In this opinion piece for Raconteur’s Intellectual Property report, we consider how inventors will be impacted as AI develops.A lot has been said about how artificial intelligence (AI) might revolutionise the world of intellectual property (IP), perhaps replacing humans as inventors or taking over existing IP systems. Let’s cut through the hype and consider what we can realistically expect.The term AI is often used loosely. I use it to mean machine-learning, whether guided or loosely constrained, to detect patterns or produce inferences or outputs based on what the machine has “studied” rather than its original programmer. Machine-learning usually needs lots of data to learn and the line with data analytics is often blurred.Starting with the basics, there are undoubtedly advances in the way machine-learning operates or can be computationally implemented efficiently and these advances may be patentable, just as for other inventions.However, many commercial applications of AI involve taking generally known AI techniques and applying them to a data-crunching problem and this alone is unlikely to be considered inventive.Nonetheless there may be protectable IP in the detail of how this is done effectively in a given case. An expert can advise on whether there is likely to be commercially worthwhile protection to seek in a particular application or if simply keeping the data is the key.More colourful debate has involved whether a machine can itself be an inventor or an author, or speculation about one AI filing its own patent applications and another “official” AI examining them. I participated in a public debate with the UK Intellectual Property Office and AI evangelists on the practical, legal and moral implications of this.Patent applications have been filed for an invention naming an AI as inventor, with a notion that this was deliberately done to test boundaries. The UK and European patent offices have both ruled that an AI cannot be an inventor. Academic debate may continue on such questions as how do you determine the term of copyright which depends on the life of an author, if the author is a machine. But for now, at least for businesses, the issues are thankfully clear.The real impact of AI will take place behind the scenes; companies will use AI in design and competitor analysis, but let the human directing the AI take the credit. There is a close parallel with the issues when a semi-autonomous vehicle has a collision: the driver is responsible. To a pilot these issues are nothing new; the captain is ultimately in command and responsible whether or not he or she chooses to rely on autopilot or other systems to assist in navigation or control.In one sense this is just normal use of technology, in the same way computer-aided design and computer-aided manufacturing simplified getting from concept to product or word processors and spreadsheets and databases assisted document production and accounting and filing.A new issue is that AI may make it easier for what I term “artificial inventing” based on analysing apparent gaps in the prior art; it is often more productive to task smart humans to make positive inventions whereas an AI can work 24/7 just looking for gaps. There are also so-called AI tools for searching and assisting with preparing patent applications which I have seen; AI will creep into the field of analysing, selecting, examining and even writing patent applications from all directions.Some balance to this is that machine-learning works well with a training dataset to spot patterns in “what is”; good examples being image processing or identifying anomalous behaviour. However, inventions must be unique and it is less straightforward for AI to deal helpfully with an open-ended “what isn’t”.We should embrace AI tools where they can help, but don’t expect them to replace expert strategic human insight or fundamentally change IP in the next few years.This article was first published in Raconteur ‘s Intellectual Property report in January 2020 (page 7).