It was interesting and there were even things to learn from it, but it wasn't magic and couldn't give us better general tools.
Comments
Well you have to run the AI, which is significantly more costly than just running A*
Also, the AI has a finite size, which means it has to be scaled up for bigger graphs. I doubt it would work at all for non-euclidian 2d graphs.
The exciting thing is that they made an AI do a toy problem, not that it was efficient.
But people are dumb and tech companies have brain control over them. "AI good and solve all problems"
It reminds me of the "slime mold solves mazes" thing. It was interesting and there were even things to learn from it, but it wasn't magic and couldn't give us better general tools.