Fight fire technology with fire

[ad_1]

CEO of NTT Research, a research leader in physics and computer science, cryptography and information security, and medical and health informatics.

getty

Technology gives and technology takes away, as they say. We like modern communications and digital infrastructure, but are concerned about cyberattacks, loss of privacy, bias, energy consumption and more. Certain combinations such as surveillance cameras and artificial intelligence (AI)-driven facial recognition look particularly menacing and scary.

What’s the right way to maximize the pros while minimizing the cons? One tactic is for industry and government to set limits. Another approach is to fight technology with more and better technology.

Limits And Innovation

Setting limits is important. An example is the AI ​​Bill of Rights proposed by the White House. Industries may engage in self-regulation, but cybercriminals and other bad actors regularly violate laws and agreements. Regulations should also try to avoid inhibiting innovation.

Quantum computing, artificial intelligence, and other tech geniuses may deliver a lot of wishes, but there’s no practical way to put them back in the bottle. Innovators will keep inventing. In many cases they will also find solutions to the problems that technology has created. It’s a bit like fighting fire with fire.

AI and quantum computing

Cybersecurity offers a case in point. Security experts believe that machine learning (ML) is becoming a critical element of Security Information and Event Management (SIEM) platforms, which store and analyze threat-related information. With cyber attackers leveraging automated software bots, businesses can counter by tapping into ML to harden their defenses, identify attacks, and respond faster than any human security analyst can.

Quantum computing is a related example. The fear is that, sooner or later, a fault-tolerant quantum computer will be used to crack existing cryptographic systems. It’s a scary prospect, but as I noted in a previous article, the race is on. Post-quantum cryptography has become a common topic at many academic conferences. The National Institute of Standards and Technology (NIST) is creating new standards and commercial solutions are emerging.

Energy consumption is also a concern. For example, the computational requirements of deep neural networks, which are pervasive in running AI technologies today, are very concerning from an energy consumption perspective. One way to address the inefficiencies of digital circuits is to reinvent and demonstrate computing as a physical process, as some physicists have recently done. It is potentially much more energy efficient.

As for other threats posed by AI, regulators are starting to act. There is the proposed AI Bill of Rights mentioned above, as well as the European Union (EU) AI Act, which is expected to become law in 2023. Two of the five principles in the US document concern algorithmic discrimination and data privacy. As in other cases, we can address these concerns through both limitations and technology enablers.

Addressing privacy and bias

One way to prevent AI from exhibiting bias is to stop developing applications that do. In recent years, several major tech companies, including Meta, have either abandoned facial recognition or limited their involvement. One policy suggestion is to allow consumers to opt out of AI-powered applications. That said, opting out would be difficult to implement, in part because it would mean abandoning the potential for otherwise useful and effective solutions.

When implemented, video surveillance combined with facial recognition software raises questions of bias and privacy. Here too new technologies offer possible solutions. In a recently conducted hackathon that my company hosted, the winning demo used a combination of attribute-based encryption (ABE) and ML to obfuscate all faces in a closed circuit television (CCTV) feed, while allowing streamlined control of encrypted data view to authorized users.

As it happens, the winning team came from Belgium, where such an application would have come in handy in a subway station incident a year ago in which a woman was pushed onto the platforms. The idea is to protect the privacy of victims and bystanders by helping authorities, assuming all the proper anti-bias protections are in place, arrest someone seen committing a crime.

A bigger problem now is how to tell real from fake or how to fix biased AI. Given the huge attention and powerful potential (good and bad) that ChatGPT has received lately, people are realizing that we don’t have an effective method to do these tasks. We are now at the stage where we are looking for such technologies.

Between Hope and Fear

Technology can stimulate both hope and fear. If the “tech hit” is indeed on the rise, anxiety is on the rise as well. New laws and regulations are likely to follow.

Meanwhile, it’s helpful to remember the dynamic nature of science and technology. Working in basic research, I know that progress is erratic and may take longer to materialize. But troubleshooters, as a rule, aim for upgrades, not downgrades. And the solutions to technical challenges often lie within the technologies themselves.

Forbes Technology Council is an invite-only community for world-class CIOs, CTOs, and technology executives. Do I qualify?

[ad_2]