“We would benefit from a new agency,” Smith said in a speech in Washington. “That is how we will ensure that [Microsoft] remains in control of technology.”
> “New laws would require operators of these systems to build safety brakes into high-risk AI systems by design,” Smith said in a blog post accompanying his speech. “The government would then ensure that operators test high-risk systems regularly to ensure that the system safety measures are effective.”
Hmm... So, we will build a Turing machine (TM1) that will simulate this other Turing machine (TM2) and TM1 will tell us if TM2 will maliciously halt bringing down the world's infrastructure. Wait.... that requires the halting problem to be decidable (which it isn't).
Let us substitute regular testing in lieu of formal verification. TM1 will snapshot TM2 periodically. If TM1 doesn't find a suspicious state, it will conclude TM2 is running normally.
2 comments
[ 4.0 ms ] story [ 17.6 ms ] threadHmm... So, we will build a Turing machine (TM1) that will simulate this other Turing machine (TM2) and TM1 will tell us if TM2 will maliciously halt bringing down the world's infrastructure. Wait.... that requires the halting problem to be decidable (which it isn't).
Let us substitute regular testing in lieu of formal verification. TM1 will snapshot TM2 periodically. If TM1 doesn't find a suspicious state, it will conclude TM2 is running normally.
That's just great.
/s