Ethical considerations in the Generative AI era

Writing a poem and looking for creative inspiration? Understand what complex physics or cloud computing jargon means? Translate huge mounds of data into legible and logical information? All these are the key use cases for Generative AI from a layperson perspective. This ability to mimic human content is revolutionary and can be a powerful tool that essentially can transform multiple sectors and industries.

At the basic level, it has algorithms using machine learning to go through mounds of data and offer unique insights, often backed by logical arguments. In multiple sectors, the rise of these tools is changing the workspace and potentially transforming how we work, play and relax. Many believe it will help automate dull repetitive tasks and thus leave time for humans to let their creative instincts prosper and drive professional and organisational growth.

Sharpening decision making

As mentioned earlier, AI systems can analyse vast datasets, recognise patterns and predict outcomes. It will in due course help in setting business goals, and could contribute to supplementing human calls with data rich analysis quickly.

All this is great news. After all, who does not seek a tool that improves the quality of work and frees them up to innovate and learn. It's been bagging investments and is the biggest tech innovation of the 20s.

Amidst all the cheer and general optimism, it is important to address the big question of how generative AI should deal with ethics and what can be done to minimize misuse.  The argument is not new and has been spoken about by tech tycoons and researchers in AI firms as well.

In the hands of bad actors, this technology could potentially unleash disinformation and 'hallucinate' instead of providing factual responses and solutions.

Moreover, these tools could throw up false information and can be weaponised by unscrupulous elements to spread misinformation. AI models can also sweep copyrighted content or unauthorised data and publish it as legitimate information.

This could fall afoul of multiple laws, including privacy rights and breaches of intellectual property laws. The outcomes generated by a generative AI model is influenced by the key data points on which it is trained. This output can be skewed or discriminatory, if this data is not diverse or unbiased. For instance, a model trained mostly on data from a specific demographic may produce inappropriate outputs for others.

For instance, it could state that people of colour are more likely to be involved in crime or throw up erroneous results on diversity in top jobs. A report in Wired spoke about how AI had impacted the lives of three innocent people.  

In 2020, the use of AI to predict A-level exam grades in the UK demonstrated a biased algorithm that skewed results to favour students from affluent backgrounds, contributing to educational inequalities. Facial recognition algorithms have misidentified individuals, leading to wrongful arrests, particularly in minority groups. These issues have meant governments and regulators are looking at working on regulation to prevent AI misuse.

Fixing it up

There is a need to fix these issues and harness generative AI for good. Companies should start educating employees and conduct continuous testing and benchmarking to identify and rectify disproportionate errors. They should train employees on the relevance and need for ethical practices, the risks involved and responsible data use. Firming up data security with tools such as encryption and anonymisation will be helpful too.

Skepticism can put the breaks on the spread of misleading information since users can authenticate their outputs independently and maintain trust.

In addition, one should prioritise collaboration between AI and human workers. A McKinsey report agrees with this and has highlighted the importance of getting a human in the mix, and argues companies, regulators, and trainers must bring in more collaboration.  A commitment to transparency in AI development with third-party audits could aid in promoting ethical AI usage.

What next?

Developers can also mitigate biases and promote fair outcomes by incorporating diverse data. They should understand the importance of transparency and accountability.

Using good monitoring tools helps identify and address bias or ethical concerns as and when they arise. Generative AI will transform industries and improve aspects of human life. Its rise should be tempered with ensuring more diversity, privacy, accountability and transparency. By adopting responsible practices, encouraging collaboration, and continuously refining our strategies, we can ensure generative AI is a force for positive change, and unlock its true potential.

Author: Arjun Prasad, Co-founder & Chief Strategy Officer, QX Lab AI