GenAI is a revolution that needs governance
Day5 of #100daysofGenAI series
So, it all started with the launch of ChatGPT — the power of AI (GenAI to be specific) was to be seen everywhere. Don’t go with my statement, but the data says it all. Its launch attracted 1M users within a matter of 5 days.
That in itself might not tell much until you compare it with how long did it take Netflix to achieve the same scale back in 1999.
Bottomline is that it wowed everyone until it didn't.
Reason?
Simple, with great power comes greater responsibility.
Undoubtedly, GenAI is powerful and has the potential to spark innovation in unimaginable ways that can lead to a better world. The question is who would bring that change?
Well, there are good actors — the ones who can leverage these unprecedented advancements in technology to bring about positive change.
But, like two sides of the coin — the world is not limited to good actors. You get it right.
As soon as the GenAI offerings and promises started to show, it attracted the bad actors as well, which inevitably, led the “wow” moment to convert to a “fear” saga.
That's where governance is needed to ensure the right balance between fostering innovation and saving it from getting into the wrong hands.
Whenever you hear any AI expert, you’d frequently hear them say something along the lines of, “Well, GenAI has the potential to do …, but”.
The word “potential” is the key. It has the potential to be used in much-needed areas, sectors, and applications, such as healthcare, education, a cleaner environment, etc.
But, who would convert this “potential” into “reality”?
We may not know the answer yet, and frankly speaking, are we even interested in knowing this as a primary goal?
We should be more concerned about saving the status quo by preventing this powerful tool from being misused. That way, before we start pinning our hopes on a brighter future, we’d be making sure that we are saved from the harm that could come with it.
The world order can be put into chaos if we leave it unattended, so I hope I have convinced you that governance is crucial to write the success story of GenAI.
Great, so what's next? How do we bring governance? Or, for some folks, the word itself might need some backing up.
Quoting the descriptive definition by Allan Dafoe, governance is the process by which decisions are made.
Allan’s stance is thought-provoking. He also highlights how strategy, regulations, norms, policies, etc. all essentially come under the umbrella term — governance.
He then goes on to share the normative definition which includes a good set of such processes. Note that “good” could be very subjective and open for everyone’s interpretation, that's where he succinctly describes it as one that is “effective, legitimate, inclusive and adaptive”.
Why afraid of governance?
The very reason that makes GenAI appealing is quickly becoming the pain to regulate it — it's evolving creative nature.
Another key concern is that it transcends boundaries, which means that no single group or community can ensure its effective use. It calls for a global task force including a multi-disciplinary committee consisting of policymakers, legal and compliance experts, regulators, industry, and academia. And on top of it all, the governments.
This process in itself is a complex and challenging step to start with, which I will explain in my future posts. But for now, let us assume that everyone has agreed to join forces and come together. But to do what?
- Clearly call out what is acceptable vs not, which in other words, means creating guidelines and guardrails.
- Once the boundaries are established, there needs to be a protocol on repercussions because no one will regulate by themselves unless the consequences are written in bold and made loud and clear.
- Rule out the negatives by defining them and share the path of what will happen if a violation takes place. Now, cut to the point of how would one evaluate whether compliance is truly happening at ground level. It needs some checks such as — is the system doing what it has been originally designed for, which in turn calls for transparent systems. Transparency by its very nature extends to accountability which is what these governance measures would hold the wrongdoers to. Inevitably, transparency and accountability together foster trust in the system.
- Lastly, despite all the above checks, just like how machine learning algorithms are probabilistic, there is still a possibility that governance alone might not be sufficient in the early days. In that case, it should be distilled through the layers of risk to ensure that AI-powered systems are assisted with a human-in-the-loop. This works two ways — firstly, it builds trust when supported with human oversight. Secondly, it prevents the high-blast radius where the damage could happen at a larger scale.
I have just shared the baby steps towards GenAI governance, but the path to its actualization is long, that can not be covered in just one post.
Remember that dialogues are happening and developments are taking place as we speak.
It is not just limited to the who’s who of the AI world, but involves all of us, that is the reason I am creating awareness on this topic.
I have used the word ‘harm’ multiple times throughout the post and will dive deeper into all possible harms the world has witnessed until now (knowing that there might be a lot that is unforeseen), in my next one.
Watch out for this space to bring the transformation in the most ethical and responsible way possible.
And yes, tomorrow, I will discuss one of the latest open-sourced models and how it gives us a preview of responsibly developed AI.
— — — — — — — — — — — — — — — — — — — — — — — — — — — — — — — — — — — — -
Hi 👋 , I am Vidhi Chugh, working on transformative and trustworthy AI solutions.
I am on a #100dayjourney to share the key milestones, developments, and insights in the #genai space.
👉 Follow my daily GenAI bytes on the below-mentioned platforms to stay informed. 📬#generativeAI #AI #AIstrategy #100daysofgenai
Stay tuned on the following platforms for further updates:
