Artificial intelligence made exponential leaps in 2020. Self-driving vehicles are now creeping into the mainstream while advances in machine learning are changing the way we write code and discover drugs. Big A.I. firms like San Francisco-based OpenAI and Alphabet subsidiary DeepMind announced new developments in 2020 that could become backbones of tomorrow’s tech while startups continued to grow despite the pandemic.
Here are the best products, the most intriguing people and the most exciting trends of the year in A.I.
Best Product: Google Smart Compose
While many of the biggest innovations in artificial intelligence are happening outside the view of the average consumer, Smart Compose is an everyday sight on Gmail and Google Docs. Launched by Google in 2018, the time-saving feature uses the tech giant’s machine-learning algorithms to predict and auto-suggest the text a user is planning to type. Smart Compose was widely released on Google Docs only in February, and the technology is still in its early days. “I think it’ll get to a point where you put some bullet points, and it can autocomplete a paragraph,” says Forbes Awards judge Konstantine Buhler, a partner at venture capital powerhouse Sequoia.
Most Disruptive Innovator: Nvidia
Nvidia, known for making the powerful chips that power gaming PCs, Bitcoin mining rigs and A.I. networks, continues to push the envelope. In 2020, it had a series of breakthroughs that significantly reduced the amount of data needed in “generative adversarial networks,” a deep-learning method of creating images, sounds and videos that did not exist before. It’s putting the technology to use in Maxine, a videoconferencing platform that the company says can reduce the bandwidth of calls tenfold by pinpointing key facial features to create A.I.-generated images of people.
Most Intriguing Newcomer: Alpha Fold 2
For decades, protein folding has been a pain point for scientists endeavoring to create new drugs. The puzzle of how proteins curl into 3D shapes has been solved for only a tiny fraction of the 200 million proteins known to science. Alpha Fold 2, which was unveiled by Google’s DeepMind in November, uses a computational model trained on some 170,000 known protein structures to predict how undeciphered proteins will fold. Initial results suggest the model is accurate, on average, to the width of a single atom, a “potentially momentous breakthrough” for a process that has so far depended on cumbersome and costly manual techniques.
Outstanding Firm: Zoom
All those endless video meetings and strained virtual happy hours might have been worth it after all: Zoom has also had an indirect yet material impact on artificial intelligence. A.I. is only effective if given data to utilize, and pandemic-era Zoom has created a massive data bank that wasn’t available during the age of in-person meetings. “Zoom has enabled us to effectively live life, in some ways, digitally,” Buhler says. One example of its use cases: the sales software of startup Gong, which uses A.I. to automatically transcribe and annotate Zoom calls so that sales reps can quickly jump to the meat of their meetings.