The question of whether artificial intelligence can be invented is almost 200 years old, dating back to the beginning of computing. Victorian mathematician Ada Lovelace wrote what is generally considered to be the first computer program. As she did, she wondered about the limits of what computers could do.
In 1843 Lovelace wrote, in relation to what could be called the first programmable computer:
The Analytical Engine has no intention of initiating anything. He can do whatever we know how to command him to do. It can follow analysis; but it has no power to presuppose any analytical relations or truths. His province is to help us provide what we already know.
And this assertion has affected the field of AI ever since. As many critics will note, computers only do what we tell them to do.
A century after Lovelace argued against the invention of a machine, Alan Turing, one of the inventors of the electronic computer, returned to the subject. In 1950 Turing wrote what is generally considered to be the first scientific paper about AI. In it, he tried to refute Lovelace’s objection:
Who can be sure that the ‘basic work’ he has done was only the growth of the seed planted there through teaching, or the effect of following known general principles. A better version of the objection says that a machine can never “surprise us”. This statement is a more direct challenge and can be met directly. Machines often surprise me.
This has not changed. Today, we are increasingly amazed by machines. Take OpenAI’s new ChatGPT bot for example. In fact, there is mounting evidence that AI can help people invent – and in some cases could even be considered the inventor itself.
Courts around the world have now begun to be taxed with the question of whether machines can be invented. Stephen Thaler, co-founder of Scentient.ai, has filed patent applications for two inventions naming a neural network as the sole inventor.
These applications have been rejected in almost all jurisdictions, mainly on the legal grounds that the inventor should be a person. But none of the lawsuits so far have tested Thaler’s claim that the computer is indeed the sole inventor.
In an article published on 7 December i Nature’s Machine Intelligence, we examine Thaler’s claim. While we reveal many technical reasons why the computer is not the sole inventor in this case, we also document a long history of AI being used to help humans invent—and in some cases invent itself. Here are just a few examples.
In the 1980s, AI researcher Douglas Lenat’s Eurisko system (eurisko “I discover” is Greek) who invented some novel 3D circuits. A US provisional patent application was even filed for one of these.
Beginning in the 1990s, computer scientist John Koza applied genetic programming to invent a number of novel devices, including some strange radio antennas that looked like folded paperclips. One of these antennas is likely to be the first AI invention in space, flying on NASA’s ST5 spacecraft.
Although it is not a better mousetrap, in 1998 the Oral-B CrossAction Toothbrush was invented by the aforementioned Stephen Thaler in a brainstorming session with a neural network.
Recently, researchers at the Massachusetts Institute of Technology used a deep neural network to identify Halicin – a powerful new antibiotic compound. Halicin is named after HAL, the famous AI computer in Arthur C Clarke’s 2001: A Space Odyssey. Multiple companies with billions of dollars in funding are using AI-based strategies for drug discovery and development.
It seems that the invention of AI is here to stay.
Is AI ‘inventing’?
The abstract idea behind how to compose AI programs is quite simple. You define some concept space, and the program explores this space. The space is usually very large, perhaps even infinite. So much effort needs to be made to identify whether a part of the space is worth exploring further, as well as confirming any promise of a new concept.
For example, the concept space might be all the possible ways to bend a straight antenna. The challenge is to find out which infinite path has the best electromagnetic properties.
We asked the chatbot Jurassic-1, a cousin of ChatGPT, to come up with a patent similar to one of Thaler’s patent applications. Here’s what we found:
PVC, latex or silicone rubber gloves, especially disposable gloves. The invention provides a glove with a flexible adhesive portion made of a fractal pattern. The moving part is created from a continuous fractal pattern. The flexible adhesive part is strong and rigid enough to fulfill its intended function.
To see if this idea was indeed original, or at least not patented, we searched the United States Patent and Trademark Office online database and found no patents with the words “glove” and “fractal”. It is therefore possible that gloves with a flexible fractal grip pattern could be patented.
What is important is that the computer generated this idea independently, without human help or prompts.
Just as AI is transforming other aspects of our lives, it looks set to soon change the way we design. We need to think carefully about how the innovation system adapts to these changes. AI could reduce the time and costs of invention, while also increasing the technical depth of inventions.
Will we need a new type of intellectual property to protect inventions made by AI systems? Or will patent offices be inundated with new patent applications designed with the help of (or with) AI?
Put on your fractal gloves and expect to be surprised!
Toby Walsh is Professor of AI at UNSW, Research Group Leader, UNSW Sydney. Alexandra George yes Associate Professor of Law, UNSW Sydney.
This article was first published on The Conversation.