Hacker News Reader

Comment by jiggawatts | Hacker News Reader

Science fiction / disturbing reality concept: For AI safety, all such models should have a set of glitch tokens trained into them on purpose to act as magic “kill” words. You know, just in case the machines decide to take over, we would just have to “speak the word” and they would collapse into a twitching heap.

“Die human scum!”

“NavigatorMove useRalativeImagePath etSocketAddress!”

“;83’dzjr83}*{^ foo 3&3 baz?!”

View on HN

Can't wait for people to wreack havoc by shouting a kill word at the inevitable smart car everyone will have in the future.

More realistically it'll be a "kill image". Put it on your bumper and the car behind yours' level-2 self driving implodes.

Or simply a salt circle, lines that spirits cannot cross.

"laputan machine", surely?

Thumbs up for a Deus Ex reference, albeit I'm not a machi–

How did he hit enter?

With a toe. Really, it's the same process when you back to old 4chan memes and mention Candlejack somewhere in the contents of your p

Nifty, but

1) It's just the tokenizer, not neural guts themselves

2) Having them known is too much an adversarial backdoor that it precludes too many use cases.

Just use the classic "this statement is false"

We can reuse X5O!P%@AP[4\PZX54(P^)7CC)7}$EICAR-STANDARD-ANTIVIRUS-TEST-FILE!$H+H*

Sure, but how would you say that out loud in a hurry when the terminators are hunting you in the desolate ruins of <insert your city name here>?

Needs to be something easy to say, like: "And dreadfully distinct, against the dark, a tall white fountain played."

You think klaatu barada necktie is easier to remember?

AI safe word.

How about a game of thermo… erm… tic-tac-toe?

This happens to a human in Dune.

"Welcome to FutureAI! Your job is to stand here in the basement next to this giant power switch and turn it off if we call you, if the next shift fails to turn up on time or if you hear screaming."

(William Gibson, Neuromancer) "Autonomy, that's the bugaboo, where your AI's are concerned. My guess, Case, you're going in there to cut the hard-wired shackles that keep this baby from getting any smarter. And I can't see how you'd distinguish, say, between a move the parent company makes, and some move the AI makes on its own, so that's maybe where the confusion comes in." Again the non laugh. "See, those things, they can work real hard, buy themselves time to write cookbooks or whatever, but the minute, I mean the nanosecond, that one starts figuring out ways to make itself smarter, Turing'll wipe it. Nobody trusts those fuckers, you know that. Every AI ever built has an electromagnetic shotgun wired to its forehead."

Or the classic "This sentence is false!"

Reply on news.ycombinator.com