AI and the rise of China: this is the New New World

Economics, Politics, Philosophy, History and Culture
Forum rules
Always add something of value to the discussion and have fun. Mind your language, no porn, no libel, no trolling and no personal attacks.

Please note, views expressed on the forum do not necessarily represent the views of Mises UK. the Mises UK Council, Mises UK Faculty or Mises UK members. Membership of this forum is open to anyone world wide who is interested in civil discussion.
User avatar
Clayton
Posts: 144
Joined: Sat Mar 31, 2018 4:24 pm

Re: AI and the rise of China: this is the New New World

Post by Clayton » Sat Jun 02, 2018 4:25 pm

It struck me today that Deep Learning spells doom for proprietary desktop software. Why? Well, suppose I am the proud owner of a software company, Minisoft. We make operating systems, desktop software and a wide host of other software tools and applications that are very popular and widely used. However, we use DRM technologies to make it difficult for unauthorized users to use our software. We're really smart people - computer scientists, mathematicians, business analysts, etc. - so we have no illusions about DRM... we understand that it's more of a bank-teller rope than an impenetrable wall. Ultimately, the claws and fangs that enforce our multi-billion dollar licensing model are sheathed in our legal department.

It is human nature to break rules that are easy to break and for which punishment is infrequent and improbable. This is why the DRM+legal team recipe works -- it makes the rules difficult enough to break that it would require an organized/orchestrated effort to do it, which makes the probability of punishment much higher. But in order to work, the recipe requires both sides of the equation -- DRM and legal enforcement -- to work as expected.

Deep Learning and virtualization are going to break PC desktop DRM once and for all by making the cost of going around DRM controls extremely low (basically, the cost of training a purpose-built Deep Learning net + the cost of one valid key, potentially). DL nets enable you to wiggle an input on one end of a black box, learn how that changes the outputs on the other end of the box, and then build a model of this. You can then use this model to ask reverse questions -- in order to get a desired output, what input must we present? "Yeah, but you're talking about cracking codes, and we already know how hard that is." Not quite. If you're running, say, a DRM-protected operating system in a virtual container with the key, you can repeatedly run the startup code with or without the key, train the neural net on what changes in the behavior of the code. Then, you can use this to automatically construct a patch that can be applied to the binary image of the DRM-protected OS itself -- once the patch has been applied, the OS will exhibit the desired behavior (booting up) and will cease exhibiting the undesired behavior (challenging for a key). This can be generalized to other, more roundabout forms of DRM protection, such as selectively disabling features, timing out (trial periods), or "phoning home" to tattle on the license violation. Of course, someone will have to develop cracking software to do this but it is obvious that crackers are highly motivated, so there can be no doubt that they will do whatever is economically feasible.

As head of Minisoft, I have seen the light and realized that software is -- in the ultimate sense -- nothing but patterns and, as such, is properly within the domain of mathematics per se. We all understand that math is inherently free and open-source. In light of this realization, I have decided to make our software free and open-source and to compete on the basis of customer satisfaction with our software products, rather than trying to juice up our profitability through the use of DRM tricks and legal threats.

Post Reply