How I got interested in Artificial Intelligence

My journey into building an AI/ML company

Photo by Jacques Bopp on Unsplash

I would not call myself an expert in Artificial Intelligence, but my journey in this area spans a longer time frame than most.

Early days: Undergraduate Program in Pune, India

In 1995, when I was in the third year of my Computer Engineering program at Pune Institute of Computer Technology, we had the option of choosing an elective for our final year.

Artificial Intelligence was one of the electives.

I have always been a forward-leaning person regarding technology adoption, and I picked the coursework because it sounded “sexier” than compilers :-).

Although I was somewhat disappointed by the state of the art, the promise that computers would magically do grunt work for humans fascinated me.

Master’s program or how I almost didn’t land in the US because of AI?

When the opportunity to do a Master's in CS presented itself, I picked the University of Cincinnati, where Professor Raj Bhatnagar specialized in Data Mining, Pattern Recognition, and AI.

The year was 1998, and India had just concluded the Pokhran nuclear tests. The sounds of sanctions were in the air. When he heard that I wanted to specialize in AI, the US visa officer told me to return home because he needed time to decide (while he kept my passport with him!!!).

The visa officer was thinking about the implications of AI x Nuclear Tests!

I was in limbo for two months before my visa was granted. It was a somewhat stressful period.

Once in Cincinnati, I took a few AI courses, but now it is too far back in time to remember what they were. I remember writing some trees for homework more than any classical ML pattern recognition work. Prolog, for writing programs that made decisions, was interesting enough but too slow to do anything useful, and I was at a crossroads because I was in complete love with Java.

AI and ethics: Implications of AI being used in war

When the opportunity to pick a master's project came up, I found myself walking into a program for a neighboring Air Force base—this was a military-focused AI research project.

The military aspect made me pause, and I thought deeply and hard about the implications of AI for war (and thought a lot about the visa officer :-) ).

I chose to explore projects with commercial rather than military aspects, so I switched out of the ML lane and landed on a Data Mining project.

The project was fascinating as I wrote Agents (IBM had a technology called Aglets) in which a computer agent would go into different environments) to answer specific questions.

The chosen environments were representative of DMV data and health data (I forget the specific database), and the questions were like, “Do cigarette smokers have a higher propensity to land in accidents?”

The intent with the different environments was that neither DMV nor Health data would be publicly available, but you had negotiated restricted and partial access to data in those environments.

The biggest challenge was getting to any database of a reasonable size to do anything interesting.

I also remember lecturing in a student body venue about how AI could be used to answer web search queries, book tickets, or build travel itineraries, removing humans from the loop and the gruntwork.

After graduating in 2000, I found that there were really no jobs in AI or Machine Learning, and frankly, the Java division of Sun Microsystems was an epic place to land.

From here on, I lost touch with the field.

Reengaging with AI and planting the seeds of Launchable

About 16 years later, I started the Deep Learning Nanodegree coursework on Udacity. I worked on it for six months and did enough assignments to get my feet wet. However, I couldn’t finish the final project because of work pressure and, candidly, because my programming skills had rusted away. Thus, I needed way more time to complete the assignments than I could allocate.

I loved the course and concluded that AI had advanced far enough that the underlying technologies would be game-changers.

I started looking for opportunities to use the technology in the workplace but was way ahead of the adoption curve.

Eventually, I decided that rather than convince people to incorporate AI into the products at the companies I work for, I would build an AI/ML company because it was less of an uphill battle.

I wanted to build the next generation of companies, and that was the birth of Launchable.

Though I haven’t been involved in the coding side of the house, I have enough practical experience building a company and running marketing and sales for a company that is selling an AI solution. This is active “AI Product Management” work, which seems to be the new buzzword.

Lately, I have delved deeper into the tech side of the ML space and have been reading Hands-on ML by O’Reilly.

Over the next few months, I intend to explore the LLM space more deeply. I picked the ActiveLoop Langchain course last year and plan to revisit it because I parked it aside as I decided to delve deeper into the fundamentals of ML.

The end goal remains the same as when I first got into AI in my undergrad program: Incorporating AI into products to remove the grunt work from humans.