Recently, I have found it pretty overwhelming the pace at which our industry has been changing.
What was once just a hobby for me, an expression of mathematics of which I was pretty unaware of for the first 20 odd years of my life (having loved maths and statistics from a very young age), I loved exploring the simplest of computer science topics. How an image is made up of pixels and numbers, how they can be transformed, how different pieces of everyday life can be represented by a computer:
- UTF-8 representations Emojis (thanks Harvard CS50)
- Graph representations of groups, running mathematically neat graph algorithms on them
- Databases that efficiently model complex relationships for efficient retrieval
- Training convolutional neural networks to classify digits, or identify cats and dogs.
These are fun things. They are definitely valuable things, but primarily they naturally have a playfulness that appeals to me and got me interested in the first place.
AI, specifically Gen-AI, the talk is all about job replacement, supreme overlords, human employees becoming extinct, greedy tech CEOs. This is not the fun application of mathematics I first got interested in that inspired my career switch. But I do believe the problem is being grossly overstated and let to run rife on social media platforms like X and YouTube in particular.
Reflecting on the initial reasons I entered the industry, it puts the developments in a completely different light. I knew that technology had great ability to bring about positive change in the world. I was working on an app with a friend during university that we hoped would greatly limit the waste from fast fashion by storing what you already owned neatly. The alternative career path for me was consulting major life insurance companies on how deaths of their clients that year had increased or decreased their profits. It was clear to me which one I wanted to do long term.
Pondering further, I imagine what it would’ve been like having the current tools back then, I would have done anything for it. As students we were short on time, pretty broke, and didn’t know enough about coding to bring about a large scale production application with large customers. That was 2018. Fast forward to 2026, now the power is in the builders hands. The people who have ideas, and strong opinions about how good software should be built to last for the long term, they are in a fundamentally more powerful position than ever before.
The important thing now is ideas, curating brilliant experiences, bring value to society. Why do we not see this being discussed in our social media? Is it boring and not getting enough clicks? We have issues like climate change, energy transitions, electric vehicles, reducing waste, mental health crises brought about by aforementioned social media apps, growing inequality, heightened division.
Training generative AI is and has been very costly, but I hope that this spike comes down as we have less need to tune and train models once they are developed and “good enough” (7 to 8 times more energy than standard computing workloads - many largely are fit for purpose already. Our attention can then turn to optimising them. Under the hood they are essentially big matrices of weight vectors. Arguably no more painful to host than some large databases and runtime of calls should be minimal as we pass through tokens through LLM functions.
So it is time to turn the conversation to rapid prototyping. To tuning websites for accessibility because we can just get an agent to do it, rather than it sitting at the bottom of a backlog. For those simply looking for work and stable careers, we will still need humans to at the very least navigate this period of change, steer the ship, make sure the correct things are being developed and diagnosing performance bottlenecks of Gen-AI. Ensuring smooth and stable releases. Explaining why and how something has been built in order to diagnose issues. Things are in fact not worse now. In fact it is a great time for builders. However, it is time for us to spend more time pondering what to build, not how to build. Perhaps this is one of the most challenging mental models to adjust. It is not enough to just engineer things into existence now. It is deciding what needs to be and what should be engineered.