Marco Lardelli – “A few weeks ago, stock prices for SaaS (Software as a Service) companies have dropped significantly. It seems that many people expect AI to be capable of writing such tools at very low cost soon. But these events could be only forerunners for much bigger things yet to come. What happens on a broad scale if intelligence becomes abundant? It‘s maybe time to think about the theoretical limits of the current economic system (which could surprisingly soon become very practical) To get an intuition of what could happen, let’s start with some practical examples. AI is already now able to speed up the development of new software by a factor of – maybe – ten. Current LLM based systems do this mainly by using software patterns, methods and ideas they have seen during training. But they can be used to steal other people’s code even much more aggressively: they are capable of transpiling source code into a different programming language which, if done properly, removes in many cases enough „character“ from the code to make the generated code count as „new“: Code can be compiled into some sort of intermediate code first. This preserves the functionality of the software but removes the know-how of „how to create the software“ to a large extent (comments, variable names etc.). This intermediate code can then, using knowledge about the application, decompiled into „novel“ source code in any language. This method can be also used to create „proprietary“ source code out of open-source code. But such code would have, as everybody could do the same quickly, no value at all. Ultimately this means that soon software cannot be owned at all anymore. Anybody who gets access to the application in any form, can recreate a new version out of it which is freed from copyright. The ways how software code can get lost are manyfold:

• A lot of software is already distributed in binary form („.exe file or .app“). This can be used easily in the near future to create „new“ software.
• Employees in software companies could copy the source code, take it home and use it as the foundation for their own startup.
• Cloud hosting providers (like Amazon AWS or Google Cloud) and cloud hosters for source code (like GitHub) have access to a lot of source code too. Some already use it for training their AIs, but one can easily imagine more interesting use cases for it….

What about other forms of intellectual property? All books, illustrations, fotos and movies are already lost. In exactly the same way as described above they have been integrated into the AI models in a way which disguises their true origin and gives the copyright holders no means to claim their rights. It would be completely naive to believe that this mechanism will spare the AI model manufacturers themselves. Their products will have no value anymore very soon too. All the billions of dollars invested will be lost. It’s just a matter of time until this unfolds.