2
$\begingroup$

I watched a youtube clip of Elon Musk talking about his view on the future of AI. He gave two examples. One of the examples was a benign scenario and the other example was a nonbenign scenario where he speculated the possibilities of future AI threats and what harm a deep intelligence could do.

According to Elon, a deep intelligence in the network could create fake news and spoof email accounts. "The pen is mightier than the sword". This non-benign scenario put forth by Elon was a hypothetical, but he went into detail about how it could have been possible that an AI, with the goal of maximising the portfolio of stocks, to go long on defense and short on the consumer, and start a war.

To be more specific, this could be achieved by hacking into the Malaysians Airlines aircraft routing server, and when the aircraft is over a warzone, send an anonymous tip that there is an enemy aircraft flying overhead which in turn would cause ground to air missiles to take down what was actually a "commercial" airliner.

Although this is a plausible hypothetical nonbenign scenario of AI, I'm wondering if this actually could have been the case regarding the Malaysian Airliner crash. The Stuxnet, for example, was a malicious computer worm, first uncovered in 2010. Thought to have been in development since at least 2005 and believed to be responsible for causing substantial damage to Iran's nuclear program. The Stuxnet wasn't even an AI.

The Stuxnet blew the world's minds when it was discovered. The sheer complexity of the worm and the amount of time it took to build was impressive, to say the least.

In conclusion, was the Malaysian Airliner crash caused by a non-benign artificial intelligence system?

$\endgroup$
  • $\begingroup$ Stuxnet is not AI, and there is no need to posit AIs with amazing capabilities. Any reasonable definition of "hacking" for instance is beyond what any modern AI could do. This does not rule out that an event is caused by malicious intent from some "controlling" political cause, and it doesn't rule out hacking (although there's nothing that I can see here that points to it, it just seems another conspiracy theory). For now, any "smart" parts to such an event will be directed by people, and software will not be AI, but more direct, like a pre-installed back door or scanning for exploit. $\endgroup$ – Neil Slater Jun 14 '18 at 13:31
  • $\begingroup$ I never said Stuxnet was AI. "The Stuxnet wasn't even an AI...." $\endgroup$ – PandaSurge Jun 14 '18 at 13:33
  • 1
    $\begingroup$ This is just a conspiracy theorist's view. There is no reason to believe any AI was involved. $\endgroup$ – Oliver Mason Jun 14 '18 at 13:36
  • $\begingroup$ Knowing the current state of AI research, I am pretty certain that there are not any meaningful AIs out there hacking into systems to enact some plan. The AI I would most be concerned about today is social profiling for targetting of "fact light, emotion heavy" political messages, or for Gerrymandering etc. That would include automated promotion of news articles and opinionss about events such as the Malaysian Airliner, in an attempt to astro-turf one or more conspiracy theories that suit a political agenda. There is evidence that both Brexit and US Election were targeted by this kind of AI. $\endgroup$ – Neil Slater Jun 14 '18 at 13:37
  • 1
    $\begingroup$ This isn't a discussion site. Maybe post in chat when you have enough rep if you want something open-ended. By "meaningful" I mean something with capabilities implied by your question. I can know this with reasonable certainty because I follow cutting-edge published AI research, and have a good idea about what is currently possible using it. I think it highly unlikely that "weaponised" AI developed in secret is more advanced in agency and planning etc than the results you see in e.g. DARPA challenges or Boston Dynamics etc $\endgroup$ – Neil Slater Jun 14 '18 at 15:21
3
$\begingroup$

Short answer: despite the incredible advances in AI via Machine Learning (and subfields) AI is nowhere near this kind of autonomous decision-making.

I can't prove a negative, but, the level of autonomy Musk is talking about is still on the horizon.

Doesn't mean Musk is wrong about the hypothetical. imo I'm glad he keeps bringing these kinds of issues to the fore.

Engineers and mathematicians have a bit more credibility on this subject than laypeople. If it seems alarmist, there are many in the scientific community who feel it is warranted.

What Musk is describing is an extension of what Asimov felt compelled to warn us about via his "Three Laws of Robotics" back in 1942, when computers sucked. For an explication of this idea re: Machine Learning, see below.


JUST FOR FUN (SORT OF:)

Author and mathematical physicist Hannu Rajaniemi just published a story in the MIT Tech Review on this very subject.

Unchained: A Story of Love, Loss, and Blockchain

Warning: the story is both wickedly funny, surprisingly moving, and very likely prescient!

| improve this answer | |
$\endgroup$

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.