|
This chapter on AIT and machine learning was nice, don’t you think? |
|
Sure! I liked the idea that analogy is just a matter of minimizing complexity. |
|
|
|
|
Oh, it’s just a special case of induction. |
|
No, I disagree. Analogy is a form of reasoning. |
|
|
|
|
Yes, but, induction is more general. I liked the last thing about universal induction and universal AI. |
|
Yes, but it’s not computable. |
|
|
|
|
|
It helps me understand that efficient reinforcement learning depends on your ability to compute the future. |
|
To me, analogy and MDL are closer to practical use. I like the idea of optimal trade-off between model simplicity and accuracy. |
|
|
|
|
And so you must be exited to see that any form of learning amounts to performing data compression. |
|
Yes, we talked about that! |
|
|
|
|
|
Yeah. As for me, my favorite remains the idea that anomalies must be simple to characterize to be genuine anomalies. |
|
Yes, that one also! |
|
|
|
|
