In the previous post I talked in length about why doing independent AI research was a rather poor experience for me. I firmly believe that pursuing independent AI research is not the best use of your time, especially if you lack academic experience and connections. The barriers to entry, competition, and lack of recognition make it a daunting and often unrewarding path.
In this post I want to switch gears and briefly talk about what you can do instead. Basically how to have more impact with less effort.
Engage in hobby or open source projects
Probably the easiest path to get started is to work on AI projects as a hobby or contribute to open-source projects. I think there is still decent chance of discovery for cool new projects. You might have to make a flashy demo or place a couple self-promotions in relevant communities. Also it’s important to have something that is immediately useful to somebody.
This approach allows you to build your skills and create a portfolio without the formal constraints of academia.
Projects like llama.cpp, RWKV, stable-diffusion-webui and contributions by individuals like TheBloke show that it’s possible to make a significant impact and even become relatively well-known within the AI community through open-source work.
Get a non-research job or join a startup
One practical route is to seek employment in a non-research role. Machine learning skills are in high demand across various industries. Even if you are not super far in your ML journey, there are a lot of AI related startups right now and some of them can’t be too picky about technical hires.
At least you’ll have some work to do and you don’t need to publishing papers. It’s not research and you almost certainly won’t be training new ML models from scratch. But surely there are places where you can practice deploying models and building AI products, while getting paid.
Start a side hustle: shareware, SaaS, or freelancing
If you have the skills for independent research, you likely have the skills for creating software or services that others might find valuable. Consider starting an AI adjacent side hustle such as software products or freelancing. Altho it can be difficult to get started, it can be rewarding both intellectually and financially. And even though you can’t expect overnight success, in my opinion it’s still easier to make a break than in research.
Educating or entertaining
If you have the right skills and drive for video creation, video sharing sites like YouTube and TikTok are some of the easiest places on the web to get traction. “The algorithm” is actually pretty fair to small and upcoming channels and plenty of people have gone viral with one good video.
The downside is that making engaging technical content is very difficult and you will be competing against all the attention grabbing content made for pure entertainment. A lot of people want to be influencers these days, so you have to stand out in some way.
Another option is writing for blogs or sites like Medium. Sometimes you see posts like this pop up to the top of aggregators like HackerNews or Reddit. I don’t know how likely it is that your writing will be discovered so I’m not going to comment on that.
Academia
While some independent researchers might consider transitioning to academia, it’s not an appealing option for me. The academic route has become highly competitive, with an increasing number of PhD candidates and graduates, vying for a limited number of positions. If I chose to pursue this path, I’d be able to continue writing papers and contributing to the body of knowledge in my field.
However, I don’t really see it benefiting me in the long run. After PhD you have to either hope you get professor position somewhere (while doing post-doc, i.e. more work for low compensation) or get hired in some industry lab. Both of these seem very unrealistic to me at this time. The reality is that securing any of these positions requires one to pretty exceptional individual with some luck and skill in navigating a highly saturated job market.
Also based on what I’ve heard, the intense pressure to publish frequently and secure grants can overshadow the joy of discovery and innovation. So all-in-all academia doesn’t sound like a very fun career path.
Academia part 2: Maybe choose traditional computer science over machine learning?
For those who still feel drawn to the academic world, it might be worth considering fields that are close to AI and ML, such as algorithms or distributed computing. While there is still competition, it’s a lot less than pure ML. If you choose your study/research topics well, you can still learn skills that are valuable in ML.
Jobs in industry research labs
Landing a job in industry research is becoming harder and harder. As an independent researcher, you are competing against the thousands of new graduates and other candidates with more traditional qualifications and affiliations with renowned institutions. Even with a strong portfolio of independent research, standing out in the pool of applicants is going to be next to impossible.
I think your best bet is to network a lot and hope one of those connections will help you get the job you want. So my advice would be to design your projects in a way that help with the networking part.
Leave a Reply