AI is not certain, it is probabilistic — Rahul Agarwalla, Managing Partner & investor, SenseAI Ventures

To get deep insights into artificial intelligence and its implications, Pooja Singh had an exclusive interview with Rahul Agarwalla, Managing Partner & investor, of SenseAI Ventures. Continue reading to learn more about how AI is changing the world.

Artificial intelligence has become part and parcel of our lives these days. Professionals from IT to non-IT are widely adopting artificial intelligence for development. Generative AI has changed the landscape of the job industry. To get deep insights into the same, Pooja Singh had an exclusive interview with Rahul Agarwalla, Managing Partner & investor, SenseAI Ventures. Continue reading to learn more about how AI is changing the world.

As evident from the diverse applications of AI across various sectors, its integration has become prevalent. Considering this, how do you envision the functionality of AI in distinct industries, such as healthcare, for instance?
Depending on the specific use case at hand, varying levels of accuracy become imperative. Consider a scenario where decisions hold life-or-death consequences; in such instances, a 100% accuracy rate is paramount. Currently, AI has not reached this level of precision. Therefore, situations necessitating critical judgements are likely to maintain human involvement, even as AI provides recommendations. The realm of AI excels at suggesting courses of action but remains distinct from making definitive decisions.

A telling example lies in skincare, as epitomized by CureSkin, a startup that analyses skin photos, identifies issues, and offers treatment recommendations. Here, a dermatologist reviews the AI’s recommendations before finalising treatment options. Operating at a precision level of about 97%, this collaborative approach blends AI’s insights with human expertise, ensuring accurate decisions while keeping a human in the loop.

In contexts like medical scenarios, the indispensability of human oversight remains evident. Whenever deep expertise is indispensable, the collaborative AI model emerges as the optimal solution. This framework involves AI conducting approximately 90 to 95% of the tasks, with human experts conducting the final checks. The synergy between human intuition and AI’s analytical prowess underscores the efficacy of such collaborative endeavours.

In your opinion, in which sector do you believe AI could be fully and optimally utilised, or where could its application be seamlessly integrated?
Indeed, there are domains where human involvement proves infeasible. Consider the fields of cybersecurity, fraud prevention, and algorithmic trading. These sectors have seamlessly transitioned to full AI integration, even in the present day. It’s important to acknowledge that while AI operates with a certain degree of fallibility — given the absence of 100% accuracy — the advantages it brings far outweigh these inherent imperfections. A tangible example emerges in the context of fraud prevention. While the occasional misidentification of a legitimate transaction as fraudulent is an error, it pales in comparison to the potential consequence of overlooking an actual fraud.

This risk trade-off is crucial and showcases the value of AI in these applications. To illustrate further, many of us encounter the extra step of confirming transactions through OTP or additional verification measures on banking websites. These layers of security are orchestrated by AI. For instance, when an additional confirmation is sought to validate a user, it’s an AI-driven fraud prevention mechanism ensuring that transactions are authorized by the legitimate account holder. The high volume, rapid pace, and complexity of transactions render human intervention impractical, thereby necessitating AI’s autonomous operation.

Numerous sectors reflect a similar pattern, including the imminent advent of self-driving cars. While they’re already navigating California’s roads, achieving widespread adoption necessitates time and meticulous development. The gradual introduction of self-driving cars into India underscores the need for a carefully structured integration process. It’s undeniable that AI has transformative potential across diverse areas, although the pace of its implementation may vary. As self-driving cars demonstrate, widespread adoption hinges on meticulous planning, adaptation, and addressing regional nuances.

The realm of cybersecurity has experienced a significant surge recently, and the prevalence of cyber threats has escalated, with AI emerging as a pivotal player in addressing this issue. What are your thoughts on this evolving landscape?

Today, AI is also being used to attack. Cyberattacks are all generated by AI, and because it’s AI generating the cyberattacks, the attacks are much larger in size and scale, or wider attacks. So, on the defence side, you also need AI. Cybersecurity is a field that will both attack AI defences and be AI itself. It already is basically that, and it will become more and more of that.

Is it plausible to anticipate that in the coming years, the situation concerning cyberattacks might deteriorate further?
As the world becomes more digital, there will be more digital attacks and fewer physical attacks. There’ll always be criminals, and criminals will always try to make money in some way or another. So, that never goes away; we can wish it away, but it never goes away. You will always need technology to defend yourself. Till that moment, we can expect some AI tools that can help us in this way as well to prevent these attacks. And we are very interested in investing in that sector. I’ve had a previous investment in the cyber sector, and we made a good exit from that. And we are very excited about cyber security.

There are companies that have yet to fully embrace AI or haven’t incorporated AI technology into their day-to-day operational activities. In your view, could these companies potentially fall behind in the competitive landscape?
In many fields, absolutely. Either they will adapt or they will die. It’s like saying that there were guys who did not use the internet. Did they lag behind? Yes, they did. Until they started using technology well, they started using the internet. The same applies to AI. It will be used across industries and functions. You are using it today. If you have a smartphone, you have AI, and you are using it. So without AI, why would you want to do it the hard way? It’s like saying that I will use a rotary dial and not a push button. Technology changes, and you will change.

Given the increasing adoption of AI, do you believe there are inherent risks that companies and individuals might encounter? Would you agree that it’s important to exercise caution and take necessary precautions before fully embracing AI?
Absolutely no new technology comes without risk. And you must understand AI, especially Generative AI is a baby right now; it’s just born. So maybe it started to crawl. We’re still very early days, which means a lot needs to be done for it to mature as a sector and as a technology.

So whether it is bias in training data, from racism to sexism, you name it, those problems exist in the data. And our AI models have the same problem because of that. From hallucination, which is a new problem with Gen AI, to privacy and security, which we all face because our data is all being indexed by AI. So these challenges are all there and have to be handled. But beyond this, there are other adoption challenges as well, like the fact that AI is fundamentally a probabilistic technology, which means it’s not a zero-one answer. It’s not yes or no; it’s maybe. Maybe this is a cat, or maybe this is a dog, maybe this is a good thing to do, maybe this is the right medicine; maybe this isn’t the right medicine. Therefore, as mentioned there’s only one answer.

Is it reasonable to have confidence in the accuracy of the responses and outcomes generated by AI, especially when considering aspects like adaptability or generative AI?
We like certainty as humans; we like to do things to some extent. And companies, especially businesses, work on certainty. You have to give your boss a date when he asks you when it will be done. And you have to be accurate. So, we like certainty. But AI is not certain; it is probabilistic, which means that you will always have some level of doubt, some level of variance, which exists, and you have to adapt to it.

You have to learn to think like that. You have to learn to ask the right questions in generative AI. There’s something called prompt engineering, where depending on how you ask the question, you get different answers. And one answer is much better than the other answer. So we also need to learn to ask AI the right questions.

What is your perspective on the trajectory of AI in the coming years, say over the next decade, and how do you anticipate its influence on society?
In my view, AI is very simple. I am a long-term believer in the potential of AI. I think it’s the single largest value-creation opportunity of our lifetimes. The closest parallel is the internet. If you go back and look at the internet in 1996, that’s when the Netscape IPO happened and everybody suddenly realised there was something called the internet. It was still small, but it grew, and the bubble kept inflating until 2000. And in 2000, it was really massive. And then there’s overhype, maybe, and the bubble bursts. But 25 years later, if you look at the world today, internet companies are worth more than all the companies in 1996. That’s how much wealth has been created. AI will do something similar. So to us, it’s a multi-decade thing. For the next 30 years, we’ll be investing in AI startups. Today, we can call ourselves an AI fund because there are not so many people who understand or invest in AI. That’s why we call ourselves an internet fund. Hence, this is one change that we are having.

Do you believe there’s a possibility that AI could lead to a reduction in job opportunities for individuals? While there is potential for efficiency gains, do you think this trend might result in a net loss of jobs, even though new roles might also emerge?

Yes, jobs will change; jobs have always changed. If you go back to the Industrial Revolution, to electricity, to the engine, to the locomotives — all of them — there are no people taking you in horse carriages around the city anymore. They all lost their jobs, but new ones were born. Now that you have taxi drivers, don’t think of it as a net loss or net gain. If you have net wealth creation, all the people will do better. So look at wealth creation as the outcome you want, rather than job creation. We’ll only know in the future. But my guess would be that it will create jobs. But it is immaterial.

Do you personally make use of any generative AI tools?
We’ve used, as I said, human-in-the-loop kinds of processes. From writing our job descriptions when I was hiring a team at Sense AI to some portions of our website, we use Gen AI. For mailers and newsletters, we’ve been using AI tools. But we always have a human in the loop. We also have a unique voice and a unique thought process. So that can’t be generated automatically. So that’s the thing we can say, We cannot depend on AI because human intervention is definitely required.

Visit TechGig: https://bit.ly/3ijY5Gt