Toronto Declaration

Summary

Preamble

Machine Learning (ML) and Artificial Intelligence (AI) are going to change the society by bringing significant benefits and improvements to the way of life. However, without any checks-and-balance in place, they can pose serious threat to the same way of life by creating agents with biases and undermining rule of law. Toronto Declaration¹ is a huge step the area of providing checks-and-balances. It defines the humans rights standard that the intelligent machines must adhere to so as not infringe upon the basic human rights as declared by The Universal Declaration of Human Rights² adopted by UN in 1948.

Using the framework

The declaration defines a framework that puts obligation to promote, protect and respect human rights on States and private sector actors and at all levels - supranational, state and regional where decisions related to technology research and usage are being taken. It states the obligations of Governments and private actors by putting responsibility to proactively prevent discrimination in order to comply with existing human rights law and standards. The declarations further emphaizes on the importance of promoting diversity and inclusion as the key pillars of protecting and upholding the right to equality and non-discrimination specially against marginalised groups.

Duties of State

As per the declaration, States must ensure that existing mea- sures to prevent against discrimination and other rights harms are updated to take into account and address the risks posed by machine learning technologies. States must identify risks, ensure transparency & accountability and enforce oversight to mitigate and reduce the harms of discrimination from machine learning in public sector systems.

Responsbility of Private Actors

The declaration states that private sector ac- tors have a responsibility to respect human rights; this responsibility exists in- dependently of state obligations. Private sector actors that develop and deploy machine learning systems should follow a human rights due diligence frame- work as defined below to avoid to prevent discrimination: (a) Identify potential discriminatory outcomes (b) Take effective action to prevent and mitigate discrimination and track re- sponses (c) Be transparent about efforts to identify, prevent and mitigate against dis- crimination in machine learning systems.

Right to an Effective Remedy

The right to justice is a vital element of human rights law. Thus, the declaration states that the companies and private sec- tor actors designing and implementing machine learning systems should take action to ensure individuals and groups have access to meaningful, effective remedy and redress.

Trade-offs

Most of the research and innovation happens unhindered when you are trying to solve a problem in hand without thinking about regulations and laws. Discovery of nuclear fission happened because researchers - Otto Hahn, Lise Meitner and Fritz Strassman - were not thinking about the consequences of their discovery. They did not know that it will lead to the creation of Atom Bomb that will be used in Hiroshima and Nagasaki³. However, it revolutionised the world and our way of being.

If researchers and programmers will have to be bounded by the regulations as Toronto Declarations proposes, it will be an added burden on the researchers and might hinder with solving some of the most critical problems that AI and ML can solve.

However, if the Toronto Declaration is discarded, it might have deleterious im- pact on humanity, if not greater but equivalent to the one caused by atomic bomb. It is excruciatingly terrible to even think about the repurcusions it might have on human rights, free speech, equality and non-discrimination.

Conclusion - My Views

The safety of any new technology often hinges on how it’s regulated. The un- tethered and unchecked machine will have the power to distort the discourse, eventually putting people at risk of being dehumanized. Toronto Declaration is a great step in the direction of setting accountability and various levels of decision making in regards to ML and AI. This puts forward a plan to safeguard human rights and interests. The human race has fought two World Wars in the last cen- tury; the positive thing - they were apparent. The war that machines will wage using ML and AI, will be a proxy war - unclear, unprecedented and unconscionable. It took us the entire feudal age and two World Wars to appreciate human beings and proclaim Universal Declaration of Human Rights (UDHR). We still have countries where the UDHR has not been completely adopted and we are still fighting wars. We still have poverty, hunger and unjustice.

I feel if ML and AI are used effectively they can help deal with most of the prob- lems that we as a human race face. Within the framework of Toronto Declaration, ML and AI can be the best friend, for which the human race have been waiting for a long time.

References