1
$\begingroup$

I read a tweet from Elon Musk where describes Gradient descent as an evil action that AI are good at, despite the fact that it is just one of the old, inflexible and not-so-efficient error correction algorithms.

He is an intelligent man why would he say something like this?

Is gradient descent a backpropagation that also lacks the recursion and neural plasticity? Or is it suddenly became an black magic AI throws at its enemies?

From an article:

"Musk indicates that internet infrastructure is “particularly susceptible” to a method called gradient descent algorithm, a mathematical problem-solving process. Bad news is, AI is excellent at doing gradient descents, which can become devastating digital weaponry."
Source: Futurism.com

Snapshot of Musk Tweet

$\endgroup$
  • $\begingroup$ At least quote the entire tweet verbatim. $\endgroup$ – BlindKungFuMaster Apr 21 '17 at 6:44
  • $\begingroup$ @BlindKungFuMaster added $\endgroup$ – Aus Apr 21 '17 at 6:51
  • $\begingroup$ @Aus "it is just one of the old and unflexible and not-so-effecient error correction algorithm" What the heck are you saying? $\endgroup$ – nbro Apr 21 '17 at 11:17
  • $\begingroup$ @nbro, Compare it to feed-forward neural-modulation or hebian neural plasticity. Back propagation has its limitations. and it is not a thing that AI is good it, it is an algorithm used to calibrate weights for a NN. $\endgroup$ – Aus Apr 21 '17 at 11:28
  • 1
    $\begingroup$ I believe Musk would make this comment because he is aware of the valid questions raised by Neo-Luddism in relation to the dangers of technology. Rather than reject technology, Musk is trying to raise awareness at a time when "narrow" Algorithmic Intelligence is exceeding humans at specific tasks, and may advance beyond our capacity to understand what they are doing. My personal feeling is that Musk's industrial endeavors are not primarily motivated by personal profit, but out of a deep concern for the future of our species, and uses his celebrity as a platform to raise awareness of issues. $\endgroup$ – DukeZhou Apr 21 '17 at 16:25
4
$\begingroup$

I think Musk was using the terminology correctly though perhaps with hyperbole. I believe this was tweeted in the context of the botnet attacks on name-resolution services that broke Netflix and a large number of other internet services for a time. He was expressing the idea that you could train a botnet-based system to attack the internet by giving it a toolbox of targets and hacks and using well-known machine-learning techniques to optimize the effectiveness of using the attacks in combination to take down the internet. Gradient descent itself isn't harmful; it is using gradient descent to train a botnet that could theoretically result in a really devastating attack.

I also don't think he is necessarily implying that there is an actual machine-learning system at work; rather that the hackers were training themselves with a methodology analogous to gradient descent.

$\endgroup$
  • $\begingroup$ Thank you for this excellent context. This subject may require hyperbole at the present state of AGI, but I think it's important to start attempting to tie moral philosophy to Algorithmic Intelligence, particularly in a game theoretic (economic) context, which has many applications in computing, and may be said to form the basis for Algorithmic Intelligence in terms of optimal decision making in the context of bounded rationality. As someone engaged in research on general economic automata in partisan contexts, deeply concerned about value alignment, I think this is of paramount importance. $\endgroup$ – DukeZhou Apr 21 '17 at 16:18

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.