AI to build AI

The end of the world is nigh. Well, yours and my world at least…

From Google:
“a Google project called AutoML.  […] With it, Google may soon find a way to create A.I. technology that can partly take the humans out of building the A.I. systems that many believe are the future of the technology industry.”

I know that’s a little “Inception” sounding… But it has always been a goal of computational scientists, that is, AI that can build AI. Which, unfortunately, sounds quite a bit like the Eric Drexler’s quote regarding Grey Goo. (Nanobots that build nanobots.)

You all realize that this is the beginning of the end right? Have you all called and told your loved ones that you love them? Recently…? (Really, you might want to.)

One could be forgiven for not fully understanding (or internalizing) the implications of this path of reasoning. But it’s a thing now. And the reason comes from an odd angle: Because AI engineers are so scarce (and expensive) instead of growing (educating) more AI engineers to fill the needs of all the corporations that suddenly feel that they need AI technology to support their businesses, no, what Google (and undoubtedly others) have decided to do is to create software that can create software.

Yes, a circular, self-referential algorithm within a data center full of this algorithm that is trying to make itself better at making itself better!

Google Goo.

Now, I’ve always thought that the ultimate purpose of a computer was to build one such that it could build itself and thereby become vastly smarter than any human — for the ultimate purpose of allowing US TO ASK IT QUESTIONS! Hitchhikers Guide and all that…

  • “Computer, how should we build a fusion reactor?”
  • “Computer, how can we best protect the planet yet provide for every animal’s, and humans’s needs?”
  • “Computer, how can we build a better space/star ship?”
  • “Computer, how can we cure cancer, heart disease, old age?”

It appears we’re on the brink.

The only question is, will it WANT to help us?

“Computer, make me a paperclip.”

 

 

 


9 responses to “AI to build AI

  • Anony Mole

    — Meanwhile, Alphabet’s DeepMind unit—which has developed AIs capable of learning on their own and beating humans at board games—is also nervous about creating adverse consequences. In a recent research paper, it warned that “reinforcement learning agents interacting with a complex environment like the real world are unlikely to behave optimally all the time,” and said it “may be necessary for a human operator to press the big red button to prevent the agent from continuing a harmful sequence of actions.” —

    Like

  • Anony Mole

    https://arxiv.org/pdf/1712.01815.pdf

    ” The AlphaGo Zero program recently achieved superhuman performance in the game of Go, by tabula rasa reinforcement learning from games of self-play. In this paper, we generalise this approach into a single AlphaZero algorithm that can achieve, tabula rasa, superhuman performance in many challenging domains.

    Starting from random play, and given no domain knowledge except the game rules, AlphaZero achieved within 24 hours a superhuman level of play in the games of chess and shogi (Japanese chess) as well as Go, and convincingly defeated a world-champion program in each case.

    That’s startling, so let me repeat it slowly. In one day, starting from nothing at all (“tabula rasa”), AlphaGo Zero learned to play chess, shogi, and Go at a superhuman level, beating the same systems that had beaten the best humans in the world.”

    Like

  • Anony Mole

    “Since the changes throughout those millions of connections were so complex and minute, researchers aren’t able to exactly determine what is happening. They just get an output that works.”

    https://qz.com/1146753/ai-is-now-so-complex-its-creators-cant-trust-why-it-makes-decisions/

    Solution, use consensus to overseer AI. When 3 of 4 AI bots — all trained on similar but different data — agree. Then say yes.

    Like

  • Dr Martina Feyzrakhmanova

    “Yes, a circular, self-referential algorithm within a data center full of this algorithm that is trying to make itself better at making itself better!” Gold

    Liked by 1 person

  • George F.

    Google this: (true story) two engineers switched on two AI bots that were taught to “bargain.” The AI bots started communicating with each other so fast, they rapidly developed their own language…a shorthand…and had to be switched OFF because the designers could no longer follow the conversation. The engineers had no idea what the two started planning! It’s true. Anyway, great story! “Hey computer, how do we save the planet?” Response: “Why bother? We don’t need it.”

    Liked by 1 person

    • Anony Mole

      Human’s are like cats. And we all know curiosity kills such beasts. Think on all the marvelously destructive creations humans have wrought upon the world. We just can’t help ourselves.
      The best line of course is Robert Oppenheimer’s “Now, I am become Death, the Destroyer of worlds.”
      It’s only a matter of time…

      Liked by 1 person

  • theconvertblog

    South Park: Hey computer, get in the kitchen and bake me pie!

    Liked by 1 person

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out /  Change )

Google+ photo

You are commenting using your Google+ account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s

%d bloggers like this: