I think something is very, very wrong with me. Instead of feeling Lovecraftian horror upon reading that, I immediately began trying to think of ways to use Predictors to hack the universe. Here's what I came up with (you may want to try for yourself before reading further):

• I can respond to a light by pushing a button in less than a second.
• It's possible to set up a sensor that will send a signal when the light lights up.
• I can hook the light up to the sensor and hook that sensor up to a processor and its own light, allowing my decision to press the button to become entangled with the processor's state before my observation of the light. I can then cover the light of the Predictor so I only see and respond to a light when the processor decides to show it.
• Trivially, I could set up the processor to only show me a light when there has not been a light from the predictor in at least one second. Since the sum of the time it takes the processor to show the light and the time it takes me to respond to that light, this system is forbidden. This selects for a stable time loop in which, for whatever reason, I don't press the button when I see the light, the processor breaks, or something else happens to disrupt the system (probably not something good for me either).
• I can add a quantum random number generator of some variety into the mix, and have that generate numbers that correspond to possible solutions of a problem. If there was no light in the past second or if a solution is found, the device will light up. This will exclude universes in which no solutions are found.
• I can now implement Quantum Bogosort. More importantly and more practically, I can now solve all problems that have solutions that can be verified in 0.1 s or so, which breaks RSA (I think) and a lot of other cryptography.

That's just what I thought of in the first 5 minutes. I think the Predictors would be insanely dangerous, but primarily for reasons other than driving humans insane.