AI Challenges


Paul Munnis

Elon Musk is leading an anti-AI drive contending it will be the source of WWIII as nations feud over possssion of the technology. Here are a few thoughts on the matter:

Keeping AI code private, confidential, and exclusive is fraught with challenges.

The first challenge is that if what is made available is machine language then it is easily reversed engineered. If it’s embedded inside of product code it is easy to isolate and reverse engineer. If it’s in source form then it is as good as in the public domain. If it’s encrypted it must be converted to object code in order to execute it. In other words if someone is determined to grab your AI code they can do so. The best solution then is to keep the code for in-house use and charge for it’s use as a service. Even then, it’s hard to keep employees and others from nabbing a copy off for black-market sale.

The second challenge is that if the AI code is being kept in-house then the customer data it operates against must be supplied for use with the AI code. That sounds simple but we live in an era of big-data where terabytes of code must be kept private and confidential.

The third challenge is that the customer data itself is valuable and of great value to a competitor whether that competitor be commercial or private or a foreign nation.

To fix this a Cloud Data Pipe is premised but as the data flows it can be intercepted and turned.

The fourth challenge is keeping the result of the AI run output confidential so it can be acted upon by whoever paid for the run.

The sixth challenge lies in the nature of some data that is highly dynamic in origin making it progressively more obsolete as time goes by.

So development of AI code is one thing but using it is something else.