A Version strategy for AI

A Version strategy for AI

Usually as I have seen, software’s unique snapshot copy is given a mathematical version number going forward,

For example,

1) first version may be 0.1 ( I don’t know if someone would start with a 0 :) ).
2) next version might be 0.1.1 or 0.2 or 1.0 depending on the value and wealth of changes.
3) Usually people commit changes and bug fixes and increment the software version ( by semvar standard, people call them as major version, minor version and patch versions ). Many software systems are automated to version this way.

For Artificial intelligence though,
a better version scheme would be to start with ‘1’.
For each valuable change or bug fixes advancing the AI code,
versioning can be in decimal and proceeded backward.

For example,

1) first version might be 1.
2) second version might be 0.999999 ( infinite 9’s ) ( depending on the trust on the AI code :) ).
3) third version might be 0.999999….( infinite 9’s )...9999998
..
..
4) If we have achieved an immense progress,belief and trust that we have solved half of the problem, they might version their software ‘0.5’. This must really be celebrated :).
5) If we have succeeded in building the AI completely and we have achieved intelligence on the task in hand, there is no more task, no more problem and no more solution, this version might be called ‘0’.