Obviously, if a group of scientists are studying a rabbit they claim can tell the future, funding is going to be an issue. So monetizing this rabbit's unique abilities is a must. 60 seconds isn't a lot of time, unless we are talking about the stock market.
Here is the plan:
- Program a computer to turn on/off LEDs based on stock prices
- Train the rabbit to push buttons based on LEDs in the future
- Program a computer to buy/sell stock based on those button presses.
Step 1 - Materials: Build a Skinner Box. Go get a Raspberry Pi, wire, LEDs, buttons and rent a book on Python from the library (total cost:
$55.70). Hookup 2 buttons and 2 LEDs to the Raspberry Pi and then attach them to the rabbits cage.
Step 2 - Training: Train the rabbit to press button 1,
B1, when light 1,
L1, is on and vise versa for
L2. The tricky part is that we don't want want the rabbit to push
L1 is currently on. We want him to push the button if it will be on in 60 seconds. He needs push a button in the present based on the LEDs in the future. See next section for details.
Step 3 - Programming: Program the Raspberry Pi to cycle through your favorite stocks in a predictable way. Then for the current stock, compare its current value to the value it had 60 seconds ago. If it increased turn on
L1. If it decreased, turn on
L2. Finally, when a button is pressed, look up the stock that will be displayed in 60 seconds. If the pressed button was
B1, buy that stock. If it was
B2, sell that stock.
Training the Rabbit
We need to immediately reward the rabbit for pushing the correct button, but we won't know if it is the correct button for 60 seconds. How does that work?
To train the rabbit, use the same stock market setup only with fake (predictable) data instead of real data. At first, use a 0 second delay and reward the rabbit if he pushes
L1 is lit. Once he has the hang of this, introduce a small delay. If he pushes
L1 will be lit in 1 second, then reward him. Keep increasing the delay, until you get to his max precognition distance of 60 seconds.
The reason this training works is because of the fake data. We know what light will be on at what time, so we know if the rabbit is giving the correct answers. Gradually start introducing real data. The downside is that we don't know if he has given us the correct answer when we use real data. So I would use fake data
90% of the time to make sure the rabbit is staying accurate.
Making money and teaching rabbits cool tricks is fine and dandy but we're scientists. Can we see further into the future? Yes, we can.
There may be a more clever (i.e. better than
O(n)), but this should extend our grasp of the future by a couple minutes. Simply add some buttons
B2b. The teach the rabbit to push
B1a, if he sees himself pushing
B1 (in 60 seconds) and to push
B1b if he sees himself pushing
B1a (in 60 seconds). Once you have done this, you will be able to see 3 minutes into the future.
The reason for this is that when
L1 is turned on, the rabbit will see it 60 seconds before it happens and push
B1. Then the rabbit will see this 60 seconds before it happens and push
B1a. Then the rabbit will see this 60 seconds before it happens and push
B1b. This button press is observed and we know that in 3 minutes,
L1 will turn on.
Since a computer is being used, the rabbit can push buttons at his leisure because an active observer is not necessary. The rabbit can push the buttons as frequently as he wants and the computer will respond accordingly. In this example, we are asking whether to buy or sell a certain stock but in general we can ask the future any binary question. The only requirement is that we ask in a predictable way. For instance, we could ask "Is the president alive?" and the rabbit could give us as much as a 3 minute warning of the president's assassination. All the computer would need is someway to judge whether the president is currently dead or alive.
When making important decisions based of the behavior of a rabbit, there are uncertainties. For example, in order to get a 3 minute heads-up we need the rabbit to see the
L1 and feel like pushing
B1. Then we need the rabbit to see himself pushing
B1 and feel like pushing
B1a. Then we need the rabbit to see himself pushing
B1a and feel like pushing
B1b. The larger the heads-up distance, the less likely it will happen.
When making decisions based of information from the future, there are even more uncertainties. For example, what if the rabbit tells you the president will die in 60 seconds. Then in the midst of precautions taking place, an assassin gets a Ferdinand-style opportunity and takes it. Thus the president died because the rabbit told you 'the president will die'.