Two Black Boxes

Two Black Boxes

Once upon a time, there were two large black boxes, A and B, connected by a long insulated copper wire. On box A there were two buttons, marked α and β, and on box B there were three lights, red, green, and amber. Scientists studying the behavior of the boxes observed that whenever they pushed the α button on box A, the red light flashed briefly on box B, and whenever they pushed the β button on box A, the green light flashed briefly. The amber light never seemed to flash. They performed a few billion trials, under a very wide variety of conditions, and found no exceptions. They concluded that there was a causal regularity, which they conveniently summarized thus:

• all α’s cause reds
• all β’s cause greens

The causation passed through the copper wire somehow, they determined, since severing it turned off all effects in box B, and shielding the two boxes from each other without severing the wire never disrupted the regularity. So they were curious to know just how the causal regularity they had discovered was passed through the wire.

A wiretap on the wire soon revealed that things were more complicated. Whenever either button was pushed on box A, a long stream of pulses and gaps—ons and offs, or bits (10,000 bits, to be exact)—was sent swiftly down the wire to box B. But the pattern of bits was different each time! The scientists decided to open up box B and see what happened to the strings of bits when they arrived. Inside B the scientists found an ordinary digital, serial supercomputer, with a large memory, containing a huge program and a huge database, written in more bit strings. They could trace each step of causation at the microscopic level. However, when the same input string was repeated, box B always gave the same output, but did not go through the same intermediate steps because the memory state changed.

All strings, they hypothesized, are either red strings (cause the red light) or green strings (cause the green light). But they didn't test all possible strings—only strings emitted by box A. When they inserted man-made variations on strings from A, the amber light almost always flashed! It was as if box B detected their intervention. Only man-made versions of red or green strings (by copying them) produced the corresponding color. After exploring billions of random strings, they concluded there were three varieties: red, green, and amber; amber strings vastly outnumbered the other two.

The scientists opened box A and found another supercomputer. It had an internal clock that was used to generate seemingly random bit strings each time a button was pushed. Yet despite this randomness, pressing α always produced a red string and pressing β always produced a green string, with only about one in a billion exceptions.

Then the two AI hackers who built the boxes explained everything. Al built box A, an expert system with a database of true propositions about everything. Bo built box B, a rival expert system. They connected them with a wire. Pressing α instructed A to choose a random true belief (from its axioms or deduced implications), translate it into English ASCII, add random bits to reach 10,000 bits, and send it to B. Pressing β instructed A to send a false sentence. B then translated the input into its own language (Swedish Lisp) and tested it against its database. If B believed it true (since both databases contained roughly the same truths), it flashed red; if false, green. Tampering produced strings that were not well-formed English, triggering amber. Most random strings were not well-formed sentences, hence the preponderance of amber.

Thus the mysterious causal property "red" was actually truth in English, and "green" falsehood. Anyone could compose red strings by writing ASCII for true sentences. However, some philosophers argued that the property wasn't exactly truth, due to errors or limitations in the databases. Others debated whether semantic properties could be causal, or if the system was merely syntactic. The debate continued, but the original mystery was solved.