System does biometric facial reading of a woman.
Authentication by facial recognition concept. Photo: metamorworks

Marco DeMello, from CyberLabs Group: “AI will change our world in ways that we can’t even predict right now”

Earlier this month, Brazil launched its national AI strategy. To understand what this means in the global and regional context, LABS' U.S. correspondent Lauren Simonds, spoke with DeMello, today in charge of the largest cybersecurity research laboratory in Latin America

Ler em português

Marco DeMello knows a thing or two about artificial intelligence and security. The president of CyberLabs Group — formed when his cybersecurity company PSafe merged with CyberLabs, a Brazilian AI startup, in February 2021 — owns 10 software patents and has more than 20 years of leading global brands and startups that protect the world’s computers and mobile devices under his belt.

According to DeMello, we are now at a fundamental tipping point in our battle with cybercriminals that has only intensified during the last year of the global pandemic. He describes it as a “forever war” that will not stop, and one in which the old security paradigms like passwords and fingerprint scans are no longer safe or effective. In 2020 alone, more than 600 million malicious URLs were detected in Brazil, and an attack was recorded every 16 seconds.

Marco DeMello, president of CyberLabs Group, which emerged in February this year from the merger between CyberLabs and PSafe. Photo: CyberLabs Group/Courtesy

CyberLabs Group currently operates in both Brazil and the U.S., and DeMello spoke with us about the challenges, the promise and the opportunities of AI with a core focus on Brazil.

LS: The global pandemic has affected everything. What challenges or opportunities has the pandemic revealed in terms of using AI-based tools across the social and business spectrum?

DeMello: It’s a very interesting topic. A tremendous amount of AI was applied to the development of the COVID-19 vaccines, for example. It would never have happened in that timeframe without AI. That’s a clear example of how AI can be extremely positive to society.

In Brazil, we used our biometric recognition systems to help maintain social distancing regulations during the most restrictive times — without capturing any individual data. By using human recognition (not individual facial recognition) and neighborhood CCTV cameras, we could immediately identify a neighborhood or street where excessive crowds were gathering, and the government could take action. [Rio de Janeiro is one of the cities that uses CyberLabs technology for this purpose].

READ ALSO: With more than 4,000 algorithms, Fligoo wants to get to know you better

LS: Brazil recently released its new AI strategy. What does it cover, and can you talk about its strengths and weaknesses with respect to addressing the challenges of developing a world-class AI industry in Brazil?

DeMello: The focus of the Brazilian AI strategy announcement, as we understand it, is to contribute and determine an ethical set of principles for the development and use of responsible, socially acceptable AI. We need to eliminate bias and surveillance. The European Union has been leaning very hard against this type of surveillance without strong justification. 

The strategy promotes sustained AI investment, research, and development so that we can bridge Brazil’s qualified professional gap. For example, this study by the Organization for Economic Co-operation and Development (OECD) reports that by 2030, China will host 37% of the world’s STEM (Science, Technology, Engineering, Mathematics) graduates, while having 16% of the population. Compare that to Brazil, which will have 2.7% of the world’s population — but only about 2% of STEM graduates. We need to expand AI research, remove barriers to AI innovation, and train professionals for the new AI ecosystem.

The world is going through a transformation, not unlike the discovery of electricity. AI is as impactful and — will be as life-changing — as electricity, and we need to adjust and adapt to this change.

Adapting to this change will require cooperation between public and private entities, industry research centers, and academia. In this way, Brazil can become a hub for promoting and incentivizing the development of AI projects.

Brazil has several key strengths. We have a large population and a very large amount of data because Brazilians spend more time online — in terms of session length — than other people in the world. We have a lot of societal problems that we need to address, and we have many systemic inefficiencies we need to solve. AI can help do that. These are all strengths because they contribute to the development and advancement of AI. 

Brazil’s main weakness is a lack of skilled labor. We need a mechanism that motivates students to join a Ph. D. program for artificial intelligence or to choose a major focused on machine learning and AI.

READ ALSO: On the edge of becoming Latin America’s new unicorn, NotCo takes baby steps in the U.S. plant-based market

LS: The E.U. has taken a tough stance with big tech for years, and its new guidelines restrict, for example, mass surveillance and using artificial intelligence to manipulate people. The U.S. Federal Trade Commission has taken a softer approach. But a recent post indicates the agency will go after companies whose products use biased algorithms. How does Brazil’s AI strategy compare?

DeMello: Brazil is in a slightly different position. The Brazilian government, in partnership with UNESCO, hired a technical consultancy to study the potential economic and social impacts of AI.

The project involves academic experts, researchers, and professionals who are deeply familiar with issues related to AI adoption and development. It looks at attracting R&D companies, international partnerships, using AI in healthcare and public safety, and how to deploy AI to address Brazil’s problematic infrastructure.

It also addresses the economic benefits, ethical principles, and implications of AI at a societal level — like privacy, bias, and surveillance. Brazil wants a fair and just adoption of AI to eliminate issues with algorithmic bias. There’s a lot happening.

LS: Regulating AI seems an impossible task. What are the challenges inherent in regulating a technology that can do both great harm and great good? Is it too late — can anyone put this genie back in the bottle?

DeMello: I alluded to this earlier, but AI is like electricity in the sense of its impact. You can’t put electricity back in the bottle. Once electricity was discovered, the entire world shifted to electricity. AI will change our world in ways that we can’t even predict right now. 

We need to embrace these changes, not stymie them. But, we must do this responsibly. The government can’t cherry-pick with overregulation. That’s not productive.

We need clear guidelines and a small number of very strong regulations with sharp teeth so that big tech does not violate them. Overregulation will set us back years, rather than help us develop and evolve a more ethical and just system. That should be the focus.

The biggest pushback with true weight behind it, is that AI takes away jobs. A lot of people are going to be unemployed because of AI. How do you deal with that?

You don’t save these jobs with outdated policies, political maneuvering, or unrealistic regulations that make no sense. We need to adopt balanced privacy and data-protection policies, and then help people get new jobs in a world where AI performs many of the repetitive tasks. We need to prepare society and not try to stop it.  

LS: What are the top priorities for improving AI in Brazil? How would you rank ethical use, data privacy, cybersecurity, and customizing AI so it can accurately reflect the country’s diverse population?

DeMello: You cannot separate ethical AI use from a person’s cybersecurity or data privacy because they’re so intertwined. You must address the three issues together, because individuals need their right to privacy, they need to be safe, and they must be allowed to benefit from AI innovation.

Brazil has the highest percentage of people unfairly arrested and/or convicted due to mistaken eyewitness identification. Of all the unjust convictions that are overturned, 70% are due to bias and racial discrimination during eyewitness identification.

At CyberLabs, eliminating racial bias has been a fundamental tenet of our design architecture and technology from day one. We’ve beaten algorithms from all the major big-tech companies in public tests, and we believe we have the most advanced facial-recognition AI in the world — with absolutely no racial bias.

READ ALSO: Google phone cameras will read heart and breathing rates

LS: Brazil tops the list of countries around the world being affected by cybercrimes. Why is cybercrime so prevalent in Brazil and other parts of Latin America?

MD: Actually, Brazil and the U.S. are very close in numbers of cyber-attacks and fraud. That’s because of the explosion of social media and social engineering. Cybercriminals have enough data from your social-media profiles and from data leaked on the Dark Web to perpetrate very sophisticated social-engineering attacks.

This is done by robots. The idea of a teenager wearing a hoodie in a basement attacking someone may be good for a Hollywood movie; but it’s fantasy.

Our cybersecurity unit at PSafe detected the largest data breach in Brazil’s history. The personal records of 223 million people, some records included photos for facial recognition, were leaked to the Dark Web and commercialized for bitcoins.

Organized crime controls the Dark Web, and it will certainly use that data to implement intelligent bots to attack individuals. You cannot defeat AI with standard algorithms. The only way to fight this type of cybercrime is with AI.

The tools employed to protect your data and privacy one or two years ago no longer work. Full stop. Companies that provide defense have a duty to provide cybersecurity that’s based on artificial intelligence. By that, I mean heuristics, behavioral analysis, and intelligent context that determines whether an action can take place or not.

LS: It seems old security measures like passwords and fingerprint scans no longer cut it. What’s needed in terms of a new approach to cybersecurity? What will take the place of the old methods — and how does advanced AI play a role?

MD: The most important point I can make is how critical it is for governments and private institutions to wake up to the fact that passwords are no longer effective. Hackers no longer break in; they log in. They have your passwords.

It’s a 50-year-old form of authentication — why are we still using passwords to identify people? The only solution for our new reality, especially post-pandemic, is to use AI and implement an intelligent means of identification and authentication within the context of what a user is doing.

Using continuous biometric authentication eliminates the “I forgot my password” issue. You don’t have to remember a billion passwords anymore. You don’t have to know something; you just have to be someone. You confirm you are the person through your face or your voice, and you’re done.

Passwords are the bane of society, and we need to get rid of them now. It’s urgent. Biometric authentication is far stronger, and it will solve the huge problem cybersecurity that we have today. It’s 2021. We need to remove passwords from any critical decision-making.

LS: Is there enough AI and programmer talent to help combat the cyberwars? If not, where will it come from? How do you plan to beef up talent in the weeks and months ahead?

MD: The short answer is yes. CyberLabs sponsors a lot of scholarships to study AI and machine learning in Brazil. For example, we sponsor the Center for Excellence in AI. This center — part of an initiative to bridge the AI divide between government, companies and universities — has consistently delivered cutting-edge global research.

AI’s key value proposition is that it learns and improves over time. If you train the model sufficiently, eliminate bias, and you have the right priority set, you don’t need an army of AI developers. You need a focused group of AI developers building an AI that improves itself and keeps getting better.

LS: The Android OS still owns the market share in Brazil and other parts of Latin America due, in part, to the high cost of Apple devices. Does that cause inherent issues because it’s more complicated to protect the myriad Android device choices vs. relying on trusted Apple?

MD: Android is the most targeted operating system, because there are more Android devices than Windows devices in the world, and Brazil — at 90% Android versus 10% iOS — is no exception.

AI bots can perpetrate many types of assaults on open platforms like Android. They only need to get one shot right to get your data, but your defense must be right 100% of the time.

The human brain can’t win this fight. You cannot safely maintain your devices regardless of the platform they use. Android is more vulnerable than iOS, but iOS is not as safe as people think. It’s a matter of when — not if — you fall victim to an attack.

Artificial intelligence is the only way to protect systems because organized crime will continue to expand its onslaught of artificial intelligence-based attacks. And the defense needs to be just as strong, or your fight is lost before it even begins. 

Cybersecurity software is essential for every one of your devices. I don’t care which company you trust. Do your research and find a company to provide cyber security locally on your device and use it 24/7. It’s not a fair fight, and only AI can fight AI.

LS: Do you have any far-out or surprising predictions for where AI leads us in the next five, 10 or 20 years?

MD: I’ll give you my favorite prognosis: continuous biometric authentication will replace passwords within 10 years. Also, AI and speech-to-text and text-to-speech technology will dramatically transform communication for blind, low-vision, deaf, and mute people. Accessibility will take on a whole new meaning within the next decade.