Artificial intelligence may speed up the process of invention while cutting expenses and increasing the technical breadth of discoveries. Can machines even invent new things without human help? Yes, they already have.
Since the very inception of computing, there has been a debate about whether artificial intelligence is capable of invention. Ada Lovelace, a Victorian mathematician, is credited with creating what is usually regarded as the first computer program. She pondered the capabilities of computers as she went along.
Since then, the AI industry has been plagued by this claim. Computers just carry out our instructions, as many detractors will point out.
One of the pioneers of the electronic computer, Alan Turing, revisited the question a century after Ada Lovelace argued against the advent of machine invention. What is usually regarded as the first scientific publication (read below) on AI was written by Turing in 1950. He attempted to disprove Lovelace’s objection in it:
Subscribe to GreatGameIndia
This remains the same. These days, technology constantly surprises us. Consider the new ChatGPT chatbot from OpenAI as an illustration. There is growing evidence that AI can assist people in inventing, and in certain instances, it may even be the actual inventor.
The issue of whether machines can invent is currently being discussed in tax courts all around the world. Two ideas in which a neural network is credited as the sole creator have been submitted for patent protection by Stephen Thaler, a co-founder of Scentient.ai.
Nearly all jurisdictions have rejected these applications, typically on the basis that an inventor must be a human. However, none of the court cases to far have put Thaler’s assertion that the computer is the lone inventor to the test.
The 1980s saw the invention of several cutting-edge 3D circuits by AI researcher Douglas Lenat’s Eurisko system (eurisko is Greek for “I discover”). Regarding one of these, a provisional US patent application was even submitted.
John Koza, a computer scientist, developed various unique devices beginning in the 1990s using genetic programming, including some quite peculiar radio antennae that resembled bent paperclips. Since one of these aerials was used by NASA’s ST5 spacecraft, it is most likely the first artificial intelligence invention in orbit.
More recently, a potent new antibiotic molecule called Halicin was identified by Massachusetts Institute of Technology scientists using a deep neural network. The legendary AI computer HAL from Arthur C. Clarke’s 2001: A Space Odyssey inspired the name Halicin. AI-based approaches are being used by numerous businesses with financing in the billions of dollars for medication development.
AI innovation seems to be here to stay.
Is AI ‘inventing’?
The abstract concept underlying how AI programs can invent is straightforward. You describe a conceptual space, and the program explores it. The space is usually exceedingly large, possibly limitless. As a result, much work must be expended in determining whether a portion of the area is worth further investigation, as well as in confirming any promise of a novel notion.
For example, the universe of conceptions could be all the different ways to bend a straight aerial. The problem is to identify which of the infinite number of ways has the best electromagnetic characteristics.
It is plausible that AI will soon change how we invent, just as it is already changing other facets of our lives. We must carefully consider how the innovation system will adjust to these developments. AI may speed up the process of invention, cut costs, and improve the technical sophistication of new discoveries.
Will the inventions produced by AI systems require a new type of intellectual property to be protected? Or will patent offices be overrun with new patent requests for inventions made possible by (or with the help of) AI?
According to a retired professor of computer science at Oregon State University today’s large neural network artificial intelligence are already slightly conscious.
Read the document below: