We put the AI in AInxiety

It seems like the right time to write about this (lol). Another year passes by, and so do a billion more papers. Can someone make it a teeny bit slower? My unread papers in Zotero are mocking me at this point. The theme is to talk about the uneasiness in my professional life, specifically with what I do, AI.

Don’t get me wrong, it’s a fantastic time to be doing AI, but it’s more important to stay sane (or even happy?), let’s talk about the latter here.

Disclaimer: I write purely/mostly from an early PhD academic viewpoint, maybe my friends in the industry have differing experiences (or maybe not).

I will try to organize this by what I think are the sources of the AInxiety, and what it means or leads to as an early academic or even anyone doing AI.

Things that trickle down from limited spots

The term ‘limited spots’ here is overloaded. It starts with first competing to get into a CS grad program (you should look at some of the applicants these days), competing for resources (compute, etc) while being in the program, finding internships, and eventually jobs.

One would imagine after spending (at least) half a decade doing a PhD you’d pretty much be guaranteed a fantastic role, but hey, this is not an MBA (only a slight dig). You could argue that this is the known academic path, yes, this is the typical journey, but the sheer number of people doing AI compared to the “decent spots” is not a pretty number, especially in the last 2-3 years.

As a consequence, you feel a need to “stand out”. This translates to being “experienced in publishing at top AI/NLP venues such as XYZ” for internship applications, not even that, this might even be the bar for even getting a PhD admission in the first place.

But wait, isn’t the point of a PhD to become that researcher? It used to be.

So what does this lead to?

Pressure to publish: More true in the last few years, with everyone publishing X number of A* papers every year, this has certainly become an implicit thing. You need to be “productive” and demonstrate that you’re “it”, by publishing in these elite venues (which I don’t think should be the aim but a consequence of doing interesting research).

Things that trickle down from the rich become richer

What I mean by this is that within the community, there are certainly groups/individuals that have more resources than others, in terms of computing, workforce, etc; (very often this is more than well deserved)! Some people (from the outside) don’t realize this is also true in academia, maybe more so in experimental/compute-intensive research, and especially now in AI. This certainly puts them at an advantage to solve certain problems that other’s cant;

But what does this mean?

Fear of being scooped: With many low-hanging fruits in AI these days, many of those ideas you might want to pursue might just be grid-searched away by people with immense resources at their disposal. This creates a state of constant pressure to put your work out as soon as possible; which shouldn’t be the ideal state of mind for a researcher to do deep (or meaningful) thinking.

Things that trickle down from the sheer pace/scale

There are things that I can attribute to the pace of things; which in some way ofc comes down from what I’ve already talked about. But I think a recent change in the last few years, could be attributed to the fact that the industry now has an equal if not higher involvement in the AI research community.

For me, this pace leads to the following:

Impossible to stay up-to-date: I’m pretty sure now there exist lists of lists of daily top-X AI papers, and this might not even be an exaggeration. It’s just impossible to a) even know what new is coming out each day, and b) even read the abstract of that paper. It’s baffling. I feel awful that I can’t even catch up to the latest in my niche, despite training to eventually be an expert.
Fear of obsolescence: Unfortunately for xAI researchers, I don’t mean obsolescence of humans (sorry folks). I just mean the constant thought of hey, will what I do today be even useful tomorrow because I have seen way too many works become not-sota-anymore within literal hours. Yes, I am looking at you oAI, Anthropic and everyone else that’s benchmark-maxxing. (I understand where you come from, but it still is a reality).

Is it all that bad? (Read: why are you even doing this?)

A very fair question after reading this, is if it feels this way why are you even doing this?
The answer to that is I really enjoy what I do. I find what I do to be extremely meaningful and enjoyable, but is it all I’d like to think about 24x7? Absolutely not!

Despite my not-so-subtle venting, the vent is not directed toward the work itself but the state we find ourselves in as a community.

Then, are we just doomed? Finding Sanity.

I’ve thought about this for a while, and I’ve found a few things (that are in my control), which have helped me find my sanity (my friends would disagree and tell you I’m not).

Honest conversations with colleagues: The lab/advisor layer is extremely important. Having the space to have honest conversations about your feelings/state with your peers and advisor is one of the most calming experiences in this chaos. It gives a great sense that we’re all in the same boat, and are here to see each other win.
Venting to friends & family: or writing blogs into the abyss (lol). Just a simple co-existence and vent session with my goofy friends is more often than not a cure to a lot of issues. It helps put things into perspective and makes me realize it’s not that deep. Pro Tip: Having non-CS friends helps a lot, I don’t want to talk about what I do at the dinner table.
Disconnecting from social media: underrated, or is it idk, but not using twitter for a few weeks helped a lot. I stopped looking at other people’s work and inadvertently comparing them with my own self, which is extremely unhealthy. Regular breaks from social media - works like a charm.
Finding your place: One thing that’s helped me is finding areas that I can work in at the pace that I’d like to. Believe me, spending time doing this is extremely helpful in the long run. This, coupled with coming to terms with how much time I want to spend doing what I do, and the potential consequences of not hustling 24x7 has helped.

Last but not least, if you think you need to talk to someone and do not have any outlet, please reach out to me.