Editor’s note: This article originally was published in the February issue of Florida Trend magazine.
AI can’t replace humans, but the technology is making inroads in more and more business sectors.
In an oft-quoted interview with Life magazine in 1970, Marvin Minsky, an MIT researcher and pioneer in artificial intelligence, predicted that scientists were about three to eight years away from creating a machine as intelligent as the average human. Such a machine would “be able to read Shakespeare, grease a car, play office politics, tell a joke, have a fight,” Minksy said, and it would learn at such a “fantastic speed” that it would reach genius level within just a few months.
Fifty years later, Minksy’s vision of a machine on par with the human brain still hasn’t been realized — but popular AI tools, such as Google’s search engine and Apple’s Siri, have become part of everyday life, and machines are learning how to master an array of complex tasks, from operating self-driving cars to spotting tumors to monitoring crops.
“It’s no longer, ‘is artificial intelligence going to work?’ It does work, and there’s numerous applications that are out there — medical imaging, credit fraud detection, movie selections — that use sophisticated artificial intelligence algorithms to make business better,” says Jeff McFadden, chief technology officer for Xonar Technology, a Largo-based company that designed an AI-enabled security surveillance system to detect weapons. “It’s really moved out of that research area and moved into where it’s really a commercially viable technology set.”
The leaps forward have been made possible by powerful computer processing engines and advances in machine learning techniques. Computers with enough horsepower can crunch large data sets and use a series of algorithms to extract patterns and glean insights from that information. In Al recommendation systems, like that employed by Netflix, the algorithm looks at an individual’s viewing history, considers the preferences of other members with similar tastes and evaluates other information to come up with a viewing suggestion. An algorithm in an intelligent credit card fraud detection service, on the other hand, flags suspect purchases by looking for outliers or anomalies that depart from a consumer’s normal purchasing behavior.
McFadden’s Xonar uses an algorithm to spot concealed weapons based on their ultra-wide band radar signature. The system transmits an electromagnetic pulse toward a person as they walk through it. The pulse bounces back to a receiver. That reflected wave is then analyzed by machine’s algorithm to see if it matches the shape, density and other characteristics of various weapons in its learning library. It’s more discriminating than traditional technologies, which rely primarily on metal detection, and Xonar can differentiate between a knife or handgun and other harmless metal items, such as keys or money clips, making for fewer “false positives.” It’s also less obtrusive. McFadden says friends who’ve breezed through the system in place at the entrances of Ruth Eckerd Hall in Clearwater didn’t even realize it was there.
Inspired by neural networks in the brain, these “deep learning” systems can remember and build on observational patterns they find in data. In essence, they become smarter over time, but it’s vital to control what data the systems receive. “The big AI in general, deep-learning specifically, learns everything. It learns what you want it to learn, but it learns what you don’t want it to learn,” McFadden cautions.
In some cases, AI can outperform its human counterparts. A recent study in the journal Nature found that a Google AI system did a better job in predicting breast cancer from mammograms than radiologists did. And a new AI software engineered by the British company DeepMind has created a system of algorithms called AlphaFold that can rapidly and reliably predict the 3-D shape of proteins — a task that usually takes months or years. The breakthrough is expected to speed up and reduce the costs of pharmaceutical development.
Here’s a closer look what companies and researchers across Florida are doing with AI technology and why the experts say it will never completely replace humans.
Gaps in Groves
Designer: Yiannis Ampatzidis (UF spinoff Agriculture Intelligence)
Product: Agroview, citrus tree counter
UF technology is helping orange growers count their trees and optimize tree growth.
Orange trees were easier to count before citrus greening struck. Fields were relatively uniform then, and farmers who had planted 160 trees per acre on 10 acres knew with reasonable certainty that they had about 1,600 trees. But when greening swept through Florida about a decade ago, groves became riddled with gaps where diseased and dead trees once stood. To get an accurate count, growers had to hire workers to drive through their groves, counting each tree, one by one, with a clicker.
“It was very time consuming and very expensive,” says Yiannis Ampatzidis, a University of Florida scientist stationed in Immokalee. To address the problem, Ampatzidis and his research team developed a system called Agroview that uses special images taken from drones and the ground to count the trees and assess their conditions. The method isn’t perfect, but it’s close. It tallied 175,977 citrus trees at one Hendry County farm with close to 98% accuracy. It’s also quick, taking about 10% as long as a manual count.
Florida growers seized on the technology in the aftermath of Hurricane Irma, when the U.S. Department of Agriculture began requiring them to submit accurate tree inventories to maintain insurance coverage. Others have used Agroview to pinpoint gaps in groves so they can replant where dead trees once stood. Growers can also use the technology to detect disease and optimize the health of plants.
Agroview combines multispectral imaging and leaf analysis. The drones carry special cameras that capture image data of the trees across a broad portion of the electromagnetic spectrum, including wavelengths invisible to the eye. Data from those images are then fed into Agroview’s AI software, where they’re cross-referenced with data from leaves that have been analyzed in labs to identify plants with nutrient deficiencies, diseases or other types of stress. Armed with that information, growers can fine-tune their application of fertilizer, pesticide and other inputs. If the system indicates low levels of nitrogen or phosphorus, for instance, the grower can increase his application of those nutrients in the affected areas. If nutrient levels are high, the grower can ease up. If disease is detected, the grower can apply a treatment early, before it spreads through the entire crop.
Farmers can purchase Agroview through a UF spinoff company called Agricultu
re Intelligence. The company will also collaborate with drone operators if growers don’t have their own drone to collect the data. “Right now, we’re working mainly with citrus, but tomato and other crops, other profiles, will come soon,” Ampatzidis says.
UF researchers are also collaborating with a farm equipment company in Clewiston called Chemical Containers to develop a “smart sprayer” that uses AI and “sensor fusion” for precise application of insecticides and weed killer. “If you have a traditional sprayer, it will spray everywhere with the same amount, but that doesn’t make sense,” Ampatzidis says. “If you have a big tree, you need more chemicals to be sprayed, but if it’s small, half the size, in theory you need half. If it’s a gap, you don’t spray at all.”
Designer: Goodby Silverstein & Partners
Product: Dali Lives, a re-creation of the artist
Salvador Dali, the Spanish surrealist painter, died in 1989 at age 84. But visitors to the Dali Museum in St. Petersburg can interact with a life-like version of the artist that’s been re-created using artificial intelligence. Unveiled in 2019 on what would have been Dali’s 115th birthday, Dali Lives was created using AI and a faces-wapping technique known as a deep-fake. Using archived interview footage and other historical materials, an AI algorithm designed by San Francisco-based Goodby Silverstein & Partners was able to master Dali’s mannerisms.
Following 1,000 hours of machine learning, the AI tool generated a likeness of Dali’s face that was superimposed over an actor’s body and synced with a voice impressionist to create a talking digital replica of the flamboyant artist. The resulting exhibit includes 125 interactive videos, with 190,512 possible combinations depending on user response, meaning no two visitors are likely to have exactly the same experience. Museum visitors are treated to a selfie taken by the digital Dali that they can receive via a text message before they leave the museum.
Designer: Nvidia/University of Florida
Product: HiPerGator supercomputer upgrade
Nvidia and UF are partnering to create “higher education’s most powerful AI supercomputer” (rendering). All colleges at the university are building AI courses related to their area of expertise.
After graduating from the University of Florida in 1980 with his engineering degree, Chris Malachowsky landed his first job at Hewlett-Packard in California, where he designed a central processing unit, or CPU chip. He leveraged that experience into his second job at Sun Microsystems, where he worked on computer graphics until 1993, when he and some colleagues decided to start their own company, Nvidia.
In the years since, Nvidia has transformed visual computing and come to dominate the artificial intelligence landscape with its graphics processing unit, or GPU, technology. Now, the Silicon Valley company is partnering with the University of Florida to supply an AI supercomputer that works with UF’s existing HiPerGator system to create “higher education’s most powerful AI supercomputer,” which will be capable of delivering 700 petaflops of AI performance, or 1 quadrillion operations per second.
The public-private partnership is anchored by a $60-million donation — including $25 million from Malachowsky and $25 million in technology, training and services from Nvidia. UF will contribute $20 million. The school has also committed to hiring 100 faculty members focused on AI and plans to incorporate AI broadly across its curriculum.
While the school already offers courses in machine learning and AI ethics, each college at the university is building AI courses related to its area of expertise. UF’s College of Business, for instance, is working on an AI course that will focus on AI in financial technologies. “We want to make it possible for every student who graduates from the University of Florida and who’d like to learn about AI the opportunity to either become acquainted with it, to become competent in it or become an expert in it,” says Joe Glover, UF’s provost and senior vice president of academic affairs.
Glover says anyone in the State University System will be able to use its new supercomputer for educational purposes to teach students about machine learning and AI. Researchers at all the State University System schools will have access to the new supercomputer for educational purposes at no charge to teach students about machine learning and AI. UF will also provide limited support and training for using its computer resources.
The partnership comes at a critical moment, with the federal government warning that United States is lagging in churning out AI-trained workers. “We think we have a unique approach here to helping to solve that issue. We intend to create the next generation of the AI-enabled workforce at scale, graduating 5,000 to 10,000 people who are going to pour into the economy and bring those skills with them to whatever their chosen occupation is,” says Glover.
COVID, Cancer & Cats
Designer: Ulas Bagci
Product: Predictive health care software
< p class="">Ulas Bagci, an assistant professor in University of Central Florida’s department of computer science, has been working with an international team of researchers to develop AI tools that are helping doctors and nurses across the globe manage COVID-19 patients.
Using images from chest CT scans, their algorithm is able to predict which patients have COVID-19 and the infection’s severity, helping health care providers identify which patients need to go to ICU, which are likely to need intensive care or die and which ones can go home based on inflammation it detects in the lungs.
The AI technology has proved especially helpful in countries such as Italy, Japan, China and the U.S. “The hospitals are full of COVID patients. They need to manage them in the optimal way,” Bagci says.
Bagci has developed similar predictive algorithms to diagnose lung cancer and pancreatic cancer from CT scans and MRI images. He says the algorithms can accomplish in seconds what it takes radiologists “ages to do.” But he says the AI tools are not a replacement for their expertise, but rather an enhancement to lend them support and point them to the problem. “A human should always be in the loop for a trustworthy AI,” he says.
In the classroom, Bagci likes to use the “cat behind the tree” example for his students. “Let’s say there’s a cat behind a tree and on one side you see the head and the other side you see the tail. You understand, as a human, that there is only one cat. But artificial intelligence is too artificial. It allows for two cats. It will give labels like cat 1, tree, and cat 2,” he explains. “With high-level knowledge, we are much better than AI, and for high-risk applications, none of these deep-learning algorithms give you reason. It’s good. It’s really helpful, but it’s not going to replace humans. True intelligence is not there.”
AI-Powered Patient Monitoring
Designer: Chakri Toleti (Care.ai)
Product: Care.ai, patient monitoring software
Chakri Toleti, a serial entrepreneur from the Orlando area, was taking a sabbatical “in between companies” in 2018 when he learned that his 78-year-old mother had fallen in a bathroom in India. She was stranded there for nearly 30 minutes before a caregiver found her. She recovered, but the incident motivated Toleti to create a company called Care.ai that makes AI-powered autonomous patient monitoring systems for hospitals, nursing homes and other health care facilities.
Toleti says the system was inspired by his experience building motion capture systems at Disney 25 years earlier and sensor technology like that used in self-driving cars. Care.ai relies on similar tools — including sensors and an AI-powered “learning library” of behaviors and movements — to predict when a patient is at risk of falling or wandering off. If it senses something worrisome, it alerts nurses. It can also detect whether workers are washing their hands, delivering medications or coming in to check on patients when they should, in effect creating a “self-aware” room.
When COVID-19 hit last year, Care. ai tweaked its platform to screen hospital visitors for signs of infection. The touch-free screening tool — which is used at Tampa General Hospital, Rush University Medical Center in Illinois and other facilities across the nation — has a contact-free thermal sensor to detect feverish visitors, sending a message to staff if a person’s temperature exceeds 100 degrees.
Real Estate Modeling
Designer: Olivia Ramos (Deepblocks)
Product: Deepblocks, early property analysis software
Growing up in Cuba, where she lived until she was 10 years old, Olivia Ramos spent lots of time in the office where her mother worked as an architect. “At the time, they had no computers, so it was all a bunch of pencils and rulers, and I fell in love with all the little gadgets,” she recalls.
Two decades later, Ramos is perfecting her own gadget — a high-tech software application called Deepblocks that uses data and deep learning to streamline and automate the process of early property analysis. Developers using the software can zoom in to a specific parcel, set their building parameters — square footage, number of units, parking, etc. — and the program spits out a 3-D visual of the project and an analysis with a projected return on investment that takes into account everything from market demographics (such as rent-to-income ratios) to local zoning rules.
“Zoning data, the rules of the city, are usually 400- page PDFs and are really expensive to go through and understand,” says Ramos, who has a master’s degree in architecture from Columbia University and a master’s in real estate development from the University of Miami. “We developed models that understand that data and extract that data from those documents. You just select a piece of land, and it tells you what you can do.”
The software can shave considerable time off development planning. It took one customer a year to do 21 iterations of a particular parcel that Deepblocks can help do in a few hours, and users can do as many models as they want, Ramos says. It currently includes parcel data for more than 1,000 U.S. cities and zoning data for 30 cities.
The Miami startup has a staff of six, including CEO and founder Ramos, and has raised $2 million through two funding rounds. It’s raising $3 million in a third round. Real estate pros can buy a subscription to Deepblocks for $1,620 a month or $12,600 for a year. The software has seen a “big growth in adoption” amid the COVID-19 pandemic because people can’t travel as
easily to visit potential markets, Ramos says.
The goal is for the software to make suggestions on opportunities in the market and determine the highest and best use of any property, Ramos says. When that happens, she believes Deepblocks will help tackle even bigger problems, such as a lack of affordable housing.
“It’s really, really hard to make an affordable housing project profitable, and it requires a lot of government help, so if we use the inefficiencies and understand what to build and how to build and where to build it, then that projects a lot of savings on the front end,” she says. “Every single penny you save in affordable housing in cost, it’s going to make that project more likely to happen.”
Designer: Holland & Knight /Joe Dewey
Product: Draft responses software
Joe Dewey, a financial services and real estate attorney and “innovation partner” at Holland & Knight, says his firm has built an AI system that can generate about a dozen draft responses to pleadings for cases in high-volume practice areas. While a human must still sign off on the final version of a document, the law firm has been able to shave about two to three hours off the process of preparing a pleading. “With 50 to 100 cases a month, that starts to add up,” he says.
Dewey sees bigger things on the AI horizon for the legal world — such as a deep-learning tool that could help an attorney draft a motion based on prior decisions issued by a particular judge, or an algorithm that could synthesize all existing case law on a topic and create a memorandum. “If you could get something that could do that at 70% to 80% accuracy, that would be a very valuable tool,” he says. “The technology will need to evolve, but I think that’s the direction we’re headed.”
The Miami lawyer is skeptical that any machine could rival the human brain in a broad way. “For the most part, the machine learning models are good at one task. It’s just a statistical model at the end of the day, and it applies the statistics to make a prediction about something,” Dewey says. “They’re very powerful with the tasks they’re trained to do, but very limited, usually to a very narrow task. Beyond that, they’re fairly dumb.”
An increasing number of companies are leveraging AI to automate more mundane business tasks.
- The Tampa Bay Rays and Rowdies use an AI-power contract management system called IntelAgree, created by the Tampa-based startup CoLabs, to streamline their contract process.
- A 4-year-old Miami company called Chirrp has developed a platform that harnesses IBM Watson’s conversation technology to create more “human-like” chatbots for businesses.
- Vinsa, a company that grew out of the AI-consulting firm Levatas in Palm Beach Gardens, uses its computer vision models along with robots created by Boston Dynamics to automate “labor-intensive” tasks such as reading and monitoring analog utility gauges at electric, oil and gas sites. Vinsa has also built computer vision models that can see whether workers on construction sites are wearing masks and complying with other safety requirements.
- Florida Atlantic University’s College of Engineering and Science is implementing AI-oriented degree programs, including a master of science with a major in AI (the first such degree in Florida), a multi-disciplinary master of science in data and analytics and a joint degree that funnels honors students into a master’s of science in data science.
- Researchers at FAU’s College of Engineering and Computer Science recently landed a five-year, $2.4-million grant from the National Science Foundation to train graduate students in data science technologies and applications.
- With a $1-million gift from Rubin and Cindy Gruber, FAU is building a 3,400-sq.-ft. artificial intelligence lab within its Wimberly Library.
When Ivan Garibay was invited to Washington, D.C., a few years ago to talk to lawmakers about artificial intelligence, many quizzed him about “the singularity” — a theoretical point in time when AI will surpass human intellect — and inquired about whether smart machines will be job killers.
Garibay, founding director of the University of Central Florida’s Complex Adaptive Systems Laboratory, told them the age of AI will be like other industrial revolutions. “Some types of jobs disappear, but it also brings new types of jobs,” he said. “A few years down the road, we probably won’t need Uber drivers or taxi drivers or truck drivers because AI’s getting better and better at guiding cars, but that doesn’t mean that many jobs won’t appear in that new AI economy.” As traditionally, retraining people with new skills, he says, will be key to weathering the disruption.
A 2019 report from the Brookings Institution suggests that the impacts of AI won’t be evenly felt. Those in “better-paid, white-collar occupations” are among the “most exposed” to AI, although the “most elite workers — such as CEOs — appear to be somewhat protected,” according to Brookings’ analysis. The report says that jobs involving “pattern-oriented or predictive work” may be “especially susceptible,” whereas low-paying “rote” jobs involving food preparation, health care and personal care may be less affected.
As for the notion of the singularity — “we’re very, very far away,” Garibay surmises. “That is something I don’t even see in generations, having true intelligent machines that could replace or be a threat to humankind. It’s almost like the Wizard of Oz. It’s always amazing when you see the results, but when you see behind the curtain, you realize what’s there is not as impressive. It’s just mathematics and fast computers. I don’t see it coming. Not in our lifetime at least.”