The Hardest Part of Learning AI (That No One Talks About)
Staying relevant in a fast-paced industry
It’ll be no surprise to you whatsoever that the AI field is a rapidly evolving one. Almost every day there’s a new model released - each one claiming to be better than the last. For people starting out in AI - and by that I mean the first few years, not months, of learning - this constant cycle of new technology can feel like climbing a mountain with an ever-growing peak.
You might imagine this is leading to the imposter syndrome conversation. Never quite feeling good enough or knowledgeable enough at what you do. But in reality, I think there’s a much greater challenge for early ML/DL engineers, and that is staying relevant!
Breadth Over Depth
To get good at anything, you need practise. And practise in AI is often encouraged in the form of personal projects where you build and deploy models in a variety of contexts. The deeper you dive into one project, the more experience and skills you’ll gain in that area.
Now extrapolate this to a larger scale, PhD scale for example. You’d be forgiven for thinking that a PhD project offers the highest quality and depth of study to class the student as a true expert in that field. But I am here to tell you, that is most definitely not the case for a PhD in AI. In fact, in fast-paced fields I would argue that quantity (rather than depth) of projects can offer the greatest gains in terms of learning.
I know it sounds like I’m shooting myself in the foot here - I’m clearly a PhD researcher in AI but I’m telling you that focusing too deeply on a specific project might not be the best way to become an expert. The reason I’m saying this is because I spent the majority of the first year of my PhD, learning about specific models and how to use them in context. But within six months of my second year, I noticed that the majority of the models from year one were rapidly being replaced by newer, more advanced machine learning methods. This quickly taught me that success in AI comes not from mastering one model or narrow area but from building a broad foundation of adaptable engineering skills that can keep pace with the fast-paced evolution of the field.
Too much time spent focused on a single project, risks the field having moved on by the time your project’s complete, potentially rendering all your hard work obsolete. It also becomes increasingly challenging to keep up with the latest developments.
Key arguments
Cutting-edge quickly becomes old news
The half-life of knowledge in AI is crazily short compared to other industries. With new techniques and methods are released each day, a technology that was ground-break only last month, can quickly become irrelevant. Keeping on top of these developments is the best way to ensure your work remains relevant in the field.
Adaptability is worth its weight in gold in fast-paced industries
Employers (much like the rest of us in the industry) appreciate how rapidly the AI landscape can change. One new model with a revolutionary capability could disrupt the entire ecosystem. So candidates with breadth of experience (demonstrating adaptability) are likely to be high on the hiring list favourites.
Diverse experience cultivates better engineers
Maybe this one goes without saying; someone with greater experience in a range of different tools and technologies often has a more rounded skill set, equipping them better for problem-solving and intuition.
Everything I’ve said above makes it sounds as if depth is the enemy of success… we all know that’s not true. There is also significant value in exploring projects, models and methods in sufficient depth that you fully understand their strengths and limitations. I simply hope that by sharing my experience through these first two years of my PhD, I’ve encouraged other aspiring ML/DL programmers recognise the importance of ensuring breadth (as well as depth) throughout the learning process.
I’ll be writing another piece on this soon, covering how you can stay relevant in this fast-paced industry so be sure to subscribe to receive that when I do.