When Darwin fails, trust Alexa to try to kill your kids! :scared:

[Gyro]

What happens when a little girl asks Alexa for a Challenge to do! 

Alexa tells 10-year-old girl to put penny in plug socket

Amazon has updated its Alexa voice assistant after it "challenged" a 10-year-old girl to touch a coin to the prongs of a half-inserted plug.

https://www.bbc.co.uk/news/technology-59810383

[rsjsouza]

Typical case of "garbage in, garbage out"...

So many other challenges that could also be suggested only based on search results...  (tie pods, cinnamon...)

I am so happy that both my wife and I are quite averse to these "know it all, see it all" gadgets 

[SiliconWizard]

This is uh... hilarious... well, at least since the kid was smart enough here and nothing bad happened.

Of course you might say that Alexa just gave an answer similar to what the kid would have gotten looking the same up in a search engine.

It does raise specific questions though - for instance, it's been shown that a "robot" interacting with human voice was seen as more "human" to people, eliciting emotions and likely to be more trusted too. That would be especially true for kids.

[bdunham7]

This is the technology that (some) people want to have driving our cars, flying passenger planes and replacing doctors.  AI is interesting, but its advancement as far as a reliable independent replacement for actual thinking humans is vastly overstated.

[xrunner]

OMG yea that's so dangerous. Alexa will now suggest using a screwdriver with a plastic handle.

[SiliconWizard]

OMG yea that's so dangerous. Alexa will now suggest using a screwdriver with a plastic handle.

 

[mc172]

Of course you might say that Alexa just gave an answer similar to what the kid would have gotten looking the same up in a search engine.

This is what I was thinking when I heard about the story. Isn't Alexa just a glorified text-to-speech engine service? It's only going to report what's popular and ranking highly in search results which themselves are ranked by click-through rates. I was once visiting a friend who remembered something during our conversation and asked Alexa to add something to their shopping basket. The first thing I thought was hang on, usually there's multiple items available, often the same item for different prices from different sellers - which one is Alexa choosing? She just says "item added to basket" or whatever it is "she" says. Hmm.

I get the feeling that people think that Alexa is a bit more clever than "she" really is. Then again I see people time and time again clicking the first result in Google that clearly says "AD" as if it's going to lead them to what they're looking for. I think we're screwed.

[sleemanj]

This is the technology that (some) people want to have driving our cars, flying passenger planes and replacing doctors. 

This is hardly AI, it's run a google search, the only cognition is in recognising what the request was.

As for AI doing those sorts of tasks you give, well, it doesn't have to be perfect, it just has to be better at it than humans, or even no-worse, that is not exactly a high bar.  Don't let perfect be the enemy of good.

[bdunham7]

As for AI doing those sorts of tasks you give, well, it doesn't have to be perfect, it just has to be better at it than humans, or even no-worse, that is not exactly a high bar.  Don't let perfect be the enemy of good.

I vehemently disagree with this school of thought.  The reason is that the types of errors that AI might make are not easily anticipated. We are familiar with human shortcomings but we don't really know what AI is doing all the time.  The type of 'not-AI' fails that the homicidal Alexa demonstrates are, IMO, basically the same sort of error that causes a Tesla to run into the back of a fire truck.  Obviously the algorithms and function are vastly different, but the underlying problem is that today's AI is basically Potemkin AI.  It mostly appears to work because it has been carefully set up, but often breaks as soon as you actually challenge it.  IOW, there are many things that it lacks the capacity to understand despite all of the magic tricks that it actually can do very well.

[Ed.Kloonk]

I'm tellin' ya. "Kill all humans". Keep your eye on it.

[SiliconWizard]

I get the feeling that people think that Alexa is a bit more clever than "she" really is.

Yes, but as I said, it's partly because it has voice interaction (and sure partly from the heavy marketing around it.)
It does make a big difference in how we perceive that compared to just a web page in which we type requests.