If most businesses live by the “Move fast and break things” maxim, Zerodha’s philosophy would be to “Pause and consider.” Granted that stock trading platforms aren’t normally looking to cause disruptions in tech, a lot of its steadfastness comes from Kailash Nadh, CTO of the platform. Nadh, who leads a lean team of 30 software engineers, isn’t one to rise to the bait of hype. Having the vantage point of someone who has been coding for around 24 years himself, Nadh is wary of the ebbs and flows of the hypecycle and its symptoms.
Aside from having a perspective on tech within the human context, his heft of experience has also given Nadh the chance to back passion projects. A co-founder of the non-profit FOSS or (Free and Open Source Software) United, Nadh has been an open-source enthusiast for a while. Zerodha’s engineering runs on open-source and saves the fintech company more than $10 million every year.
We talk to Nadh about the criticisms against open-source, generative AI hype and his hope for tech.
Poulomi Chatterjee: Aside from open-source models, there is a growing need for transparency around the details of how a model was trained, what kind of data was used etc. For example, Meta AI’s Llama 2 model which is open-source was found lacking in transparency. What is the difference between the two?
Kailash Nadh: Legally, open-source means making your source code readable for all humans so that it may be distributed and modified. So, being transparent about your model isn’t an ask in open-source. Or an open-source model doesn’t necessarily translate into being transparent in that sense. Although by nature, open-source code is transparent because you can take a look at the source code and the motivations behind the making the model become obvious from just the numbers.
But seeking transparency is a moral demand and something that will generally be good for society. The two terms shouldn’t be conflated.
(For top technology news of the day, subscribe to our tech newsletter Today’s Cache)
PC: There is still a section within technology that considers open-source software a bigger risk and more vulnerable because it is accessible to everyone. Is this a fair judgment?
KN: I believe in 2023, this is an urban myth. There is enough data to suggest otherwise. Complex systems will always have bugs depending on the quality of people working on it and the knowledge they have. Open-source software isn’t inherently safer because technically all software is prone to vulnerabilities varying on these factors. But it is safer because it has countless eyes looking at it all the time. The collaboration and scrutiny that OSS is under because so many people are invested in it is a virtue that makes it safer. And there are mountains of evidence in this day to prove this. If a piece of proprietary software has a bug, you need to reach out to the vendor who then take their own sweet time to fix it. This is exactly why there’s a large number of internet businesses like Meta’s Facebook or Zerodha running entirely on open-source software.
PC: There were reports that ChatGPT can trade better than humans. How do you look at the generative AI phenomenon and to what extent does Zerodha plan to implement it?
KN: Anytime there’s a frenzy or a hype, everybody just gives in to it - it was Big Data sometime back or blockchain before this but with generative AI its different. There is significant substance underneath the hype this time and it is a genuine breakthrough. But no, we don’t have any financial applications around ChatGPT. Zerodha is generally never in a rush to follow a trend unless it is objectively the right tool for the right use case.
I do use ChatGPT minimally - it is a great language model. My Google searches have dropped by 90-95% because this is more efficient. And the tech team uses it heavily for coding assistance.
PC: Is there worth still to becoming a software engineer since AI tools have becoming so good at coding? Will generalists still be valuable now that we have new fads like prompt engineering job roles paying big paychecks?
KN: Software engineering will absolutely stay valuable. But it will depend on the skill level. Tasks that need a lower skill level like patching bugs, maintenance can be easily done by AI tools now so you have to acquire new skills and level up. We are far from AI systems completely consuming all jobs but the cat is definitely out of the bag. It’s just been 12 months and there’s so much change already another few months and there will be another AI tool.
And these massive salaries are a product of the hype - you could come across a position for an ML engineer who gets paid $1 million per year but that is just 0.00001% of the opportunities. Tomorrow, there will be a million more ML engineers. Remember, when the startup boom happened and they were paying unreasonable amounts of money to people but that changed soon. This will also fade.
PC: A few days back, Andreessen Horowitz published a ‘Techno-Optimist Manifesto’ where it listed pretty much anyone questioning trust and safety around tech or sustainability issues as “enemies.” How optimistic are you about tech and how do you feel about their post?
KN: I am not a fan of Andreessen Horowitz and I refused to read the thing. But I would consider myself a tech pragmatist.
Maybe a better question to ask would be how optimistic are you about the future of humanity because the impact of tech will vary according to the framework of the people using it. I am not very optimistic about our future. Humans are far too young and intelligent a species to have brought about such drastic changes. Look at what we’ve turned the earth into in just 100-150 years. I am not sure how long it can go on this way.