The title of this post is inspired by the 2017 revolutionary Google paper 'Attention Is All You Need,' which I believe started the era of large language models. The paper introduced the transformer, which is the foundation of LLMs today.
It’s hard to know what’s gonna be the sh*t
I did not fully comprehend the significance of that paper/or the idea of a transformer model until this year. There were many signs in the past 8 years, though. My company's CTO walked into my office and asked me if I wanted to chat about attention (He heard that was gonna be the sh*t for building NLP models). During my master's program at UW, I saw that Transformer models often perform better than traditional NLP models if set them up correctly. Huggingface (a company that provides an open-source library for using transformer models) became a huge success and empowered my non-DS friends in building cool NLP applications. But the moment that hit me was when my mom started asking me about NLP and ChatGPT and when my dad started consulting ChatGPT about legal questions (this is not recommended…).
Technological innovation is interesting because living in the moment, you are not sure what's going to be the game changer. There are believers. Jensen Huang believed and bet on the power of GPUs. Elon Musk believed and bet on electric vehicles among other things. Steve Jobs bet on the iPhone. And then there are the rest of us who could not be certain about the paradigm shift until widespread adoption.
The Best Skill to Learn?
I recently read an article titled 'Why Taste vs Skill Will Change Your Career (Forever)' — an interesting article if you want to take a detour. The article delved into the significance and equilibrium between taste and skills for designers and PMs. However, it got me thinking: what do taste and skill mean for me in this era of thriving AI tools and automation? What’s the best skill to learn now??
Skills no longer hold as much weight unless you're engaged in a craft like woodcarving or specialized labor like plumbing — because automation is either difficult or not cost-effective yet. Instead, the new game seems to be sourcing the right tools. It ultimately boils down to weighing the opportunity costs of mastering a skill and considering its longevity. While some skills can be built upon and expanded, others simply become obsolete over time.
In an increasingly automated environment, people complain that it is difficult to determine the lifelong skills necessary to stay relevant. However, it was never easy to identify the right skills to learn anyway. My grandparents believed that driving was the best skill to learn and encouraged all my uncles to become commercial drivers. They even invested the whole family's savings to buy a large car to start a shuttle business. They were wrong because the majority of the population learned to drive quickly. People can make a living by driving, but it does not set them apart. Now, a lot of people around me believe transitioning to an AI-related career will provide job security, but I am not sure that is necessarily true. Over time, skills get normalized. We either need to become the very best in the field or we just need to learn new skills.
This phenomenon is no different from the pattern seen in college major selection. Students often prioritize majors that offer job security or potential prosperity, rather than pursuing their interests. 20 years ago, fields like finance and electrical engineering were popular choices, whereas today, students are drawn toward majors such as computer science and statistics. The hottest major will probably change again in the next decade.
Judgment is all you need
Taste, or I'd like to use the word "judgment," is key and long-lasting, though. It fundamentally sets you apart from others who could not discern good from bad or truth from lies.
I would argue that judgment is all you need. And that has not changed at all before or after the so-called AI era; it's just that people with good judgments will stand out more, and people with bad judgments will fall behind faster now.
Good judgment + hard work is the recipe for survival. Call it fortunate or unfortunate, going forward, we might not have to work as hard anymore, and the weight of judgment in this formula is higher now.
Here are a few examples of judgments at play:
1. Your copilot writes you a new piece of code; you have to judge whether it's good-quality code or even the correct implementation.
2. You want to automate a task. You googled or used "there's an AI for that." You are presented with 10 AI tools that achieve the same thing. You have to judge fast and accurately which tool you need to go with.
3. You write a new blog or marketing campaign content with ChatGPT. What are the things that you need to fact-check? Is it good enough to even be posted? You are the judge.
4. Is the AI safe enough to use? We talk about AI privacy and security issues a lot. The truth is we can't have a catch-all system, and we can't let security issues stop us from AI adoption. You have to be the judge of whether it is "safe enough" to use. Be the judge of the balance between efficiency and safety.
But I don’t really want to do any of these
Thanks for reading until the end. I wanted to add one more thought before closing so you know I am no Tony Robbins.
Have you ever felt that you know exactly what to do but you just don’t feel like doing it? I have.
I have learned the importance of judgment. Yes, I need to use my brain. Yes, I need to rely on good judgment so my skills don’t become obsolete. However, sometimes I envy people who read a tweet and just believe it. I envy people who just choose an occupation and stick with it, for better or worse. That seems to be a simpler life.
Well, well.
Till Next Time
I'm soliciting ideas to write about. Send me an email if you'd like me to write about a topic.
I greatly value and appreciate your subscription!