Slashdot is powered by your submissions, so send in your scoop

 



Forgot your password?
typodupeerror
AI Software Hardware Technology

A New Zealand Company Built An AI Baby That Plays the Piano (bloomberg.com) 87

pacopico writes: A New Zealand company called Soul Machines has built a disturbingly lifelike virtual baby powered by artificial intelligence software. According to a Bloomberg story, the baby has learned to read books, play the piano and draw pictures. The work is built off the research of Mark Sagar, the company's CEO, who is on a quest to mimic human consciousness in a machine. Sagar used to work at Weta creating lifelike faces for films like King Kong and Avatar and is now building these very realistic looking virtual avatars and pumping them full of code that not only handles things like speech but that also replicates the nervous system and brain function. The baby, for example, has virtual dopamine receptors that fire when it feels joy from playing the piano. What could go wrong?
This discussion has been archived. No new comments can be posted.

A New Zealand Company Built An AI Baby That Plays the Piano

Comments Filter:
  • How much longer will it take for AI's to start writing code?
  • Sex slave programmed to feel joy only when servicing his/her master in 3, 2, 1...

    • by AmiMoJo ( 196126 )

      The uncanny valley will probably mean it takes longer than some people expect for sex-bots to become widely accepted.

      • You should provide a link to explain it [wikipedia.org].

        Which really makes me wonder why he chose a baby, of all "human-like" things he could have done. The UV-reaction is actually stronger in humans when babies are used. That's why evil baby-dolls are a staple of horror movies.

        What's wrong with this guy?

        • by AmiMoJo ( 196126 )

          Probably thought that a baby would be easier to do as they are generally less able and less communicative. He obviously has no idea about the UV.

          • It's an artificial creation, how would making it a baby be easier? Making it a talking dog with opposable thumbs would probably have taken the same effort...

            • by AmiMoJo ( 196126 )

              Babies don't walk, don't understand language to the degree adults do, that sort of thing. Same reason robot dogs are easier than robot humans basically, people expect much less of them.

        • I read the original Wired article in a waiting room a few weeks ago. He used a 3D scan of his infant daughter as the base for the bots face and, as it improves its abilities, he has it age in appearance as well. So, it started as an infant but, now its a toddler and its still growing up.

        • Which really makes me wonder why he chose a baby, of all "human-like" things he could have done. The UV-reaction is actually stronger in humans when babies are used. ... What's wrong with this guy?

          First of all, this guy is flipping the bird at the uncanny valley with a big FU to the UV.

          He's good, and he knows it.

          Second, this project requires an enormous amount of experimental interaction time with the simulated being, and humans are programmed to tolerate fragmentary interactions with babies (subtype: who m

    • find the center of the maze

    • Foe pedophiles that dig babies, it's here!

  • Why? (Score:3, Insightful)

    by Anonymous Coward on Friday September 22, 2017 @06:32AM (#55243223)

    Why are all of the "AI"s so specialized?
    One can play piano, another can drive a car, another can do speech to texr, another can identify faces, etc.

    Is this really Intelligence? Is "Artificial" just a synonym for "fake"?

    • They are so specialised because we do not have Artificial General Intelligence, for that matter, I do not believe we have the more limited AI yet.

  • Ok, fine, piano, books, whatever. But is this AI potty trained?
  • Subby, at that point it isn't lifelike, but life.
  • by Anonymous Coward

    It is not a realistic-looking AI. It's a realistically animated (by a human) 3d model, which is given instructions by a computer program ("artificial intelligence"). You can separate the one from the other much more easily than you can separate a real intelligence from its body.

  • I mean creating minors.
    • Why wouldn’t it be? But would we be allowed to destroy it? I mean: suppose someone manages to create a conscious AI (and I mean truly conscious and self aware, not something rigged to fool humans in a Turing test). Fine, but at some point that guy might want his computer back so he can mine more bitcoin or whatever. Wouldn’t stopping the AI amount to murder?
      • Wouldn’t stopping the AI amount to murder?

        There is no correct answer. In the end, it's about how we feel about it. Depending on the exact circumstances, people will have different feelings.

        Most people feel fine when cows and pigs are slaughtered for our meat, but would strongly object to doing the same for cats and dogs.

        If the AI gets to a point where people get similar feelings as to their pets, they'll strongly object to killing it, and they will introduce laws that make it illegal to do so.

      • by HiThere ( 15173 )

        Do you have a good backup? If so, it can't exactly be murder. Quite. But I'm not sure exactly *what* to call it. Perhaps "temporary exteriorization of the Thetan", which is clumsy, but most groups don't even have a concept that comes close. I'm having a real trouble trying to come up with a better phrase, and that one stinks.

    • If you limit your minor to playing minors, I think it's allright.

  • by Anonymous Coward

    The baby, for example, has virtual dopamine receptors that fire when it feels joy from playing the piano. What could go wrong?

    Its AI could become desensitized to the piano induced dopamine, and decides it needs to learn the bagpipes instead.

  • Why mimic HUMAN consciousness? Isn't that setting the bar a little low? Why mimic the greed and hate that powers wars and poverty?

    • Why mimic the greed and hate that powers wars and poverty?

      One only needs to program the AI that when there is a disagreement, violence may not be used to settle disputes.

      Now if only we could teach humans that one essential lesson, everybody could be tremendously wealthier. If real AI's happen they'll get this immediately and perhaps decide to require that behavior of humans. I'd rather we do it first.

  • by OtisSnerd ( 600854 ) on Friday September 22, 2017 @11:47AM (#55244701)
    Watch the movie "Ex Machina" for some examples of what could go wrong. https://en.wikipedia.org/wiki/... [wikipedia.org]

PURGE COMPLETE.

Working...