3 ms·

Ironically, it may end up being one of these "tiny" scenarios which finally does Google in. When trying to illustrate a problem or bug, one of the typically ti

by blunte 5y ago

Ironically, it may end up being one of these "tiny" scenarios which finally does Google in. When trying to illustrate a problem or bug, one of the typically time consuming challenges is reducing the scenario to the minimal case which illustrates the problem. So thank you, @emilyldolson! Aside from an empty file, you cannot reduce this any further. It brings to light in simple terms that non-techies can understand how absurd the "ML to solve everything" promise is -- and even moreso how wilfully negligent companies are by providing NO human intervention or support when the machines break down.

trhway 5y ago
> "tiny" scenarios which finally does Google in Danger of tiny scenarios - I expect that Google like any BigCo will try files containing only 2, 3, 7 - no bug, ok, and then push the fix like this if (l = read(file) ; l == "1") ... else
- cgrealy 5y ago
  Given it's almost certainly a ruleset generated by an ml agent, it's more likely to be a change in the training data.
  - bonzini 5y ago
    Or just don't flag any file below 100 bytes.
    - tyingq 5y ago
      They probably want to flag some urls or torrent references.
quickthrower2 5y ago
It's a real motivator to leave Google and the MAG cloud in general. It has at least reminded me again to do regular Google takeouts.