I hope “plush giraffe perturbation” becomes a standard test and every robotics lab now has to go get a plush giraffe
If you ask Visual Chatbot, there are always at least a couple of giraffes.
That's because it learned from questions that humans asked & answered about images, and apparently people rarely asked "how many giraffes are there" when the answer was zero. http://demo-visualdialog.cloudcv.org
Looks like the @OpenAI model handled the giraffe so well because it trained in a variety of universes.
No simulation could be perfect, so their flaws varied to prevent the AI from learning to hack the matrix. As it inevitably would do. https://openai.com/blog/solving-rubiks-cube/ …
"Universe where there is a pesky giraffe" turned out to be not that much more difficult than "universe where friction is weird" or "universe where everything is rubber".
The AI also adapted to "universe where everything is shoddy"
In my book I talk a lot about AI's tendency to exploit the flaws in its simulations.
I am pleased they let me title a chapter "Hacking the Matrix, or AI Finds a Way"
can't argue with science
You can follow @JanelleCShane.
Tip: mention @threader_app on a Twitter thread with the keyword “compile” to get a link to it.
Enjoy Threader? Sign up.
Threader is an independent project created by only two developers. The site gets 500,000+ visits a month and our iOS Twitter client was featured as an App of the Day by Apple. Running this space is expensive and time consuming. If you find Threader useful, please consider supporting us to make it a sustainable project.