The buzzword status of artificial intelligence and its assorted subtopics can obscure products that are actually using AI, but there are some genuinely impressive applications in use right now.
The broadly defined AI sector reeled in $22.7 billion in investment last year — led by the automotive, data science and healthcare industries, according to ABI Research. Market watchers say that as enterprise organizations in industries like healthcare look to replace aging legacy software in the next couple of years, AI-driven solutions are poised to fill that lucrative void.
That trend isn’t just restricted to healthcare. Companies throughout the Bay Area are building technology that can write, read and act on written language; augment farming operations; and streamline access to credit. Some of the entrepreneurs associated with these products are claiming their AI technology is ready (or close to ready) for use by globe-spanning enterprise organizations.
Improve Access to Credit
First recorded among ancient Sumerians around 5,500 years ago, credit has long been an engine of social mobility.
But Upstart co-founders Dave Girouard, Anna Counselman and Paul Gu believe the United States’ score-based system of assessing potential loan recipients has become a barrier to credit access. In a December 2019 study, the company found that while 80 percent of Americans have never defaulted on a credit product, just 48 percent of the country has access to prime credit.
By leveraging AI, banks can improve and help de-risk their lending activities...”
Upstart leverages AI to rethink the loan approval process. Trained on more than half a million loans and around 7 million repayment transactions, Upstart’s algorithms assess more than 1,500 variables per consumer. The company recently released an API designed to help banks and other lenders make credit decisions.
“Upstart’s AI models have significantly outperformed traditional FICO-based models before and during the COVID-19 crisis,” said Girouard, the company’s CEO, in a statement. “By leveraging AI, banks can improve and help de-risk their lending activities during even the most rapidly changing economic environments.”
Producing Language
Using a set of principles centered around broadly distributed benefits, technical leadership, long-term safety and cooperative orientation, OpenAI is racing to be the first organization to build artificial general intelligence — while making sure the results are safe and equitable for all stakeholders. And if you’re familiar with the so-called “paperclip problem,” you know that in the case of AGI, “all stakeholders” means “the entire human race.”
Much of OpenAI’s recent work has focused on AI systems that can understand and produce written human language. Last year, the company rolled out GPT-2, a text-generation system that can produce original written content based on a prompt. Developers have used it to create their own text tools.
Given any text prompt, the API will return a text completion, attempting to match the pattern you gave it.”
OpenAI is now working on GPT-3 and funding further AI research by selling access to the new suite of tools through an API. Would-be users request access to a private beta, and the company assesses each application for potentially harmful use cases.
“Unlike most AI systems which are designed for one use-case, the API today provides a general-purpose ‘text in, text out’ interface, allowing users to try it on virtually any English language task,” the company wrote in a blog post this month. “Given any text prompt, the API will return a text completion, attempting to match the pattern you gave it. You can “program” it by showing it just a few examples of what you’d like it to do.”
Efficiency in Agriculture
For farmers, the margins between loss and profit can be razor-thin. A piece of land’s viability can vary from year to year, and sustainability is a major concern for intensive farming operations — yet many of these calculations traditionally rely on a farmer’s intuition. In recent years, however, farmers are increasingly plugging field data into AI-driven analytics platforms to help inform decisions. Agtech platforms are also turning algorithms into visual datasets like satellite photographs to detect issues.
“Many in agriculture are becoming accustomed to near real-time imagery showing the crop health of their fields,” wrote Matthew Hayes, a data scientist at San Francisco-based agtech company Granular, in a recent blog post.
This higher resolution imagery will allow the detection of crop health issues in near real-time at an unprecedented level of detail.”
Hayes pointed to cubesat swarms — low-cost miniature satellites that deploy in large numbers to produce detailed image data on farms around the world every day — as an infrastructural advance that will allow further agtech innovation.
“This higher resolution imagery will allow the detection of crop health issues in near real-time at an unprecedented level of detail,” wrote Hayes.
Granular’s suite of data analytics platforms use AI to estimate costs and expected yields for specific land parcels, helping to inform budgeting decisions. The software also helps farmers plan future crops to optimize yields, profit and long term sustainability.
Solving IT Issues
Most AI-enabled IT troubleshooters merely track user complaints or gather a few details before passing them on to IT teams. But technology that can automatically understand and resolve IT tickets is now available, driven by advances in natural language understanding — an NLP subtopic focused on reading comprehension in machines.
One leader in the field is Mountain View-based Moveworks, which has trained its algorithms to decipher meaning from the language used inside enterprise organizations. It’s a very different style of conversation than the way we would address an Alexa or Google Assistant, for example. In an interview with TechCrunch last year, Founder and CEO Bhavin Shah said Moveworks solves between 35 and 40 percent of IT issues, empowering knowledge workers to focus on business-critical innovation.
The Moveworks bot converses with employees in natural language straight on the messaging tool they use every day...”
“Providing rapid, 24/7 support to a remote workforce is not a realistic possibility for human agents alone, particularly given the increasing volumes of tickets (during the COVID-19 pandemic),” Shah wrote in a March blog post.
“The Moveworks bot converses with employees in natural language straight on the messaging tool they use every day…Moveworks then determines what an employee is asking for using a combination of machine learning models, before autonomously resolving the issue via deep integrations with other software.”
Artificial intelligence software traditionally replaces repetitive manual tasks. But as it becomes better at problem solving and communication with humans, watch for AI to move into more service roles as well. Moveworks says demand has surged since mass work-from-home policies went into effect, and Business Insider recently predicted it will be one of 15 tech startups to come out of the COVID-19 pandemic even stronger than before.
