Artificial intelligence, not so smart
Compiled by Karen Smith
Most of you have heard of Artificial Intelligence, commonly called AI. It has uses in many diverse fields, from illustrating comics to identifying tumours pictured in medical images. I have used an AI-powered app to identify plants in the field.
AI companies claim these computer applications ‘learn’ by extracting patterns, linguistic structures and contextual insights from huge amounts of source material. The murky side to this is where the source material originates and how much is protected by copyright. It has been widely reported that AI companies in the USA, such as Facebook’s parent company Meta, Google and Open AI, are being sued in various courts by authors, musicians and illustrators, for example, demanding compensation for unauthorised use of their intellectual property.
Meta reportedly downloaded massive amounts of pirated material from ‘shadow libraries’ to train its large language model AI. Open AI and Google justify this practice as covered under the ‘fair use’ provisions of the Copyright Act and are lobbying the US government to legalise it in the name of national security. They argue that limiting AI training on copyright material could weaken America’s technological edge and slow down innovation.
Earlier this year, a Chinese company called Deep Seek released a new AI model called R1. This AI is offered at a fraction of the cost of the US models and caused a few ripples on the US stock exchange. This development was remarkable because the US has an embargo preventing China from obtaining high-level computer processors, so the Chinese AI uses less powerful but more economical components.
US company Open AI is now claiming the Chinese AI has pilfered data from the US company. Open AI claims the Deep Seek AI used other AI models as a teacher. Deep Seek queried Open AI at a massive scale until it became as ‘smart’ as the teacher. This method of training only needs a fraction of the resources other models use while using much less power.
Of course, US company Open AI is complaining loudly. Its terms of use explicitly state that nobody may use its AI models to develop competing products. However, their complaints are not attracting much sympathy.
Anyone with a stake in intellectual property will await the next chapter.
Artificial intelligence, I’ll take real intelligence every time.
References:
https://www.theregister.com/2025/03/11/meta_dmca_copyright_removal_case