I joined IBM in 2006. Back then, people would talk about programming languages – Java, C++, PHP, SQL etc. Each of them apparently having a usable life of at least a decade. A student could just be a Java expert and hope to contribute massively to the industry for at least 5 years, as a programmer. An expert coder would be sought after and much in demand. But the pace at which technology has evolved has changed things. A lot.
Today, the perceived value of a programming language evolves
much faster, with newer languages and frameworks constantly emerging and
reshaping demand. There are some exceptions, like Python, of course. But think
of it like an open source ecosystem supported by millions of developers
worldwide who have ensured it stays relevant, adding new versions and libraries
every single day.
So, does a language
like Python keep developers relevant? Someone might have said yes. But then
came in the large language models (LLMs). Trained on massive datasets, the
ability to predict next possible token based on input tokens suddenly became a
game changer. Without a true understanding of how systems work in a human
sense, and instead relying on patterns learned from massive datasets, suddenly
the models could write content, create images as well as generate code! No
language appears to be too difficult anymore, including Python. Suddenly it
became less important to memorize syntax and more important to be able to come
up with innovative ideas to solve business problems.
Vibe coding became the new normal. What could take months to
develop suddenly could now sometimes be done in a few hours. Suddenly it became
important to communicate better in written languages, than understanding the
syntax of programming languages.
But at what cost? Developers started to care less about
learning the basics and are more intent on taking shortcuts to solve business
problems. While development became faster, in case of any bugs, troubleshooting
became a nightmare. But then that also changed fast, as soon as AI came to the
rescue and started to help fix the problems as well. All the while humongous
data centers kept chugging in at full steam, requiring significant amounts of
electricity and water to power computation and sustain cooling systems. In
large hyperscale facilities, cooling alone can consume millions of gallons of
water per day.
So, somewhere, we are looking at a few things –
1.
Developers and technologists now have the magic
wand AI that enables rapid productivity gains, as major business problems get
solved
2.
But human insights are still essential as AI
still makes mistakes and does not understand business problems in a “human
sense”, but rather models responses based on statistical patterns.
3.
While AI promises super high productivity and
lesser reliance on human expertise for traditional coding jobs, thus also
reducing costs, the data centers powering AI are massive and comes with
significant infrastructure and resource costs, particularly in terms of energy
and cooling demands.
The question is – are we missing the balance somewhere?
