It was only a few years ago that business AI ethics became a constant headline for the very first time. In December 2020, Timnit Gebru was fired by Google. Even though there are conflicting accounts of the event, this termination pushed forward the debate around AI ethics, a debate that was somewhat sidelined in the tech world before this incident happened. To get an idea of how AI will develop in 2021 and beyond, inputs were taken from the AI ethics leaders regarding the opportunities and challenges in the field. Here’s taking a look at what the top AI experts think about the ethical aspects of AI development.
For quite some time, policymakers were dependent on the stories told by the corporations, research papers, and media, and portrayed the image of how effectively the AI systems are working. However, AI expert Deborah Raji thinks that as these systems enter into the real-world systems and are put to use, it then becomes apparent how they are failing that can harm those who are affected.
There is also a need for auditing and accountability. Auditing independently from third-party actors like regulators, consultants, and internal teams to actually get some sort of external systems to analyze and challenge the stories told by the companies creating them. Not only that, these different actors are working in isolation and not synchronized which results in inconsistency as to how these auditors and algorithmic auditors are currently working.
According to Rumman, an expert in the field of applied algorithmic ethics, there should be an enhancement in the algorithmic choices for users. This means that there should be more control, understanding, and clarity in terms of computational systems that rule this experience such as making features more reachable so that people learn how this works or making product changes where people can customize it and let them modify according to their needs. In terms of implementation, algorithmic choices give difficult choices since there is no clear meaning of “to start.”
Abhishek, Founder and Principal Researcher at the Montreal Air Ethics Institute, believes that AI ethics will struggle in regards to the formalization of the process of conducting audits for bias and other ethical issues. This formalization is necessary for solving the problems regarding lack of regularity in the auditing guidelines and consistency in their application, making results more relevant.
The challenge with this will be the need to rectify various proposed rules and resourcing from a proficient and financial resources point of view.
Christine, another prominent AI expert thinks that the biggest advancement and a challenge will be the interest in and money spent on everything related to formulating research questions, designs, and system development, where not only direct users of this are thought about but also the non-users who will be impacted. Technology would be able to inform us about the bad situations as well and not only things it could do for us. There is a lot we are not aware of regarding social interactions in technology that can make participatory design challenging for us.
R. David Edelman
Edelman, an American policymaker and academic, thinks that AI ethics will make its way into a firm AI policy. Due to the increasing demand for AI across the world, we have moved past the uncontrollable questions and began using law and rules to train AI’s faults. With the right security, we can make use of it in the field of finance, medicine, education, and beyond. We need to ensure long-lasting standards that can be applied to the use of this technology.
The biggest risk will be utilizing the time focusing on the problem instead of using that energy and constructing the technology and tools needed to protect that is important.
Dobrin, Vice President and Chief Data Officer for IBM Analytics, is of the opinion that there will be a shift towards human-centered AI where the person is either going to be affected by AI or will use it in some way. This entails going back from looking at the output to the data instead so that the privacy is maintained and the data is collected with the required consent.
In terms of challenge, well-thought accurate regulation is needed. By this, he means that there is not only general regulation of AI which particularly use cases or results are supervised.
When it comes to analyzing the major challenge in AI ethics, it would be the flourishing industry of deepfake, which is the real threat lurking in the dark.