Why Independent Research is Hard

I have been doing independent research for the last ~8 months at this point. (Independent research for me is the practice of working on a problem and trying to find novel ways to solve it solo¹)

Current Research Focus

I’ve been working on a few problems for the past few months, namely:

• Multilingual voice assistants/agents
• World models or state predicting models for robotics navigation
• Language models and their use in helping humans solve difficult problems
• Fast inference of large models on local hardware
• Better, faster, efficent Robotics systems.

PS: If you’re an individual or a lab working on this, please slide into my DMs.

Roadblocks

For each of these problems, I’ve hit a few roadblocks:

1. Not enough compute
2. Not enough training data
3. Not enough brain cells to iterate quickly on projects
4. Not enough time

Funny how the opposite of all 4 is what you get being associated with a research lab. 3 of those 4 problems are solvable by just joining a lab; actually, the 4th one too if I get paid and can go full time (drop out maybe, lol).

I don’t come from an academic background, so the “academia and its jargon” is also kind of difficult to get used to and digest. While most of it is run by idiots² who value credentials over actual work, academia and its supporters have tried so hard to make this wall for outsiders to not easily transition into it, and that kind of sucks.

Advantages of Being a Independent Researcher

Now here comes advantages of being a lone wolf (I had to) researcher:

1. Agency: It’s insane how many times I’ve just gone into rabbit holes of my own to learn things I’m curious about, and that is something I value a lot.

2. Ability to Build and Sell: Not really sure about this one, but I can just sell things I research on later on? Or build a product out of it? How cool is that.

3. Look Really Cool While Doing It: I know people who do this or have done this in the past, and I can’t express in words how cool they are.

4. People Underestimate You: This one is really funny. I’ve had so many experiences where people underestimate you greatly, and that is great because you can just blow their expectations out of the water when the time comes.

5. You Meet Really Cool People: I’ve been posting my work on the internet for 2-3 months now, and the number of talented, high-agency individuals I’ve met has been amazing. There’s so much to learn from them.³

A Few Tips

1. Start by looking up the existing literature in the field, pick one, try to reimplement it from scratch.
2. Stalk people who lead the field. Literally 70% of my baseline knowledge of a field starts by reading blogs and papers written by researchers of that field. God bless personal websites.
3. Nice way to skim through a research paper:
   • What is the input and the output of the function shown in the paper?
   • Ask questions to yourself about if the algo can handle hard problems. At what point does it break?
   • Think about incremental improvements to existing algorithms and benchmarks.
   • Constantly be thinking of harder benchmarks and new capabilities.⁴
4. Put your work out there. I can’t express how important it is.⁵
5. Try to get in contact with people who lead or who actually innovate in that field.

home