Next generation inventions: a computer-human perspective

Technologies which incorporate automation and autonomous artificial intelligence (AI) can perform tasks with human-like ability. As machine learning and computer language often surpass human expertise, certain inventions would be impossible without the support of artificial intelligence.

Many AI-derived inventions receive patent protection. However, despite their vital role in generating inventions and patentable ideas, computers are not regarded as inventors.

In his 2016 paper "I Think, Therefore I Invent: Creative Computers and the Future of Patent Law", Ryan Abbott addressed the uncertainty over rights ownership for computer-generated inventions. Abbott argued that, while AI systems have been making patentable inventions for over 20 years, patent law has failed to keep up.

When computers invent, who is the inventor?
Patent applications must identify one or more named (human) inventors – to date, this requirement has not caused any issues. Although AI systems are regularly used in R&D activity, human input is still necessary for:

• developing inventive concepts;
• designing AI ​​systems;
• identifying problems; and
• evaluating results.

When several people are responsible for developing an invention, courts and IP offices are responsible for determining inventorship. However, if AI systems develop patentable inventions without human involvement, the conventional evaluation procedure (ie, identifying the human inventor behind a patent) is no longer relevant. A new generation of machines which act independently (ie, without human involvement) appear to be redefining inventorship and the inventive act.

Robot scientists
Automation in experimental laboratories is on the rise. So-called ‘robot scientists’ are systems that integrate AI algorithms with physical laboratory robots to perform scientific experiments autonomously. They work with minimal human intervention to provide experiment consumables and remove waste. Existing robot scientists have demonstrated the ability to:

• observe;
• design hypotheses;
• perform experiments;
• test hypotheses;
• use automated laboratory equipment to perform experiments; and
• interpret results.

This technology is an important step towards the full automation of scientific discovery. One robot scientist, Eve, is being used in drug development – for example, it has identified compounds for combating drug-resistant malaria. For 5,000 molecules, Eve determined the properties of the most efficient molecules and considered that only the remaining elements of the set that they predicted would be the most effective. Through this process, it ‘discovered’ a new anti-malarial application for a known cancer-inhibiting drug.

Computer-generated claims
Questions remain over the inventorship of patents generated using AI systems, and despite the lack of human involvement, these patents must still be protected by patent law. While companies such as Cloem apply natural language processing technologies to assist applicants of computer-generated claims, no existing patent offices support laws which claim computers as inventors. The increased involvement of computers in generating and implementing new ideas is likely to transform inventorship as we know it.