HN user
The core difference with AI is that it is adaptable. Sure, flailing around and happening to do something useful isn't clever, but if we can reinforce useful behaviours something more closely resembling intelligence can emerge.
Only if people don't know about the dummies, things tend toward zero sum with awareness of the situation.
http://www.progonos.com/furuti/MapProj/Normal/ProjPoly/projP... These maps aren't too bad.
The helium reduces resistance which results in lower vibration of the read/write heads, which in turn allows for reduced space between the platters and the heads, which allows for a reduced space between the platters and more platters per drive. If the same drive was filled with air rather than helium, the increased vibration of the heads would cause them to crash into the platters - which would scrub the disk clean (which would be bad).
Deep learning has really only recently become successful with new learning algorithms such as constrastive divergence and convolutional neural networks. Previous efforts were focused around backpropagation, but due to the signal loss across many layers there was never enough information in the output layer to successfully train the network.
Another issue is that when a big drop falls, the small drops that it leaves behind are equally sized and equally spaced.
OK, but that's a very large data set that to track - does AV software actually do that?
A lot of new malware is self-modifying. In situations where this is the case checksums lose most if not all of their value. Heuristics in such situations are really the only viable option.
There really needs to be a change to patent law around independent derivation of a concept. At very least we need to look into generalised thicket busting laws. The current situation is fundamentally unscalable.
Concatenation of the hashes seems like an unjustified risk, that in certain circumstances will allow weaknesses from either algorithm to flow throw to the final hash. If you really want to combine the hashes, XOR seems like a safer bet to me (since the algorithms are unrelated there should be no potential cancellation of entropy).