The ‘training’ of any AI engine uses the method of reducing errors statistics and FR displays are not going to be dumb terminals like a TV screen when installed at public places like airports, offices, etc. The huge photo images being captured by FR engines are like fodder for AI to ‘train’ itself and grow.
Facial Recognition (FR) for automatic identification and recognition of a person is a deep learning Artificial Neural Network technique which forms the part of Artificial Intelligence (AI). Anyone with even a basic understanding of AI can vouch that FR as public use technology is still far from being implemented worthy.
The usual gimmick of showcasing technology as a fad to garner popularity or financial gains is not the same now with self-learning AI machines. The ‘training’ of any AI engine uses the method of reducing errors statistics and FR displays are not going to be dumb terminals like a TV screen when installed at public places like airports, offices, etc. The huge photo images being captured by FR engines are like fodder for AI to ‘train’ itself and grow. “This task undertaken without adequate regulations in place can be risky and push humans to the thin end of the wedge. It is a must that this risk is acknowledged well in time, and strong regulations are evolved to safeguard Indian citizens’ database. For example, issues related to storage of Indian database on foreign servers, etc,” says Milind Kulshreshtha, Artificial Intelligence, and C4I expert.
According to Kulshreshtha, historically, the profiling of citizens as part of National census every other decade was a non-intrusive means to plan governmental social programmes. But now private players capturing the citizens’ photographs for analysis, profiling, etc. as AI gambit has multiple implications like:-
(a) Collating citizens photograph linked with personal details for use and recording by private players including Airport Management firms basically infringes privacy rights. The Supreme Court directive on linking of Aadhaar is no brainer here. Supreme Court terminated Section 57 of Aadhaar Act which wrongfully permitted the private entities to avail Aadhaar data and this SC judgment bars mobile phone service providers and banks to seek Aadhaar authentication from customers. Needless to say, FR is also dwelling on similar lines as Section 57 of Aadhaar Act and risks exposing citizens’ database to private players for use/misuse. As a citizen, do we have to look up to Supreme Court to acknowledge this fact for corrections, but unfortunately that may come too late for effective safeguards. Expecting private players not to undertake any of such tasks if they deem it to be unethical, does not fit their business model itself.
(b) Face Recognition Misuse: Our face can be noted as one of the most important identity features for humans. Any attempt to track citizens through CCTV networks at airports, railway stations, etc. is unethical practice. The database collected at Airports can easily be extended to CCTV mapping and tracking movements of people even beyond the place data was collected. Regulations are required to address the issues at a larger level rather than just individually taking ‘I Agree’ consent in small prints from uninformed citizens.
“Facial data gathered from specific demographics can be used to train FR algorithms of Artificial Neural Network for ethnicity-based FR accuracy. The social media networks transfer huge amounts of personal information about Indians to servers in the US. This includes individual names, location to their activity online.” “Are we unknowingly allowing our citizens to be part of someone else’s mass surveillance programme?” he asks.
Irish data protection commissioner has objected to Facebook’s method of transferring data from the EU to the US for business purposes. Such investigations are required to set the legalities for the basis of existing business data transfers, especially for cloud computing.
(c) Fake Video Statements: With the help of AI, the algorithm can take the input photo of a human face and adjust it as per other imagery to create fakes, especially videos. It is possible by AI to make the videos look real by modulating the facial features like adjusting the lip movement, lines around the mouth, chin, cheeks, eye movements, etc. for a natural look. The Standard Contractual Clauses (SCCs) in practice today are not fit to address the vulnerabilities of data mishandling services. These are the kind of efforts required from the top end to drive AI regulations since it is not part of AI Industry’s role.
In the words of the AI expert, with Artificial Intelligence activities like FR unwittingly being increased in our daily life, the ethics of the use of ‘people targeting’ has become more covert. This Human-Technology (HUMTEC) evolutionary process is at a serious juncture at this stage since governments too are unable to take a top-down view with technology moving fast and intelligently. “Interestingly, few important private players have come forward to fill in this gap but looks like here a wolf has been set to watch the sheep. It is impossible to believe these private players for ethics leadership since it risks being clouded and individualistic. When cornered by Governmental committees, private players usually cop a plea like in the case of “Report on the Investigation into Russian Interference in the 2016 Presidential Election” conducted by the US, he points out.
What is the Way Forward?
Recently, the government has taken very clear measures for AI implementation in India with agencies like Niti Aayog, AI Task Force, etc. spearheading the task. The domains of focus has been Manufacturing, Fintech, Healthcare, Agriculture/Food Processing, Education, Security, Environment, Retail/Customer Engagement etc. covering almost all major aspects of national policies.
Unfortunately, the underlying factor of ethics which is considered as the common thread running through each topic has not been addressed adequately. AI is not like implementation of regular initiatives since AI is a concept evolving on niche technology, that too by market groups who have pumped in billions of dollar in R&D and are looking now for fast returns. In his view, governmental organizations cannot be controlling such groups and therefore, Regulations in AI are of utmost importance.
The government may do well by first addressing the Regulations part of AI rather than jumping to implementing AI applications. “Simple regulations like the Apps to process images locally on smart devices (mobiles, Tablet) instead of sending them to the cloud server located in another country for analysis or, the information of measurements of features on a person’s face can be extracted while the actual photo deleted later,” he suggests.
Maybe industry requires some more technology advancements like higher capability processors, better battery usage, etc. but this should not deter the government from framing the AI Regulations.