After a series of false starts and unsatisfied potential, artificial intelligence (AI) has proven itself valuable across a variety of domains and problems over the last 15 years. While the list of domains and problems is long (spam detection, web searches, speech-to-text, machine translation, just to name a few), the majority fall into a broad bucket of “function optimization.” In these types of problems, there is a correct answer and a lot of data, and the AI’s job is to learn how to answer correctly for new data, based on all the examples it’s seen and learned from.
While solving these types of problems will always be a natural fit for AI, it is now being used to solve a new class of problems, where there is no “right answer.” In these types of problems, the objective isn’t to get something right, it’s to create something new. These are fundamentally creative problems, and to meet these challenges, AI is becoming creative.
Philosophers and theorists can argue about whether AI that solves these types of problems truly possesses “creativity,” or whether that requires some additional human-like “spark.” Answering that question may be an interesting thought exercise, but the fact remains: Artificial intelligence is increasingly finding success in creating wholly new things that are intended for a human audience.
- The famous auction house, Christie’s, just became the first auction house to sell a work of art created by an algorithm. The work eventually sold for $432,500, nearly 45 times the high estimate!
- “Deepfake” technology is successfully replacing faces in videos. The applications of this technology range from the light-hearted (virtually putting Harrison Ford into the recent Solo movie) to the nefarious (making it appear as if politicians or leaders said things they didn’t).
- Natural Language Generation (NLG) technology, like that being created by Narrative Science, is solving a variety of business communication problems across a range of industries by explaining what’s happening in data via plain English. Applications of this technology are numerous and include report automation, dashboard annotation, and ensure everyone in the organization is empowered to understand and act on business data.
What’s changing that enables this shift toward creative AI? The most straight-forward answer is the explosion in the amount of data that is now being captured. All this data pushes creative AI forward in two ways.
First, the huge amount of data means that deep learning approaches can be applied. Deep learning is a “black-box” approach, which means that the AI’s decision-making process is inherently impossible to understand or explain. This makes the approach a big risk in business settings, where transparency, accountability, and accuracy are paramount. Deep learning approaches are powerful, however, in areas where accuracy is subjective and “sloppiness” is acceptable, such as virtually painting new portraits.
On the NLG side, the vast amount of data gives the computer something useful to communicate and create. NLG has been around for 50 years, but for many of those years, there wasn’t much useful for the computer to say. Now, though, after organizations have invested enormous amounts of money getting their data clean and well structured, there’s increasing awareness of the need for “last-mile” information and insight communication. There’s a huge amount of potential value that is still untapped in business data, because the expertise and effort required to extract it (e.g. via dashboards, Excel, etc.) is too high for many people in the organization. This kind of problem is a perfect fit for NLG, and I’m convinced that people will increasingly rely on NLG-powered stories and conversation to better understand their world as its reflected in data. Implemented correctly, like we do at Narrative Science, means that these systems can convey useful information and insights, without sacrificing the accuracy, transparency, and accountability that is so important in business.
The trend toward creative AI is going to continue and even accelerate in the coming years. While there are certainly risks (e.g. public trust in a world of “deep fakes”), it’s also exciting to be entering a new phase of AI. We’re already beginning to see how NLG can help bridge the gulf between how people and computers understand the world by allowing the machine to explain what it sees in the data in a way that people can understand and act on. As AI becomes more creative, human and artificial intelligence will be able to function more and more as a team, instead of two entirely different species. Is your organization poised to take advantage?