Why is Generative AI (GenAI) commanding so much airtime everywhere, from boardrooms to living rooms? Because it changes everything in a big way. Customer service can chat in a way that feels personal and right on point. Creating content isn’t just about staring at a blank page and waiting for inspiration. Teams can do more without the grind of repetitive tasks and get a jumpstart that feels like it’s halfway to finished already.
That’s the kind of shift we’re talking about with GenAI. But is there a dark side to this shiny new tech?
At its core, GenAI’s USP is in understanding and generating new content, whether that’s words, pictures, or even code. We are essentially deploying a super-powered assistant that’s always on, ready to help businesses come up with ideas, solve problems, and serve customers better.
It is easy to imagine a future not too far away where work feels less like work. But are we inadvertently setting off towards a dystopian future in this journey?
The Minefield of Innovation
With a big leap, like the one we’re seeing with GenAI, a whole set of new challenges and complexities pop up. And they touch on everything from ethics to security. Here is a rundown of some such scenarios.
- Bias and fairness: GenAI is trained on mountains of public domain data and has a tendency to mimic the less-than-perfect parts of human thinking. Imagine a scenario where a company rolls out a new GenAI system for screening job applications. It sounds great on paper, but then it starts favoring certain resumes over others, not based on merit, but on biases it learned from the data it was trained on. It’s a tricky situation that shows how AI’s output can sometimes be a mirror reflecting our own societal biases1, often without us even realizing it.
- Data privacy and security: Picture a well-intentioned assistant appointed to make life easier by pulling up personal schedules and emails and accidentally sharing private information in a public setting. That’s a much bigger problem than the one the assistant is deployed to solve. That’s precisely the issue with GenAI systems as well. The systems are capable of inadvertently leaking sensitive personal information (PII) or proprietary business data. A quarter of the organizations2 have already banned the use of GenAI despite its many benefits. The consequences here extend beyond mere privacy violations, potentially leading to legal repercussions and a loss of public trust. We need to keep a tight lid on privacy and security, especially as these technologies creep deeper into our daily lives.
- Malicious use: This is another dark turn on the path of AI innovation. It’s one thing for GenAI to generate helpful content, but what happens when it’s used to create fake news or scam emails so convincing that they are hard to distinguish from the real thing? These scenarios are not hypotheticals. They are real risks that need smart, thoughtful responses.
- Hallucinations and reliability: GenAI, for all its advancements, can sometimes produce misleading information or “hallucinate,” leading to decisions based on incorrect or fabricated data. Even in software development, the reliance on AI for code generation has been found to produce insecure code with a notable frequency3.
A net-positive impact on sustainability
There are many more issues at play with GenAI systems. Security vulnerabilities such as prompt injections, jailbreaks, and extraction attacks open doors for cybercriminals to manipulate these systems4. The copyright and intellectual property infringement, where even well-intentioned use of Gen AI can infringe on third-party rights5.
Navigating this minefield requires a mix of innovation, vigilance, and a hefty dose of ethics. We need to find that sweet spot where we can harness AI’s full potential while keeping an eye out for the pitfalls. We cannot afford to just respond to these challenges as they come. We need to anticipate them, creating a safer, more responsible future for AI.
Crafting the Shield: The Strategy of Responsible AI (RAI)
There is no doubt we need to create and practice a new art of defense in the age of smart machines. How do we do that? The collective wisdom, after years of experimentation and numerous discussions with industry leaders, ethical technologists, and public sector figures, boils down to three immediate steps.
Responsible by Design: Embedding responsibility right from the design phase of the GenAI systems involves integrating ethical considerations, fairness, and transparency throughout the AI lifecycle – from training data preparation to deployment. We must build GenAI with a conscience. Conduct impact assessments to identify vulnerabilities, engage in adversarial testing to uncover hidden flaws, and use tools that automate the inclusion of ethical principles in the development process.
Runtime guardrails: The GenAI value chain is complex. Businesses often don’t start from scratch when creating AI models. They usually get them from various providers, who themselves pull resources from elsewhere. Plus, the risk varies based on how you’re using the AI, what industry you’re in, and who’s using it. So, how do we solve this issue? Start by placing safety checks for what flows in and out of AI systems. Preempt everything from unintended data leaks to potential copyright issues with guardrails, ensuring nothing problematic influences the outcomes.
It is important to set straightforward rules and ensure we play by the book, especially with privacy and data protection laws. It is also wise to have plans ready for unexpected turns, evolving as we encounter new challenges and deepen our understanding of GenAI.
The Governance compass
This cross-pollination of information ensures that decisions are informed by a comprehensive understanding of the entire product lifecycle, from concept to customer feedback and back to the drawing board again.
Loved what you read?
Get practical thought leadership articles on AI and Automation delivered to your inbox
Loved what you read?
Get practical thought leadership articles on AI and Automation delivered to your inbox
After more than a year of diligent experimentation, research, and development, we share insights on how GenAI fits into and dramatically adds value to every stage of the product manufacturing lifecycle.
Treating GenAI governance as a one-off setup isn’t enough. Constant vigilance is necessary to spot and rectify issues swiftly. We can achieve that by regularly monitoring our AI models’ health and usage, ensuring they are performing as intended. With a consistent approach to governance and a commitment to responsible AI, we’re not just aimlessly moving toward the future; we’re making deliberate, informed strides. Our goal is a world where AI not only enhances business but also contributes positively to society at large. Achieving this vision of Responsible AI will take us closer to this goal.
Disclaimer Any opinions, findings, and conclusions or recommendations expressed in this material are those of the author(s) and do not necessarily reflect the views of the respective institutions or funding agencies
- https://www.washingtonpost.com/technology/interactive/2023/ai-generated-images-bias-racism-sexism-stereotypes/
- https://www.cfo.com/news/27-of-organizations-ban-genai-risk-generative-artificial-intelligence/706526/
- https://snyk.io/reports/ai-code-security/
- https://www.theverge.com/2023/11/29/23981725/ai-researchers-talked-chatgpt-into-coughing-up-some-of-its-training-data/
- https://www.nytimes.com/2023/12/27/business/media/new-york-times-open-ai-microsoft-lawsuit.html
