Posted on

AI is struggling to adjust to 2020

2020 has made every industry reimagine how to move forward in light of COVID-19: civil rights movements, an election year and countless other big news moments. On a human level, we’ve had to adjust to a new way of living. We’ve started to accept these changes and figure out how to live our lives under these new pandemic rules. While humans settle in, AI is struggling to keep up.

The issue with AI training in 2020 is that, all of a sudden, we’ve changed our social and cultural norms. The truths that we have taught these algorithms are often no longer actually true. With visual AI specifically, we’re asking it to immediately interpret the new way we live with updated context that it doesn’t have yet.

Algorithms are still adjusting to new visual queues and trying to understand how to accurately identify them. As visual AI catches up, we also need a renewed importance on routine updates in the AI training process so inaccurate training datasets and preexisting open-source models can be corrected.

Computer vision models are struggling to appropriately tag depictions of the new scenes or situations we find ourselves in during the COVID-19 era. Categories have shifted. For example, say there’s an image of a father working at home while his son is playing. AI is still categorizing it as “leisure” or “relaxation.” It is not identifying this as ‘”work” or “office,” despite the fact that working with your kids next to you is the very common reality for many families during this time.

Image Credits: Westend61/Getty Images

On a more technical level, we physically have different pixel depictions of our world. At Getty Images, we’ve been training AI to “see.” This means algorithms can identify images and categorize them based on the pixel makeup of that image and decide what it includes. Rapidly changing how we go about our daily lives means that we’re also shifting what a category or tag (such as “cleaning”) entails.

Think of it this way — cleaning may now include wiping down surfaces that already visually appear clean. Algorithms have been previously taught that to depict cleaning, there needs to be a mess. Now, this looks very different. Our systems have to be retrained to account for these redefined category parameters.

This relates on a smaller scale as well. Someone could be grabbing a door knob with a small wipe or cleaning their steering wheel while sitting in their car. What was once a trivial detail now holds importance as people try to stay safe. We need to catch these small nuances so it’s tagged appropriately. Then AI can start to understand our world in 2020 and produce accurate outputs.

Image Credits: Chee Gin Tan/Getty Images

Another issue for AI right now is that machine learning algorithms are still trying to understand how to identify and categorize faces with masks. Faces are being detected as solely the top half of the face, or as two faces — one with the mask and a second of only the eyes. This creates inconsistencies and inhibits accurate usage of face detection models.

One path forward is to retrain algorithms to perform better when given solely the top portion of the face (above the mask). The mask problem is similar to classic face detection challenges such as someone wearing sunglasses or detecting the face of someone in profile. Now masks are commonplace as well.

Image Credits: Rodger Shija/EyeEm/Getty Images

What this shows us is that computer vision models still have a long way to go before truly being able to “see” in our ever-evolving social landscape. The way to counter this is to build robust datasets. Then, we can train computer vision models to account for the myriad different ways a face may be obstructed or covered.

At this point, we’re expanding the parameters of what the algorithm sees as a face — be it a person wearing a mask at a grocery store, a nurse wearing a mask as part of their day-to-day job or a person covering their face for religious reasons.

As we create the content needed to build these robust datasets, we should be aware of potentially increased unintentional bias. While some bias will always exist within AI, we now see imbalanced datasets depicting our new normal. For example, we are seeing more images of white people wearing masks than other ethnicities.

This may be the result of strict stay-at-home orders where photographers have limited access to communities other than their own and are unable to diversify their subjects. It may be due to the ethnicity of the photographers choosing to shoot this subject matter. Or, due to the level of impact COVID-19 has had on different regions. Regardless of the reason, having this imbalance will lead to algorithms being able to more accurately detect a white person wearing a mask than any other race or ethnicity.

Data scientists and those who build products with models have an increased responsibility to check for the accuracy of models in light of shifts in social norms. Routine checks and updates to training data and models are key to ensuring quality and robustness of models — now more than ever. If outputs are inaccurate, data scientists can quickly identify them and course correct.

It’s also worth mentioning that our current way of living is here to stay for the foreseeable future. Because of this, we must be cautious about the open-source datasets we’re leveraging for training purposes. Datasets that can be altered, should. Open-source models that cannot be altered need to have a disclaimer so it’s clear what projects might be negatively impacted from the outdated training data.

Identifying the new context we’re asking the system to understand is the first step toward moving visual AI forward. Then we need more content. More depictions of the world around us — and the diverse perspectives of it. As we’re amassing this new content, take stock of new potential biases and ways to retrain existing open-source datasets. We all have to monitor for inconsistencies and inaccuracies. Persistence and dedication to retraining computer vision models is how we’ll bring AI into 2020.

Read More

Posted on

General Atlantic to invest $870M in India’s Reliance Jio Platforms

Mukesh Ambani’s Jio Platforms has agreed to sell its 1.34% stake to General Atlantic, the latest in a series of deals the top Indian telecom operator has secured in recent weeks.

On Sunday, New York-headquartered private equity firm General Atlantic said it would invest $870 million in the Indian telecom operator, a subsidiary of India’s most valued firm (Reliance Industries), joining fellow American investors Facebook, Silver Lake, and Vista Equity Partners that have also made sizeable bets on the three-and-a-half-year old Indian firm.

General Atlantic’s investment values Jio Platforms at $65 billion — the same valuation implied by the Silver Lake and Vista deals and a 12.5% premium over Facebook’s deal, the Indian firm said.

Sunday’s announcement further illustrates the growing appeal of Jio Platforms, which has raised $8.85 billion in the past one month by selling about 14.7% of its stake, to foreign investors that are looking for a slice of the fast-growing world’s second largest internet market.

General Atlantic, a high profile investor in consumer tech space that has invested in dozens of firms such as Airbnb, Alibaba, Ant Financial, Box, ByteDance, Facebook, Slack, Snapchat, and Uber, has been a key investor in India for more than a decade though it has avoided bets in consumer tech space in the country.

It has cut checks to several Indian startups including NoBroker, a Bangalore-based startup that helps those looking to rent or buy an apartment connect directly with property owners, edtech giants Unacademy and Byju’s, payments processor BillDesk, and National Stock Exchange of India. The PE firm, which has invested about $3 billion in India, said last week that it was looking to invest an additional $1.5 billion in Indian firms by next year — this time focusing on the players operating in consumer tech category.

Reliance Industries chairman Ambani, who has poured more than $30 billion to build Jio Platforms, said the telecom network would “leverage General Atlantic’s proven global expertise and strategic insights across 40 years of technology investing.”

“General Atlantic shares our vision of a digital society for India and strongly believes in the transformative power of digitization in enriching the lives of 1.3 billion Indians,” he added.

Prepaid SIM cards of Reliance Jio at a retail store. (Photo: INDRANIL MUKHERJEE/AFP via Getty Images)

Launched in the second half of 2016, Reliance Jio upended India’s telecommunications industry with cut-rate data plans and free voice calls. Jio Platforms, a subsidiary of Reliance Industries, operates the telecom venture, called Jio Infocomm, that has amassed 388 million subscribers since its launch to become the nation’s top telecom operator.

Reliance Jio Platforms also owns a suite of services including music streaming service JioSaavn (which it says it will take public), smartphones, broadband business, on-demand live television service and payments service.

“In just three and a half years, Jio has had a transformational impact in democratizing data and digital services, propelling India to be positioned as a leading global digital economy,” said Sandeep Naik, MD and Head of India & Southeast Asia at General Atlantic, in a statement.

The new capital would help Ambani, India’s richest man, further solidify his last year’s commitment to investors when he said he aimed to cut Reliance’s net debt of about $21 billion to zero by early 2021. Its core business — oil refining and petrochemicals — has been hard hit amid the coronavirus outbreak. Its net profit in the quarter that ended on March 31 fell by 37%.

In the company’s earnings call last month, Ambani said several firms had expressed interest in buying stakes in Jio Platforms in the wake of the deal with Facebook . Bloomberg reported last week that Saudi Wealth Fund was also in talks with Ambani for a stake in Jio Platforms.

Facebook said that other than offering capital to Jio Platforms for a 9.99% stake in the firm, it would work with the Indian giant on a number of areas starting with e-commerce. Days later, JioMart, an e-commerce venture run by India’s most valued firm, began testing an “ordering system” on WhatsApp, the most popular smartphone app in India with over 400 million active users in the country.

29-year-old Akash Ambani, the oldest son of Mukesh, said in a statement, “Jio is committed to make a digitally inclusive India that will provide immense opportunities to every Indian citizen especially to our highly talented youth.”

Read More

Posted on

Can API vendors solve healthcare’s data woes?

A functioning healthcare system depends on caregivers having the right data at the right time to make the right decision about what course of treatment a patient needs.

In the aftermath of the COVID-19 epidemic and the acceleration of the consumer adoption of telemedicine, along with the fragmentation of care to a number of different low-cost providers, access to a patient’s medical records to get an accurate picture of their health becomes even more important.

Opening access to developers also could unlock new, integrated services that could give consumers a better window into their own health and consumer product companies opportunities to develop new tools to improve health.

While hospitals, urgent care facilities and health systems have stored patient records electronically for years thanks to laws passed under the Clinton administration, those records were difficult for patients themselves to access. The way the system has been historically structured has made it nearly impossible for an individual to access their entire medical history.

It’s a huge impediment to ensuring that patients receive the best care they possibly can, and until now it’s been a boulder that companies have long tried to roll uphill, only to have it roll over them.

Now, new regulations are requiring that the developers of electronic health records can’t obstruct interoperability and access by applications. Those new rules may unlock a wave of new digital services.

At least that’s what companies like the New York-based startup Particle Health are hoping to see. The startup was founded by a former emergency medical technician and consultant, Troy Bannister, and longtime software engineer for companies like Palantir and Google, Dan Horbatt.

Particle Health is stepping into the breach with an API -based solution that borrows heavily from the work that Plaid and Stripe have done in the world of financial services. It’s a gambit that’s receiving support from investors including Menlo Ventures, Startup Health, Collaborative Fund, Story Ventures and Company Ventures, as well as angel investors from the leadership of Flatiron Health, Clover Health, Plaid, Petal and Hometeam.

Image via Getty Images / OstapenkoOlena

“My first reaction when I met Troy, and he was describing what they’re doing, was that it couldn’t be done,” said Greg Yap, a partner with Menlo Ventures, who leads the firm’s life sciences investments. “We’ve understood how much of a challenge and how much of a tax the lack of easy portability of data puts on the healthcare system, but the problem has always felt like there are so many obstacles that it is too difficult to solve.”

What convinced Yap’s firm, Menlo Ventures, and the company’s other backers, was an ability to provide both data portability and privacy in a way that put patients’ choice at the center of how data is used and accessed, the investor said.

“[A service] has to be portable for it to be useful, but it has to be private for it to be well-used,” says Yap. 

The company isn’t the first business to raise money for a data integration service. Last year, Redox, a Madison, Wis.-based developer of API services for hospitals, raised $33 million in a later-stage round of funding. Meanwhile, Innovaccer, another API developer, has raised more than $100 million from investors for its own take.

Each of these companies is solving a different problem that the information silos in the medical industry presents, according to Bannister. “Their integrations are focused one-to-one on hospitals,” he said. Application developers can use Redox’s services to gain access to medical records from a particular hospital network, he explained. Whereas using Particle Health’s technology, developers can get access to an entire network.

“They get contracts and agreements with the hospitals. We go up the food chain and get contracts with the [electronic medical records],” said Bannister.

One of the things that’s given Particle Health a greater degree of freedom to acquire and integrate with existing healthcare systems is the passage of the 21st Century Cures Act in 2016. That law required that the providers of electronic medical records like Cerner and EPIC had to remove any roadblocks that would keep patient data siloed. Another is the Trusted Exchange Framework and Common Agreement, which was just enacted in the past month.

“We don’t like betting on companies that require a change in law to become successful,” said Yap of the circumstances surrounding Particle’s ability to leapfrog well-funded competitors. But the opportunity to finance a company that could solve a core problem in digital healthcare was too compelling.

“What we’re really saying is that consumers should have access to their medical records,” he said.

Isometric Healthcare and technology concept banner. Medical exams and online consultation concept. Medicine. Vector illustration

This access can make consumer wearables more useful by potentially linking them — and the health data they collect — with clinical data used by physicians to actually make care and treatment decisions. Most devices today are not clinically recognized and don’t have any real integration into the healthcare system. Access to better data could change that on both sides.

“Digital health application might be far more effective if it can take into context information in the medical record today,” said Yap. “That’s one example where the patient will get much greater impact from the digital health applications if the digital health applications can access all of the information that the medical system collected.” 

With the investment, which values Particle Health at roughly $48 million, Bannister and his team are looking to move aggressively into more areas of digital healthcare services.

“Right now, we’re focusing on telemedicine,” said Bannister. “We’re moving into the payer space… As it stands today we’re really servicing the third parties that need the records. Our core belief is that patients want control of their data but they don’t want the stewardship.”

The company’s reach is impressive. Bannister estimates that Particle Health can hit somewhere between 250 and 300 million of the patient records that have been generated in the U.S. “We have more or less solved the fragmentation problem. We have one API that can pull information from almost everywhere.”

So far, Particle Health has eight live contracts with telemedicine and virtual health companies using its API, which have pulled 1.4 million patient records to date.

The way it works right now, when you give them permission to access your data it’s for a very specific purpose of use… they can only use it for that one thing. Let’s say you were using a telemedicine service. I allow this doctor to view my records for the purpose of treatment only. After that we have built a way for you to revoke access after the point,” Bannister said.

Particle Health’s peers in the world of API development also see the power in better, more open access to data. “A lot of money has been spent and a lot of blood and sweat went into putting [electronic medical records] out there,” said Innovaccer chief digital officer Mike Sutten.

The former chief technology officer of Kaiser Permanente, Sutten knows healthcare technology. “The next decade is about ‘let’s take advantage of all of this data.’ Let’s give back to physicians and give them access to all that data and think about the consumers and the patients,” Sutten said.

Innovaccer is angling to provide its own tools to centralize data for physicians and consumers. “The less friction there is in getting that data extracted, the more benefit we can provide to consumers and clinicians,” said Sutten.

Already, Particle Health is thinking about ways its API can help application developers create tools to help with the management of COVID-19 populations and potentially finding ways to ease the current lockdowns in place due to the disease’s outbreak.

“If you’ve had an antibody test or PCR test in the past… we should have access to that data and we should be able to provide that data at scale,” said Bannister. 

“There’s probably other risk-indicating factors that could at least help triage or clear groups as well… has this person been quarantined has this person been to the hospital in the past month or two… things like that can help bridge the gap,” between the definitive solution of universal testing and the lack of testing capacity to make that a reality, he said. 

“We’re definitely working on these public health initiatives,” Bannister said. Soon, the company’s technology — and other services like it — could be working behind the scenes in private healthcare initiatives from some of the nation’s biggest companies as software finally begins to take bigger bites out of the consumer health industry.

Read More