Look Beyond Your Classes

Intro

I am writing this post to argue two points:

1. There are good reasons for you to be paying attention to the zeitgeist and current events in your field, even as an undergraduate.
2. You can’t count on your classes to help very much with this.

These two points lead to the conclusion that you should look beyond your classes.

I think this applies most to students in fast-moving fields. I would consider machine learning/AI to be a fast-moving field. My perspective is mostly based on experience in the field of AI.

Status: First draft with some editing. Main ideas are here, I think it’s readable, but probably rough around the edges.

Point 1: You should pay attention to current events

Here’s are several reasons why you should pay attention to current events in your field:

1. New kinds of jobs emerge out of the bleeding edge. Once-stable jobs can also disappear due to developments on the bleeding edge.

2. It’s fun and interesting to hear about breakthrough developments. The frontier comes with a sense of possibility and opportunity. Nobody knows what’s going to happen, but it’s going to be interesting! Wanting front-row seats to the main event can be a powerful motivator.

3. The frontier makes use of the things that you’re learning, in new and exciting ways. This might be a powerful motivator for you while you’re learning the basics of a topic. The basics might seem dry and uninteresting otherwise, seeing how they are applied at the bleeding edge can make them feel more alive, real, and meaningful.

4. Talking points and buzzwords are not everything, but talking points can open conversations.

   • For example, if you are looking for research opportunities, demonstrating some awareness of what’s happening in the field may reflect positively on you.

5. In a fast-moving field, formal curricula can move slower than industry standards. Sometimes much slower.

6. If you aren’t excited by the zeitgeist, maybe you aren’t excited by your field. If you don’t like your classes, many things could be to blame. It could be that you don’t like your field that much, but it could also be that you don’t like your instructor, or the class format, or having to sit in a lecture hall, or learning from textbooks. Maybe you think it would be easier to know if you chose the right major if you had a chance to see the concepts applied in the real world. Paying attention to breaking news in your field is one way to approximate this. (On the other hand, if you are frustrated by the zeitgeist and you think everyone is going in the wrong direction, you might be in exactly the right field.)

7. Excitement and enthusiasm are a social phenomenon. One mindset says “I just need to focus on my classes, and muster up the motivation to study hard. I am the source of my own motivation.” There is some merit in this mindset, I don’t mean to discount it. But contrast it with another view: “I get excited when the people around me are excited. When I’m part of a group of enthusiastic people, I feel more motivated.” Sometimes I receive questions about how to “decide” what to do. I think this is a good question. My experience has been one part actively “deciding” and one part passively experiencing contagious enthusiasm. If I hadn’t seen others getting excited about big questions in AI, I wouldn’t have been as excited by them myself.

8. Frontier developments in one field can spill over onto other fields. For example, deep learning spilled over onto protein engineering via AlphaFold. NLP is having spillover influence on social science via large language models. Often this occurs at actionable time scales — you could notice, for example, that Technology A is having success in Field A, but hasn’t been applied to Field B, even though Field A and Field B share much in common. This could be an opportunity to “skate where the puck is going”.

9. Expand your self-image. Sometimes people on the cutting edge are suprisingly young. On a bad day, this can be cause for imposter syndrome or lead to discouraging comparisons. But on a good day it can be very inspiring.

Arguments against point 1:

These are valid arguments, I experience these sometimes. My general response is that paying attention to the zeitgeist is a spectrum, not an all-or-nothing commitment. I am trying to nudge you just a little in the direction of paying more attention.

• Trying to keep up with all the new developments is too stressful, or too time-consuming.

Point 2: Your classes aren’t enough to keep up with current events

There are disincentives and impracticalities that prevent your classes from keeping you up to date with the pulse of your field.

• The set of “frontier” topics is usually (much) larger than the set of core truths.

• As an undergraduate instructor, I’m responsible for making sure you know the “core truths”. “Frontier” is often equated with “graduate-level”. Giving students some perspective on recent developments is encouraged, and I try to do this. But there isn’t a hard expectation for me to show you all of the latest and greatest.

• The frontier is uncertain.

• Faculty disagree about what’s happening on the frontier.

• Much of what’s on the frontier will look insignificant in hindsight.

• Many actors on the frontier do not fit the mold of academic authority. Alexandr Wang (Scale Founder and CEO) is only 27, and has no advanced degree.

• The academic environment has an Overton window that sometimes lags the Overton window of the bleeding edge. Technical understanding flows from the bleeding edge, gets solidified over time, and eventually makes its way into undergrad curricula. But even when the technical understanding is there, sometimes the range of topics that can be comfortably discussed in an undergrad classroom feels narrower. I’m having a hard time being more descriptive than this.

An example:

Here’s an example of how paying attention to current events played out in my own life. This is my best effort to recall a sequence of events from memory. I’ll update with sources and corrections as I have time.

2020: Connor Leahy appears on the Machine Learning Street Talk podcast, predicting the arrival of “prompt engineering” as a job. I think Connor was being completely genuine, but when I heard this, I was skeptical. The show hosts were skeptical too, as I recall. Superficially, it seemed like a Fisher-Price version of software engineering, probably not something that a serious engineering team would be doing. At the same time, Connor made good arguments.

To my recollection, Connor also made the case for GPT-3 exhibiting “reasoning”. This was around the same time as the Stochastic Parrots paper. At the time, many academics completely dismissed the idea that GPT-3 could be doing anything like reasoning.

Fast forward three years later…

2023: Prompt Engineering makes the news, high TC prompt engineering jobs attract media attention.

2024: A google search for “prompt engineer” lists 19 open job postings local to me, at companies like Anthropic and Scale AI.

As of 2024, the field is still divided on whether LLMs are “reasoning” or not. From a functionalist perspective, it seems like the ability of LLMs to solve “reasoning” problems has improved (a lot) since 2020. From the philosophical or cognitive perspectives, my sense is that the jury is still out. The claim is taken much more seriously now, though, and many who completely dismissed the idea have eaten their words since.

What am I saying here? Should I have dropped what I was doing in 2020 and declared myself a prompt engineer? Maybe, but that’s not really the point.

Will prompt engineering still be a real job 10 years from now, or will it have been a fad? A good question, but also beside the point.

I guess the point I’m making is that it’s satisfying just to have been able to follow that sequence of events from an inside perspective, to have at least known that it was something to keep an eye on. I’m not dissatisfied for not having dropped everything to make a career of prompt engineering, but I do feel satisfied not being too suprised when it makes the news.

Likewise, I’m not asking you to reorganize your whole life plan around this week’s trending topics, just to keep an eye on them.

Another example:

This Tim Urban blog post from 2015https://waitbutwhy.com/2015/01/artificial-intelligence-revolution-1.html was one of the main reasons I started thinking about a career in AI. The ideas in this post felt like a huge elephant in the room. “Superintelligence” was a topic that would have been met with half-humorous half-nervous laughter. What if these ideas were true? I really wanted to know if what was described in this blog post might actually happen. The idea that something that seemed so far away from the norm could be possible was exciting and motivating. It’s since become a more acceptable topic for serious academic discussion, although still not taken seriously in some circles.

Ok I’m sold, what should I do?

If this half-baked first draft has convinced you to look beyond your classes, I would recommend that you:

• Make an X (Twitter) and follow interesting people, especially those who:
  • Work with technology that interests you
  • Have jobs you think you want
• Subscribe to podcasts or youtube channels
• Attend industry gatherings or conferences
• Read and subscribe to blogs

Further Reading

Why does getting a job in tech suck right now? Is it AI?