1) Can you tell us a little about your role and what are your key responsibilities as a director of machine learning in Ikigai?

To be precise, machine learning and AI are one of my key responsibilities. I work with Ikigai enabling AI for everyone who might not know how to code. People might be interested in using artificial intelligence even when their workflows don’t include coding skills.

I work on developing machine learning solutions that companies can utilize in their workflows. For this, I work with an entire team that creates new machine-learning solution algorithms. Apart from this, Ikigai has something called an Ikigai Academy where we teach professionals how artificial intelligence can improve their work. We have a number of courses, and I lead the side in developing the courses and everything related to academics.

2) What is the most challenging part of your job?

I would say understanding the requirements of our clients is the most challenging part of our job. For instance, to determine what new customers want and enable them from the technical side considering sometimes the requirements are genuinely complex. Also sometimes we are working with low amount of data, however, it is also fascinating to work with such complexities at times. Currently, we are working with MIT founded startup so our job can be both challenging as well as interesting.

3) Considering that AI has been a buzzword for a while how do you see it evolving?

With new developments like ChatGPT, AI has become popular and has been in the discourse. It is something that’s in everybody’s mind space.

The way I see AI evolving is that it is becoming more of a companion to us. We are right now at the sweet spot where things are changing in the domain. For example, now AI is helping us write emails.

In fact, two years back, there was some level of AI whenever we used to type in the mobile keyboard with predictive words. Also, it had the capability to mend current mistakes in typing. In fact, you could have incremented entire phrases and sentences with autocomplete which was a wow factor.

Now with ChatGPT, you can have entire paragraphs and suddenly it hits people that AI can do more. There has always been a little amount of machine learning in many techniques for instance cleaning images in Photoshop and then came Dali.

Think about this, generating images just by a description that appears like magic. And, yes there is some level of magic involved considering it's reasonably complex for a person to do. Although after that people start thinking can it replace an artist or a writer? So, the way I see AI evolving is continuing to make these kinds of advances. Advances that we traditionally thought would be done by some person. However, AI will be a companion for doctors in diagnosing people, it will help in law enforcement, loan sanctions, etc. Gradually, it will be omnipresent and will continue to do so.

4) How do you think CXOs can adopt data analytics and AI technologies?

It is a necessity for CXOs today. If they don’t adopt data analytics and AI technology, they’d be lagging behind in the competition. In terms of adoption, well, AI works on the backbone of data. The more the data, the better will be your predictions. Second would be the type of data you collect to feed your AI model because it determines what you’re interested in. Therefore, it is essential you collect a vast range of data and have an aim to optimize it for a given metric.

 Adding to it, suppose you want a large number of viewers, therefore, you need to define the kind of users you want and not just have clickbait views. You have to define what it means when people click on your article or watch your video. Are they completely watching it or watching 80% of it would be considered a view? As long as you are following this, you are successfully using AI. People will give you strategies, however, it is you who needs to define metrics to understand the kind of people you want in the store or on your website.

5) How can data analytics and AI contribute to improving customer experience, especially for startups?

AI is helpful in this department. Consider companies like Netflix, Youtube, etc. that have a large user base and are continuously thinking about how to improve their experience. With AI, you can look at the core and decide which type of content keeps the user happy. In fact, you can flow downstream from one metric to another.

Today, websites seek more opportunities to show content and advertisements. AI can help in improving user experience in two ways. For instance, AI excels in finding rare things. Let’s say in the olden days, you would have critics or experts who would watch a movie or it would be recommended by someone you know.

With AI, you don’t depend on these small layers of people. For instance, these people could only watch 10 or 15 movies in a week. However, at this point, we literally have thousands of hours of video uploaded every second.

Companies like Star Movies or Netflix continuously look for what people are watching or listening to understand what user wants. For instance, every time you are watching a video on Youtube, there is a small program running in the background. This program helps them understand what people want. For example, if you watched a video for a few seconds and turned it off, this would mark it as negative feedback. On the contrary, if you watched it twice or shared it, this would be a strong positive. These small interactions lead up to this magic. The system is not listening to your microphone rather it's collecting data from your past. The content that you have viewed and based on that comes up with recommendations.

6) Can you mention some cutting-edge applications of AI and data analytics for various industries?

Talking about cutting-edge applications, there’s a bunch for instance LLM (large language models) like ChatGPT. It is an application that has gone from zero to 100 million users pretty fast.

Now, companies are looking forward to fine-tuning these large language models for their own company. They have a large corpus of documents that is internal that can be used to train models instead of the World Wide Web. This would enable them to answer people’s questions. It is a deed that is being done by a number of firms.

The second would be explainable AI. Companies are concerned that the responses generated by these AI are simply inscrutable.

We don’t always understand why something was done the way it is, and in many cases, it may not particularly matter. However, you’re shown one movie or something else that makes a difference. For instance, when someone’s loan is approved or denied, AI can decide the tenure of imprisonment based on the crime. There are people who would like to understand how a decision was taken. We can make AI explain itself and get behind the decision-making and make sure that we do not take any information into account that wasn’t supposed to be taken. Also, AI doesn’t discriminate against genders or races.

Another part of this is from a viewpoint of privacy for things like federated learning. We can train machine learning systems with large amounts of data. Therefore, every time you’re using the system, you are giving large amounts of data that any third party can store or sell.  People won’t be willing to reap the benefits of AI while giving up a lot of privacy. This is where things like federated learning come which reduces the amount or dilutes the amount of information.

Another part is to make AI efficient. For example, it is known that training a model like ChatGPT takes up massive amounts of resources. These things can be handled more efficiently while we can worry about things like climate change. Considering, all these systems take large amounts of power, we can use the requirements to reduce compute requirements making things more efficient.

Another part of this would be Edge Analytics. This is when we collect data from a device and compute as much as possible on the device, thereby reducing the amount of data that is transferred to a centralized service for storage. This makes the compute close to the user rather than the company that is specifically targeting a segment.

7) What role do data privacy and security play in the adoption of data analytics and AI? Also, what are the measures that should be taken to protect sensitive information?

Based on the information collected, every company needs to take its own stance on privacy and security. It is tempting to collect as much information as possible for as long as possible. There are multiple things that come into the picture and the data becomes outdated. This data may or may not be relevant to your models. Therefore, it is important to decide to what length you wish to store it. At what point, it would be useful for making your predictions? There are other things besides just the modeling or AI part for instance legal considerations for storing the data or deleting it. The company needs to consider this for minimizing risks of any kind such as data leakage, data breach, and so on.

One simply does not collect the data with a number of anonymization techniques. For example, hashing sensitive or removing any identifiable information or PII (personally identifiable information) from the data set. You might still be able to work with large amounts of data and be effective with your AI. In case you work with fundamentally sensitive data for example if a company work with medical data, there’s no way to avoid using PII.

One simply needs to have a secure data practice. When you feed the data into the machine learning model make sure that PII data is as little as possible. What we have done in Ikigai is instead of using training mode, we have used synthetic data. For instance, if you have a set of banking customers that you want to train a particular model for. You could simulate those customers in order to generate synthetic data. The data is similar to an actual customer but does not belong to actual customers. This protects sensitive customers from breaches.

8) How do you see the future of data analytics and AI evolving? Also, what implications does it have for CXOs and tech enthusiasts?

Well as I said earlier, I see it becoming a part of our lives without us noticing it. In fact, it is already a part in many ways. To further add to this thought, today Netflix or Youtube recommend us a particular video to see, however, we will be at a point where AI will generate videos on the fly that matches our taste. Content that no one else has seen before and might not ever see it.

AI is making an impact when it's combined with robotics. We already know about automated cars and we spend a lot of time driving ourselves getting frustrated with other drivers. So, this AI would take over, the moment it finds a clean patch while running on highways. In fact, there will be cars where you simply have to enter the location, and it will take you from point A to B. I mean, I see a lot of opportunities in terms of making things efficient.

Also, the thought that AI will lead to job losses can be true in a small proportion but consider what happened 30 years back. The time when the Internet was becoming big and people thought that it would end brick and mortar stores. Retailers would be out of business. On the contrary, it led to a massive explosion in the number of jobs available in an entirely new category that was not even thought of. Even, if AI does replace some level of decision-making, people will find other levels that are important for them.I believe it would lead to an overall improvement in our lives for everybody.

9) Are there any last words that would like to share with our viewers?

Yeah sure! So, we see a lot of hype in this sector. However, not everything that comes out of AI is a force for change. We are right now in a stage where AI isn’t’ fully settled, however, it is going to lead to some level of disruption. For example, there are concerns with systems like ChatGPT like hallucinations where a system dreams up facts that can be used to generate false articles and false pictures.

There is a lot of work going on enabling systems to figure out whether something is true or false. We are trying to mark if something is from ChatGPT or written by a human. These are things, we should be aware of. There are a lot of things that need to be done. For example, there’s a lot of work left related to learning from small data sets and things like that.

Think about when you are training a computer vision system, you literally need to give it tens and thousands of examples before it learns about a particular image and classifies it. When a human baby looks at it, it can figure it out simply by looking at a single example. So, we have to get to a point where a system can learn from small amounts of data and I think a lot of interesting things are still remaining to be done and it's going to be an exciting place.

Want to connect with Farhat Habib, and follow his journey? Check out his Linkedin. To get notified of similar interviews, subscribe to MobileAppDaily. To read more interviews similar to this, click here.