Genpact looks to ethical artificial intelligence to chart path ahead

With artificial intelligence (AI) continuing to play a strong role in 2020, the road ahead will be built on “augmented Intelligence” — a mix of human and machine intelligence that will give an edge to companies, says Genpact’s chief digital officer Sanjay Srivastava.

Enterprises could see more positions like digital ethics officers and move towards providing transformation as a service, an area Genpact is heavily focused on. Srivastava said the company is in the fourth phase of evolution and will focus on areas like eliminating bias in intelligent systems and digitising newer kinds of work.

“With AI, it’s no longer that you have to protect your applications, because I can corrupt your data and then perfect AI will run on wrong data and give you the wrong results. I foresee that corporate boards are going to start thinking about ethical governance very differently. Just like we have audit subcommittees and compensation subcommittees, the world is going to have digital ethics subcommittees as a board-level subcommittee,” said Srivastava.

Concerns over bias in AI systems and their efficiency are not new, but there seems to be an increase in board-level recognition of the pitfalls of this. Genpact is using AI in multiple ways, most notable being its AI platform called Cora, which collects and analyses the available data and makes recommendations to clients.

One of the things it is working on is language. “We’re big believers in natural language processing. We apply it in invoice processing, contract reconciliation, any place you have to compare two large PDFs and derive information,” said Srivastava.

The next big area is computer vision, which works on what are called knowledge graphs, or deep learning. “You can turn your iPhone and say, where's the nearest Starbucks and it will tell you where it is. But if you look at an invoice and on page number 23, (you say) the sales tax you applied for in one state [that] you need to take to another state, there’s no in AI in the world today that can solve it, because it’s not contextualised,” he added. In this context, how do digital officers deal with ethics, which are subjective even offline? The answer, said Srivastava, is not hundred per cent laid out.

“But I do think that the parameters are now starting to get defined. So I actually work with clients on this. I have them set up a framework....one of the dimensions is an awareness of the intended and non intended usage parameters of the AI that we recommend,” he said.

Microsoft chief Satya Nadella's prescription to do away with AI bias

That artificial intelligence (AI) applications can be sometimes be biased is no longer unusual, and when that happens, Microsoft — one of the biggest proponents of AI — has a lot to worry about. But Satya Nadella, the chief executive officer (CEO) of the Redmond-headquartered tech giant, also has a simple solution to do away with such biases.

Currently on a three-day visit to India, the Hyderabad-born Nadella says it’s not the AI engine, but the teams that build it as well as the algorithms to make it intelligent have to be diverse and inclusive, so that the solutions they build do not show up any sort of ‘unconscious bias’ that they may have.

“One of the best things you can do to have ethics around AI and to protect (it) from bias, would be to have diverse teams,” said Nadella, speaking at the company’s ‘Future Decoded’ tech summit in Bengaluru on Tuesday.

“If there is one thing that is going to be the real currency for the next 10 years, is how diverse and inclusive your teams that are building these technologies are. There is no protection against unconscious human bias, other than diversity,” he added.

The Microsoft CEO said the developers have to think about privacy as a human right and there is a need to have gender and ethnic diversity in the teams that are building these technologies.

Speaking at the summit which was attended by the developer community and leaders from the information technology industry, Nadella said an economy like India’s needs to adopt digital technologies across sectors, such as retail, health care, and agritech, in the next 10 years. Reflecting on his visit to India at the start of a new decade, Nadella said the impact of digital technology has been tremendous in the last 10 years.

“It is just stunning. The entire mobile revolution came to fruition; the Cloud happened (and) a lot of consumer internet businesses got built. The last 10 years has been amazing, but it has been narrow in some sense,” said Nadella, dressed down in a checkered shirt, denims, and trainers. “But in the next 10 years, can we dream even more? The good news is that the core technology fabric is getting embedded in the world in our homes, offices, stadia, and hospitals. Whether it is a car or a refrigerator, you basically have computing in it.”

Nadella, who is slated to meet a group of students, developers, and entrepreneurs in New Delhi on Wednesday, said the company’s goal is to help every organisation become a software company by adopting the latest technology. They can create their own digital innovations on top of it. “The worst mistake anyone can make is reinventing the wheel.”

As technology becomes ubiquitous, he said there is a responsibility for developers and companies to think a lot more broadly about the impact on society. One such responsibility is building trust into the technology, like every bank which builds a mobile app and is doing transactions, will need to also foster privacy and provide cybersecurity for assets and customer data. He said, today some $1 trillion is lost in the world economy due to cyberattacks and crime.

“Breaking the world is not going to work in the next 10 years. You have to think about that inclusive nature of economic growth, that trust in technology, and sustainability,” said Nadella.

Nadella also said that Indian businesses and start-ups are emerging as innovation leaders, signalling how technology can drive growth, enhance customer experience, and tackle some of the country’s most pressing societal and environmental issues.

Giving examples of some top start-ups using Microsoft’s technology, he said Udaan, a business-to-business e-commerce platform, has scaled up very fast. He also highlighted firms such as Sun Mobility, which provides battery-as-a-service, and Bionic Yantra, which is aiming to restore mobility for severely injured people through wearable robotic exoskeleton aiding rehabilitation.

“It is fantastic to see these unicorns and (start-ups) coming out of India, building successful businesses and models and most importantly, having an impact,” said Nadella.

Nadella said there is unprecedented opportunity to apply technology to drive economic growth that is inclusive, trusted, and sustainable everywhere, including in India. “That’s why we are partnering leaders in every industry across the country to help them build their own digital capability, transform their organisations, and achieve more in this new era.”

This AI system may help doctors treat patients with traumatic brain injury

Researchers, including one of Indian-origin, have developed an artificial intelligence (AI) based system to help doctors treat patients with traumatic brain injury (TBI) -- a significant global cause of deaths, especially in low-and-middle income countries.

TBI patients are unconscious, making it challenging for doctors to accurately monitor their condition during intensive care, the researchers, including those from the University of Helsinki in Finland, said.

The study, published in the journal Scientific Reports, noted that many tens of variables are continuously monitored in the ICU such as pressure within the skull, mean pressure exerted on the blood vessels, and the force driving oxygen flow into the brain.

According to the researchers, these parameters indirectly give information regarding the condition of the patient, with intracranial pressure alone yielding hundreds of thousands of data points per day.

Since, there are millions of daily collected data points from the ICU for a patient, the researchers develop an artificial intelligence (AI) based algorithm to predict the outcome of individual patients, and give objective data regarding their condition and prognosis over the course of treatment.

"A dynamic prognostic model like this has not been presented before. Although this is a proof-of-concept and it will still take some time before we can implement algorithms like this into daily clinical practice, our study reflects how and into what direction modern intensive care is evolving," said study co-author Rahul Raj from the University of Helsinki.

The study noted that the new AI system can predict the probability of the patient dying within 30-days with accuracy of 80-85 per cent.

The scientists created two algorithms -- the first algorithm, a simpler one based only on objective monitor data, and the second a slightly more complex system that includes data on the level of a patient's consciousness, the study said.

"As expected, the accuracy of the more complex algorithm is slightly better than for the simpler algorithm. Still, the accuracy of both algorithms is surprisingly good, considering that the simpler model is based upon only three main variables and the more complex upon five main variables," said Eetu Pursiainen, study co-author from the University of Helsinki.

The researchers hope to validate the algorithms using national and international external datase

Only 10% of Indian CEOs confident about reliability of AI applications: PwC

Only 10 per cent of Indian CEOs are confident about the reliability of their artificial intelligence (AI) applications, according to PwC India.

Even as AI has the potential to solve complex problems effectively at scale, badly designed applications can cause more harm than good, PwC India said in its report based on a comprehensive study conducted with over 1,000 CXOs and business decision makers from India and other regions, between May and September 2019.

The intent of the study was to understand the outlook towards AI in India. Findings of the report strongly indicate the need to invest in building AI systems that are responsible, understandable and ethical, ensuring customer trust.

AI can be defined as a collection of technologies which are capable of sensing, thinking and acting like rational human beings. The respondents spanned across industries such as technology, media and telecom, financial services, professional services, health, industrial products, consumer markets, government, and utilities, said PwC India.

"An overwhelming majority of decision makers globally as well as from India confessed that they may not have robust tools or processes for ensuring reliability of their AI solutions.

"Interestingly, only 10 per cent of Indian respondents were confident about the reliability of their AI applications," it said.

Deepankar Sanwalka, Leader- Advisory at PwC India, said that it is encouraging to see Indian organisations adopt or willing to adopt AI significantly in the coming few years.

However, to scale AI initiatives, organisations will have to ensure these solutions are ethically sound, compliant with all regulations, with a robust governance framework, he said.

The report further said "it is heartening" to note that India (62 per cent) is not very far behind global (65 per cent) in terms of implementation of AI.

However, the worrying part is that Indian respondents (53 per cent) significantly outnumber their global counterparts (36 per cent) in admitting that they have no formal approach to identify AI risks, it said.

Sudipta Ghosh, Leader- Data and Analytics, PwC India added that merely adopting AI will not yield desired results.

AI must be supported by strong performance pillars addressing bias and fairness, interpretability and explainability, robustness and security, he said.

"Else, the enthusiasm to implement AI projects is very likely to run into headwinds. Benefits of AI may be realised when an appropriate governance framework and dimensions are in place, and humans and machines can collaborate effectively," Ghosh said.

The study, PwC India said reiterates the need for a comprehensive Responsible AI (RAI) framework and toolkit for its widespread adoption.

It also highlights how AI's potential can be unlocked as well as maximised if a structured approach is taken towards addressing the associated risks.

India should integrate AI with education to become world leader: Sikka

Former Infosys CEO Vishal Sikka, who has announced a new AI startup with $50 million fund, believes India has the potential to become a world leader in artificial intelligence but the key to this is integrating AI into the country's education system in a massive way.

India is at "an inflection point" when it comes to AI or artificial intelligence, Sikka said.

Over the next 20-25 years, AI is going to be "a very, very big disruptor" for the Indian society because what one is seeing now in terms of automation and job losses because of automation is just the beginning, said Sikka, who announced his startup Vianai Systems last week.

"But on the other hand, if we are able to bring AI education, the ability to build AI systems to India at a very large scale, and I'm talking about like billion plus people, then India can really leap frog and become the world's leader in artificial intelligence, in AI skill and AI talent," Sikka told PTI in an exclusive interview.

Doing that requires working on multiple dimensions in parallel, he said.

Last month, at the request of Prime Minister Narendra Modi, Sikka gave a presentation before the NITI Aayog how to expand the reach of AI to the Indian society in a very big way.

Representatives of some 20 Union ministries were present during his presentation on AI and India. This, he said, required creating necessary infrastructure to bring the talent through institutions, schools and educational institutions, the ability to do AI education at a large scale.

According to Sikka, the prime minister said he personally saw whenever classes worked into digital classrooms, he was joking that children would sometimes even forget to eat their lunch because they were so engrossed in learning. "It was very encouraging. But I think a lot of that has to be done," he said and suggested multi-faceted countrywide programme like digital classrooms.

If India does nothing then this great wave of AI is going to have massive disruption over the next 20 years. But on the other hand, if it puts together programmes then this can be a huge advantage for it and "we can be a leader in the world," he said.

Referring to his interactions with the Indian government officials and the steps being taken by the prime minister, Sikka said he is very encouraged by the commitment.

His Vianai Systems is an AI Enterprise Platform startup with a mission to help businesses around the world successfully leverage AI to drive fundamental digital transformations.

He said there have been remarkable achievements in the field of AI, but conceded that significant issues too have emerged in the field of artificial intelligence.

"The technique themselves are quite opaque, not transparent. The lack of ability to explain is one of the major weaknesses of these techniques," Sikka said.

He said that there is an incredible shortage of talent in the field of AI.

With this in backdrop, Sikka said he started this venture to build a platform that can dramatically improve the accessibility of AI, that is educational, that explains to people how this works, that they can experiment with it, they can play with it, they can explore and understand what is happening.

"That is what my prototype was about," said Sikka as he went back to his PhD thesis on AI in 90s.

For its mass application in a transparent manner, he said there is a need to build a platform that is a dramatic improvement on what is available today.

Sikka has raised $50 million as the startup seed money.

"Hopefully, we wouldn't have to raise any money again," he said, adding that the objective is to bring AI to businesses on massive scale. Last week, he demonstrated the first glimpse of the platform at the Oracle Open World event in the Silicon Valley.

One of the reasons that there are so few applications of engineer machine learning is because the techniques and the tools are extremely dense, they are complicated, the experience of developer is broken profound.

"So we wanted to make it dramatically simpler to access an AI system and to execute on AI. His platform dramatically simplifies the AI system from thousands of lines of codes to a few dozens of lines of codes. This will make the systems much much simpler," he said.

"We have been thinking very carefully, and very diligently about how we can come up with a platform that is completely agnostic, completely open and yet makes it dramatically simpler, and more efficient and more exploratory to build an AI system. And with that when you think about a future where enterprises will deploy dozens, if not hundreds of AI applications, you need something like this to make that happen. And that's what we are working on," Sikka said.

Last week, he announced a panel of advisors from industry and academia for Vianai.

They include Henning Kagermann, former chairman and CEO of SAP and chairman of Acatech; Alan Kay, Turing Award winner and computer science pioneer; Divesh Makan, founder of ICONIQ Capital; Indra Nooyi, member of the board at Amazon and Schlumberger and former CEO of PepsiCo; and Sebastian Thrun, CEO of Kitty Hawk Corporation and co-founder and chairman of Udacity.

"Every company that hopes to be a leader in their industry needs to understand the transformative potential of AI, and to successfully activate it within their organisations," said Nooyi in a statement.

"The market need is immense, and Vianai, with its capable leadership and differentiated platform, is in a unique position to address it. I look forward to being an advisor to the company in the exciting times ahead," she said.

Jack Ma takes on Elon Musk over future of Artificial Intelligence

JAck Ma believes artificial intelligence poses no threat to humanity, but Elon Musk called that "famous last words" as the billionaire tech tycoons faced off Thursday in an occasionally animated debate on futurism in Shanghai.

The Chinese co-founder of Alibaba and the maverick industrialist behind Tesla and SpaceX frequently pulled pained expressions and raised eyebrows as they kicked off an AI conference with a dialogue that challenged attendees to keep up, veering from technology to Mars, death, and jobs.

However, the hot topic in the hour-long talk was AI, which has provoked increasing concern among scientists such as late British cosmologist Stephen Hawking who warned that it will eventually turn on and "annihilate" humanity.

"Computers may be clever, but human beings are much smarter," Ma said. "We invented the computer -- I've never seen a computer invent a human being."

While insisting that he is "not a tech guy," the e-commerce mogul added: "I think AI can help us understand humans better. I don't think it's a threat."

Musk countered: "I don't know man, that's like, famous last words." He said the "rate of advancement of computers in general is insane", sketching out a vision in which super-fast, artificially intelligent devices eventually tire of dealing with dumb, slow humans.

"The computer will just get impatient if nothing else. It will be like talking to a tree," Musk said.

Mankind's hope lies in "going along for the ride" by harnessing some of that computing power, Musk said, as he offered an unabashed plug for his Neuralink Corporation.

Neuralink aims to develop implantable brain-machine interface devices, which conjures images of "The Matrix", whose characters download software to their brains that instantly turns them into martial arts masters.

"Right now we are already a cyborg because we are so well-integrated with our phones and our computers," said Musk, 48.

"The phone is like an extension of yourself. If you forget your phone, its like a missing limb."

But humanity will also have more leisure time in the future as AI takes on much of the burden of transporting, feeding, and thinking for earthlings, said Ma.

"People could work as little as three days a week, four hours a day with the help of technology advances," he said.

Ma, 54, who steps down next month as head of Alibaba Group, questioned Musk's push to develop spacecraft that could help us colonise Mars.

"We need a hero like you, but we need more heroes like us improving things on earth," Ma said.

Musk countered that we must master interplanetary travel in case earth becomes uninhabitable.

Scientists like Hawking have said the same, citing the risk of nuclear war, a devastating virus, global warming or asteroid collision.

But not to worry: both agreed that human mortality is a good thing as each generation brings new ideas to the global challenges we face.

"It's great to die," Ma said, with Musk adding: "That's probably true.

In a first, an AI taught itself to play a video game, and is beating humans

Since the earliest days of virtual chess and solitaire, video games have been a playing field for developing artificial intelligence (AI). Each victory of machine against human has helped make algorithms smarter and more efficient. But in order to tackle real world problems – such as automating complex tasks including driving and negotiation – these algorithms must navigate more complex environments than board games, and learn teamwork. Teaching AI how to work and interact with other players to succeed had been an insurmountable task – until now.

In a new study, researchers detailed a way to train AI algorithms to reach human levels of performance in a popular 3D multiplayer game – a modified version of Quake III Arena in Capture the Flag mode.

Even though the task of this game is straightforward – two opposing teams compete to capture each other’s flags by navigating a map – winning demands complex decision-making and an ability to predict and respond to the actions of other players.

This is the first time an AI has attained human-like skills in a first-person video game. So how did the researchers do it?

The robot learning curve

In 2019, several milestones in AI research have been reached in other multiplayer strategy games. Five “bots” – players controlled by an AI – defeated a professional e-sports team in a game of DOTA 2. Professional human players were also beaten by an AI in a game of StarCraft II. In all cases, a form of reinforcement learning was applied, whereby the algorithm learns by trial and error and by interacting with its environment.

The five bots that beat humans at DOTA 2 didn’t learn from humans playing – they were trained exclusively by playing matches against clones of themselves. The improvement that allowed them to defeat professional players came from scaling existing algorithms. Due to the computer’s speed, the AI could play in a few seconds a game that takes minutes or even hours for humans to play. This allowed the researchers to train their AI with 45,000 years of gameplay within ten months of real-time.

The Capture the Flag bot from the recent study also began learning from scratch. But instead of playing against its identical clone, a cohort of 30 bots was created and trained in parallel with their own internal reward signal.

Each bot within this population would then play together and learn from each other. As David Silver – one of the research scientists involved – notes, AI is beginning to “remove the constraints of human knowledge… and create knowledge itself”.

The learning speed for humans is still much faster than the most advanced deep reinforcement learning algorithms. Both OpenAI’s bots and DeepMind’s AlphaStar (the bot playing StarCraft II) devoured thousands of years’ worth of gameplay before being able to reach a human level of performance. Such training is estimated to cost several millions of dollars. Nevertheless, a self-taught AI capable of beating humans at their own game is an exciting breakthrough that could change how we see machines.

The future of humans and machines

AI is often portrayed replacing or complementing human capabilities, but rarely as a fully-fledged team member, performing the same task as human beings. As these video game experiments involve machine-human collaboration, they offer a glimpse of the future.

Human players of Capture the Flag rated the bots as more collaborative than other humans, but players of DOTA 2 had a mixed reaction to their AI teammates. Some were quite enthusiastic, saying they felt supported and that they learned from playing alongside them. Sheever, a professional DOTA 2 player, spoke about her experience teaming up with bots:

It actually felt nice; [the AI teammate] gave his life for me at some point. He tried to help me, thinking ‘I’m sure she knows what she’s doing’ and then obviously I didn’t. But, you know, he believed in me. I don’t get that a lot with [human] teammates.

Others were less enthusiastic, but as communication is a pillar of any relationship, improving human-machine communication will be crucial in the future. Researchers have already adapted some features to make the bots more “human friendly”, such as making bots artificially wait before choosing their character during the team draft before the game, to avoid pressuring the humans.

But should AI learn from us or continue to teach themselves? Self-learning without imitating humans could teach AI more efficiency and creativity, but this could create algorithms more appropriate to tasks that don’t involve human collaboration, such as warehousing robots.

On the other hand, one might argue that having a machine trained from humans would be more intuitive – humans using such AI could understand why a machine did what it did. As AI gets smarter, we’re all in for more surprises.

Maude Lavanchy, Research Fellow in Behavioural Economics, IMD Business School and Amit Joshi, Professor of Artificial Intelligence, IMD Business School

This article is republished from The Conversation under a Creative Commons license. Read the original article.

Google's AI system better than humans at spotting lung cancer: Study

Scientists at Google have developed an artificial intelligence (AI) model which they claim is better at diagnosing lung cancer than human experts, an advance that could lead to earlier treatments for the deadly disease.

Deep learning -- a form of AI -- was able to detect malignant lung nodules on low-dose chest computed tomography (LDCT) scans with a performance meeting or exceeding that of expert radiologists, researchers said.

The system, described in the journal Nature Medicine, provides an automated image evaluation system to enhance the accuracy of early lung cancer diagnosis that could lead to earlier treatment.

The deep-learning system was compared against radiologists on LDCTs for patients, some of whom had biopsy confirmed cancer within a year. In most comparisons, the model performed at or better than radiologists.

Deep learning is a technique that teaches computers to learn by example.

The deep-learning system also produced fewer false positives and fewer false negatives, which could lead to fewer unnecessary follow-up procedures and fewer missed tumours, if it were used in a clinical setting.

"Radiologists generally examine hundreds of two-dimensional images or 'slices' in a single CT scan but this new machine learning system views the lungs in a huge, single three-dimensional image," said Mozziyar Etemadi, a research assistant professor at Northwestern University in the US.

"AI in 3D can be much more sensitive in its ability to detect early lung cancer than the human eye looking at 2D images. This is technically '4D' because it is not only looking at one CT scan, but two over time," Etemadi said.

"In order to build the AI to view the CTs in this way, you require an enormous computer system of Google-scale. The concept is novel but the actual engineering of it is also novel because of the scale," he said.

This research is incredibly important, as lung cancer has the highest rate of mortality among all cancers, and there are many challenges in the way of broad adoption of screening, said Shravya Shetty, technical lead at Google.

"Our work examines ways in which AI can be used to improve the accuracy and optimise the screening process, in ways that could help with the implementation of screening programs. The results are promising, and we look forward to continuing our work with partners and peers," said Shetty.

Large clinical trials across the US and Europe have shown that chest screening can identify the cancer and reduce death rates, researchers said.

However, high error rates and the limited access to these screenings mean that many lung cancers are usually detected at advanced stages, when they are hard to treat, they said.

The deep-learning system utilises both the primary CT scan and, whenever available, a prior CT scan from the patient as input.

Prior CT scans are useful in predicting lung cancer malignancy risk because the growth rate of suspicious lung nodules can be indicative of malignancy.

The computer was trained using fully de-identified, biopsy-confirmed low-dose chest CT scans.

The novel system identifies both a region of interest and whether the region has a high likelihood of lung cancer.

The model outperformed six radiologists when previous CT imaging was not available and performed as well as the radiologists when there was prior imaging.

"The system can categorise a lesion with more specificity. Not only can we better diagnose someone with cancer, we can also say if someone doesn't have cancer, potentially saving them from an invasive, costly and risky lung biopsy," Etemadi said.

Google scientists developed the deep-learning model and applied it to 6,716 de-identified CT scan sets to validate the accuracy of its new system.

They found the Ai-powered system was able to spot sometimes-minuscule malignant lung nodules with a model of 0.94 test cases.

The researchers cautioned that these findings need to be clinically validated in large patient populations.

However, they said this model may assist in improving the management and outcome of patients with lung cancer.

China's prowess in AI is making US nervous; can it win from tech cold war?

China’s growing technological prowess in areas such as artificial intelligence is making Washington very nervous. US efforts to fight back, though, could make the problem worse.

In US policy circles, suspicion of China is starting to resemble a new Red Scare. Universities are heightening scrutiny of research proposals from China and, in some cases, restricting collaboration. Chinese scientists’ visas are being delayed for conferences and exchanges. Visas for Chinese graduate students studying topics such as robotics or advanced manufacturing have been shortened to one year from five.

Last week, the M.D. Anderson Cancer Center in Houston kicked out three senior researchers of Chinese ethnicity after the US National Institutes of Health said they had potentially violated disclosure and confidentiality rules. Workers at various technology companies have been charged with stealing trade secrets in recent months.

More formal rules are coming. After President Donald Trump signed the Export Control Reform Act last year, the US put in place new policies to restrict Chinese investment in American high-tech companies. It also began a process of reexamining export controls on sensitive “emerging and foundational” technologies. That process is nearly complete: The Commerce Department’s Bureau of Industry and Security is holding seminars over the next few months to help companies understand how to comply with the tighter restrictions.

While the details are still murky, one thing is clear about the new rules: Like the old ones, they’ll apply not just to hardware shipped overseas, or even software and algorithms. They will cover individuals and ideas as well.

The disclosure of proprietary information or controlled information to a foreign national, even within the US, triggers the need for an export review process. If such an exchange were to happen elsewhere in the world — between a US national and a foreign national — it would be considered an export to that country.

Among other things, this means that US tech companies may not be able to use Chinese researchers to work on certain critical technologies, including AI. The kind of intellectual collaboration that’s produced some of the greatest advances in the field may no longer be possible.

This could be a boon for China. Beijing for years has tried to lure talent to the mainland. It’s set up generous benefits for so-called sea turtles — Chinese nationals who have “swum” abroad to study and work and then returned — from cheap housing to lucrative research grants. Those efforts have had mixed success: Fears of censorship and red tape have dissuaded some; a desire to live and work with the best minds in the world has motivated others to stay on in the US

If new export controls are enforced in their harshest form, the calculus for many Chinese scientists and engineers may change. Their job possibilities in the US will be limited. A potentially cumbersome visa process could discourage many American companies and universities from even interviewing qualified Chinese candidates.

For those who do find work, living under constant suspicion will be far from pleasant. In a letter to Science magazine last month, associations representing Chinese-American scientists said even they fear being singled out and racially profiled.

Some Chinese researchers may seek opportunities in Europe instead, or in countries such as Israel or India with thriving high-tech industries. Many more are likely to go home.

This would drain the US of some of its best talent; institutions such as the Massachusetts Institute of Technology rely on foreign students and academics for a significant part of their research capability. It’ll also limit the ability and willingness of US companies and universities to learn from Chinese counterparts: The Commerce Department’s latest “unverified list” — essentially a red flag that requires US companies to do more diligence and provide additional information about a listed company or person, while imposing license requirements — includes mostly Chinese academic institutions, research centers and tech companies.

More conversations will begin taking place outside the US In its global strategy report, MIT noted that “America’s relative economic weight in the world has been declining for decades, and as other countries grow more prosperous, a growing share of global R&D is originating outside the US” Collaborations between Chinese and European researchers will no doubt increase; Washington’s inability to persuade its allies to reject technology from Huawei Technologies Co. has shown the limits of its influence.

When it comes to certain sectors, including artificial intelligence and autonomous vehicles, an influx of new talent now could spur major advances in China. The country already offers some key advantages for researchers — including the reams of data available and fewer privacy concerns. Smart researchers could leverage those tools to leapfrog US technology.

Export controls covering technical information have been around since at least the 1940s. In the 1980s, the Reagan administration also erected barriers to scientific collaboration, fearing Soviet efforts to steal US technology. Universities were pressured to close their doors to foreigners, among other isolationist measures.

Such restrictions ultimately failed to slow the transfer of ideas and technologies. If anything, they blinded the US to its rivals’ growing capabilities. Washington might want to study that history, along with where technology is headed, before isolating itself again.