AI year in review: Opportunities grow, but ethics loom large
Artificial intelligence garnered a lot of attention from the usual players — governments, tech giants, and academics — throughout 2019. But it was also a big year for business AI, with even more growth expected ahead. In a March KPMG survey, more than half of business executives said their company would implement enterprise-scale AI within two years. That is partly what drives PwC estimates that AI will deliver $15.7 trillion to the global economy by 2030.
Impressive leaps in 2019 have enabled new business applications for virtual assistants, such as Salesforce’s updated Einstein Voice Assistant for sales and customer service apps and IBM’s intelligent agent Watson Assistant. Meanwhile, governmental deployment of AI around the world has led to abuses and concomitant regulation. But along with concerns about power in AI comes the technology’s potential to help make everyday life a little better.
Automated AI for the enterprise
The enterprise cloud market heated up with increased implementation of automated machine learning (AutoML) that allows customers to apply AI to use cases such as marketing, customer service, and risk management. The biggest players in cloud computing — Google, Microsoft, and Amazon — spotlighted AI tools and automation in their annual tech showcases. Microsoft summed it up at Ignite 2019 with its tagline for Azure Cognitive Search: “Use AI to solve business problems.”
At the Google Cloud Next conference in April, Google announced new AutoML classes, premade Retail and Contact Center AI services, and the collaborative model-making tool AI Platform. In December, Amazon launched a blizzard of AI-powered enterprise tools at its re:Invent 2019 conference.
One of the more intriguing tools from Google Cloud Next was AutoML Natural Language, released widely in December, which analyzes text from a range of document and formatting types to feed sentiment analysis, legal document parsing, and publications management. Amazon rolled out a similar tool for AWS, called Textract, in April. Microsoft, meanwhile, pumped up the subject matter virtual agents, sentiment analysis, and business process automation available on its business-focused Power Platform.
AI at the edge
At the other end of the network — the edge — software advances like federated and multimodal learning are enabling artificial intelligence on smartphones and other devices, with the promise of greater control and better privacy protections compared to AI processed in the cloud. In June, Apple introduced Core ML 3, which allows iOS devices to perform machine learning for the first time. Google incorporated federated learning into its TensorFlow development environment back in 2017, and the effort is bearing fruit: In October, Google promoted the many AI touches on the Pixel 4 smartphone, from speech recognition to greatly improved camera features.
Hardware is also becoming more efficient, with “real AI” powered by mobile chips. Examples abound: Arm is building up its product line to power machine learning and AI in a wide selection of devices. Intel promoted Keem Bay, a vision processing unit that brings inferencing tasks to edge devices. Google offered Coral AI, a range of boards and kits for neural network machine learning that work on the edge. And Nvidia released the Jetson Xavier NX to power AI for drones, cars, and other mobile edge devices.
In addition, a new focus on power efficiency could help reduce the environmental (and financial) impact of running all those AI systems. Google created a controller that keeps its experimental quantum processor cool enough to function while using just 2 milliwatts of power. On the consumer side, Facebook announced DeepFovea, an AI technique that improves the power draw of VR headsets. And even closer to home, Sense released a line of AI devices to monitor and reduce household energy use, while Evolve Energy’s AI helps solar and wind power customers find the best prices and save energy.
Shipping and shopping
Besides consumer energy monitors like the above, 2019 saw huge advances in areas like autonomous cars and the internet of things (IoT). AI also made inroads into such everyday tasks as grocery shopping.
Self-driving cars from the likes of Uber, Lyft, Alphabet’s Waymo, Tesla, and Argo are the pretty face of autonomous vehicles, and consumer sentiment reports suggest the public is warming to the idea. But commercial trucking was where the money was in 2019. Carmaker Volvo is so confident in the viability of its smart trucks that it’s going to break out its driverless financials starting in 2020, although it faces competition from the likes of TuSimple, which is testing delivery for the U.S. Postal Service; Daimler, which is testing autonomous trucks in Virginia; and Starsky Robotics, which relies on remote teleoperators to run its test fleet.
Competition in the AI assistant market is still greatest between Amazon and Google, rivalry that has spurred the performance and capabilities of voice recognition and personal assistants. Amazon and Microsoft launched the Voice Interoperability Initiative in September, along with a slew of partners — absent Apple, Samsung, and Google — that seek to allow devices to run more than one assistant. There’s good reason for Microsoft to join, since it’s become clear its Cortana is not beating Amazon’s Alexa or Google Assistant anytime soon. But Samsung’s Bixby also stepped back in acknowledgment of its market position. As for which voice assistant is best, Google Assistant keeps coming out on top in accuracy tests, although in a May 2019 test by Tom’s Hardware, Alexa and Siri were not far behind.
AI went mainstream for grocery shopping in 2019. Walmart is using AI to improve online grocery ordering, employing machine intelligence to figure out what consumers are likely to need, and it’s begun using driverless vans to ferry goods between stores in Arkansas. Meanwhile, Microsoft helped grocery chain Kroger create cashierless stores using smart shelves and other intelligent technology, while the Giant Eagle chain turned to Grabango for its own AI trial.
Politicians and corporations clashed throughout 2019 over the appropriate use and oversight of AI. Famously, one of Democratic senator and presidential candidate Elizabeth Warren’s campaign promises is to break up big tech businesses like Google and Amazon. “With fewer competitors entering the market, the big tech companies do not have to compete as aggressively in key areas like protecting our privacy,” Warren’s campaign blog states. And her rivals brought up AI specifically in the Democratic debates.
Beyond the general concern of private-market encroachments upon personal freedoms, governments from the Massachusetts town of Somerville to the nation of the U.K. are examining how the public sector should use AI technology — and coming to differing conclusions.
While China has been using facial recognition to regulate cell phone accounts and allegedly round up the Uighur minority population, most governments in the U.S. that examine the technology do so to limit or ban its use. Especially in California, San Francisco and other cities are enacting bans on use of facial recognition by public entities, particularly police departments.
Detroit has no such ban, and indeed its police chief, James Craig, is enthusiastic about facial recognition’s potential to fight crime. This led to an August 2019 Twitter beef for the ages between Chief Craig and Detroit’s U.S. Congressional Representative, Rashida Tlaib (D-MI), that ended in an awkward demonstration of the technology that frustrated both sides.
The tide could turn if President Trump wins re-election in 2020, as his administration takes a more collaborative approach to AI. If the eventual Democratic nominee is Senator Bernie Sanders (D-VT), and he wins, Chief Craig might be out of luck. But legislation to regulate facial recognition has been surprisingly nonpartisan so far.
Despite privacy concerns over government use of AI, the technology can improve everyday life, especially if proper care is taken to consider the ethics.
The amount of data being created by smart cars, roadside cameras, public transit, and other sensors is overwhelming, but by feeding it into AI systems, companies like Waycare are helping cities predict and improve traffic flow. StreetLight Data takes a different approach: By tapping into cellphone location data, it can track and predict traffic for vehicles, bicycles, and pedestrians. London is using Waze to tackle traffic congestion in the city center and reduce air pollution. Elsewhere, Alphabet division Sidewalk Labs is helping Toronto push the envelope of smart city technology, fed by weather and usage patterns, to create a high-tech innovation district.
Norway has emerged as a hotbed for startups leveraging AI to build better cities. Oslo-based Spacemaker‘s software allows city planners to estimate the effect of each planning decision and optimize for a range of goals, courtesy of machine learning. And in August, the city of Trondheim unveiled Powerhouse, a smart office building designed to generate more energy than it consumes and apply the excess to powering other smart city tech, such as road monitoring.
As 2019’s projects come to fruition in 2020 and beyond, it will be interesting to watch how AI develops in the real world. Ethical oversight will be needed to make sure the technology continues to serve humanity, rather than the other way around.
Join us at VB Transform 2020, the AI event of the year for enterprise executives, brought to you by today’s leading AI publisher. You can purchase tickets below.