What happens to software jobs with the rise of Generative AI?

It took ChatGPT just 5 days to get 1M users.

Let that sink in - just 5 days. Instagram which held the record for the shortest time to get to 1M downloads took 2.5 months to get there. Just 4 months back, LLMs (large language models) and GPT was a concept so esoteric that many even inside the Data Science community did not understand it. Today, my uncle, who is far removed from the tech world, has an inkling of what ChatGPT is. A good part of my Linkedin feed is full of mind-blowing stuff people are doing with LLMs. And there isn't a single internet-enabled business out there who is not thinking about (or has already implemented) a LLM/GPT strategy in their business model.

All breakthrough technologies are disruptive but what's so striking about LLMs is that it is pervading a very vast variety of industries and at a shocking velocity. I haven't come across a person yet who has tried ChatGPT and has not been able to use it for doing something that impacts their work as of today. That means there is a practical, high-impact use case for anyone using the tech. Very few technologies in the past have had such a wide applicability in real-time, in the past.

And it's not just the pace of adoption of the technology but the further development of the technology itself, which makes it distinct from any other technology trend the world has witnessed in the past. A paper published by a staff member of Anthropic shows, GPT had already surpassed capabilities by summer of 2022 that the expert community thought may not be achieved even by 2024 (there's more interesting stuff in there , give it a read).

There is a lot of talk (and fear) about how LLMs are going to eliminate many jobs. I personally believe that disruptive technologies can negatively impact human jobs and functions in the short term, but in the long term it only leads to value creation and generation of wealth, not only for the few elite driving the technology but also the masses. For all the vices of industrial evolution (impact on environment, more extreme concentration of wealth), it did usher in a lot of masses into the middle class, ultimately improving the quality of life for them and the future generations.

So while LLMs (and other future AI tech) will create disruptive changes in the workforce - many operational jobs will become redundant and skill requirements will change significantly, in the long run it will lead to automation of much of the menial work hence increasing productivity per person and freeing up humans to focus on meatier problems..

Of all the jobs, the impact it is going to have on software development is probably the most interesting. As we speak,? 60-70% of coding work can be done by a generative AI assistant tool like Github Copilot or AWS Code Whisperer (AWS claims developers are able to complete their tasks 57% faster with AI assistant). This is because the logical and mathematical structure of programming languages lends even more beautifully to how large language models work. Pattern occurrences are far more predictable in a structure coding language than a natural language. Hence it’s possible to generate a large majority of your code, as long as the right prompts are being given to the system. Naturally, the structure and architect of the solution has to be written by the software developer, but the majority of the underlying code can be generated.

So if Gen AI can generate a lot of the code, how does a developer's job change

• There will be a massive increase in productivity - One can cut down significant time that is today spent on just writing code. While the solution still needs to be architected, broken down into logical steps and then stitched together to produce results, the actual job of writing code can go away and one can focus on just problem solving.
• It will lead to all round Improvement in the quality of code - People often know what is to be done, they might not always know the most optimised way of writing that code. With Gen AI, one needs to focus on the efficiency of the solution, the efficiency of the code gets taken care of. Another example is - writing test cases becomes so much easier. More test coverage leads to a better quality system.
• Developers will need to get better at testing - If 80% of your code is going to be generated, one must get really good at comprehending code and testing it.

So what impact will it have on developer jobs?

• Vanilla coding skills will get commoditized -? If 80% of your code is generated, that means, a significant part of your skill has lost value. This may lead to a fall in wages in many software jobs and elimination in some software jobs.
• Value of architecting and solutioning will go through the roof - The value of those who can solution problems at a fundamental level and build well architected solutions will grow exponentially. Someone with the right skills can single-handedly do the job of a small team.
• Prompt Engineering will become a key skill - LLMs are great at producing outputs when the input is right. Even today, most code generating tools can spit out the code, if the prompt is framed the right way. It will not be an easy skill to develop, it requires real first principle thinking and problem-solving capability.

And that begets a question, what impact will it have on the industry at large?

• There will be an explosion of apps and softwares - small teams will be able to do a lot more, very quickly. That means more people will become creators.
• Non-developers will start generating code - One can totally imagine a scenario where a PM is able to come up with a basic version of an application without the need of a software developer. While real world systems will still heavily depend on solid software engineering skills, prototyping or writing code for well-bounded tasks will become a trade craft for non-developers.
• Software Development as trade may change - Instead of writing software, the trade itself may go towards doing just 3 things a) Architecting the solution and prompting for code? b) Reviewing and fixing generated code and c) Furthering development of LLMs and other AI tech ??

If you are a software developer and you are not already leveraging generative AI in a meaningful way, you better get on the bandwagon. And while your job will not be eliminated, you definitely need to evolve your skill set.?