Natural-born inventors -or- What happens when a computer creates

This week has me thinking again about some of the historical impediments to Black inventors, although in a modern context, that has very little to do with civil rights. Remember that historic rules requiring inventors to state their citizenship were used to prevent Black inventors from filing patents. Thankfully, that roadblock eased over time. But a similar question is now being raised around the world, albeit without the racist undertones: does an inventor even have to be a person?

This might make some intuitive sense, particularly for my generation and older. But there are lingering questions that start to border on philosophy. For example, what if a computer “invented” something only because of the creative way in which it was programmed? Or what about inventions that are made by AI and humans working together? It seems to me like a patent might at least be appropriate for that. But how involved does the human have to be? And can the “black box” nature of machine learning be squared with the legal requirement that a patent teach the public how to make the invention?

So far, the United States Patent and Trademark Office (USPTO) says, “absolutely.” However, the USPTO recently ruled that Federal case law and the using “whoever” in patent statutes requires an inventor to be a “natural person.” This may not be the final word. Other patent offices worldwide have preliminarily found that computer-assisted innovation can be protected. For now, though, U.S. applications for inventions that are created by artificial intelligence will be invalid because they can’t properly name the inventor.

There are potentially substantial commercial problems too. Clearly, even with its current limitations, AI has the potential to provide dramatic improvements across a range of fields. Even with relatively simple genetic algorithms, AI engines can derive new structures that offer vast improvements to systems such as aircraft chassis or engine parts. And AI analysis has shown promise in the quest to identify new drugs, such as antibiotics or cancer-fighting molecules. In this regard, an AI engine can analyze data on millions of existing – or potential – compounds with a speed and a depth that are impossible for humans alone. (There are other successes and broader questions discussed in the video linked below.)

But if companies can’t patent the results, will companies invest heavily in AI to actually generate these improved products? Some argue that they won’t because non-patented products can be too easily copied by competitors. Some argue that patents may just become less critical to commercial success, just as they already have for many commercial software companies.

For example, the falling cost of computing could democratize production so much that branding, service, quality, and other intangibles become the most important factors for a product, rather than some patented structure or function. Many have also proposed that commercial success in the future will depend on learning to use computers and people in combination to take advantage of the unique skills of both. I certainly don’t know, but it does suggest that inventors of the future will need to develop skills that go beyond the traditional technical competencies. What do you think?

Finally, after the post last week, someone passed on the video below. I think that the artist gets peanut butter wrong (it probably wasn’t George Washingon Carver’s), but it was still a catchy tour through quite a few other amazing inventions. Hope you enjoy!