From 3M Health Information Systems
AI Talk: 10 breakthrough technologies 2021
In this week’s blog, I want to focus on the annual prognostication from MIT Technology Review. This year is particularly relevant, as it marks two decades of such efforts by the publication. Not only does MIT Technology Review look ahead to 2021, they also look back at their predictions from 2001. Some predictions that were a bust from 2001: brain-machine interfaces (pretty much in the same state today as in 2001, though still relevant) and flexible transistors. Others from that time have become more relevant, like data mining and natural language processing. Clearly, these latter fields have thrived.
So, what about the publication’s current list of predictions? It is actually a pretty good list with a number of pandemic related predictions as well. Here is a summary of that list:
- mRNA technology— Decades in making, the pandemic catapulted this technology to practical relevance. Both the Pfizer and Moderna vaccines are based on this technology and it clearly has promise for a range of other diseases.
- GPT-3— This 175 billion parameter language model has the remarkable ability to write fluent text, complete imaginative story lines and do very well in random conversations. However, the uber-environmental footprint to train the model and the biases it perpetuates by using a large corpus of internet content are sobering counterweights to its use. Nevertheless, it is a big technology breakthrough which clearly needs to be tamed to be useful.
- TikTok recommendation algorithms – Their uncanny ability to connect random content creators to consumers has catapulted their use to stratospheric levels.
- Data trusts— A model being bandied around in Europe and a few other countries that would collect, protect and manage data from users for the good of society and individuals. Sort of like trade unions for data protection, but unlikely to take hold unless regulators force the hand of big tech.
- Lithium-metal batteries— A revolutionary new technology that attempts to replace the more traditional lithium-ion batteries that power cell phones to laptops to electric vehicles (EVs). The technology has the potential to fundamentally alter the EV marketplace.
- Green hydrogen— This refers to hydrogen generated by electrolysis (splitting water into hydrogen and oxygen) where the actual electricity comes from renewable sources such as wind and solar. If the electricity is from fossil fuels or natural gas, the hydrogen is referred to as “gray hydrogen.” Hydrogen in fuel cells and other forms is a potent energy source which can drive powerplants to cars and the use of green hydrogen is critical to achieving the climate goal of zero carbon emissions.
- Digital Contact Tracing— This is one technology which failed to achieve its goals. Retrospective analysis of why it failed shines a light on lack of trust in the app and its potential for privacy compromising solutions. This serves as a good lesson to remember and apply in the event of another pandemic.
- Hyper-accurate positioning— There is a makeover that’s emerging for the standard Global Positioning System (GPS) which powers our navigation apps. The next iteration, GPS-III, brings the accuracy of the system from few meters to within a few centimeters. The tech is already being used in China and will slowly show up here in the U.S. (multiple satellite deployments supporting this tech are underway). For those of you wondering how to navigate in space? To the rescue, Quantum Positioning Systems. This technology is kind of Star Wars-y at the moment, but perhaps we will see it’s in use in a few decades.
- Remote everything— The pandemic fundamentally changed how the world responds to remote work. Education and health care are two areas that changed radically, as remote learning and telehealth use grew exponentially. Even as the pandemic winds down, we will continue to see significant use of these technologies.
- Multi-skilled AI— The past few years have brought about significant advances in speech technology, vision technology and natural language processing. This has led researchers to dream about combining these different capabilities to create a multi-modal intelligent system. Current systems are super smart in one area (like AlphaGo) and super dumb in others. The goal is for systems to become more intelligent and show a modicum of common sense by combining multiple sensory inputs from speech to vision to text.
V. “Juggy” Jagannathan, PhD, is Director of Research for 3M M*Modal and is an AI Evangelist with four decades of experience in AI and Computer Science research.