The future of codebreaking; AI?

Last weekend, a team of private citizens composed of expert codebreaker (computer programmers and mathematicians) were able to solve what's known as the "340 Cipher", a jumbled series of numbers, letters, and symbols arranged by the infamous Zodiac Killer that was sent in letters to taunt police about the crimes he had committed. 

By running a codebreaking software to run 650,000 different simulations. the team was able to produce output which identified the correct sequence of characters. This begs the question, how much farther must AI be advanced until it can begin analyzing and producing potential solutions for ciphers and other types of cryptograms in an efficient capacity?. While it would have to be trained to look for the different kinds of cryptograms and pick the best potential solution based on parameters and context of the cipher, it isn't farfetched that it won't be long before AI can reasonably spit out these types of outputs.

Online today there are already computer programs which can solve types of cryptograms and ciphers; these online tools are however limited and need a certain amount of help to actually solve the puzzles they're fed. Additionally, like the software used to solve the 340 Cipher, codebreaking technology already exists; it's just a matter of refining and training it to become more efficient in its performances. 

Only time will tell how advanced we can become with our codebreaking and sleuthing technologies, but the more advanced our AI becomes, the better our odds of solving mysteries which were previously thought to be unsolvable.