Good morning friends! ☀️
This week, I want to discuss something I have been asked a few times, which is should you read research papers as a data scientist?
![](https://substackcdn.com/image/fetch/w_1456,c_limit,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2Fd0076fa2-ff5e-4726-b603-fb2591a3b8b1_772x416.png)
Machine learning and AI are very hot right now. There seems to appear something exciting and new every week, particularly within the Large Language Models (LLMs) space.
It's very similar to physics in the early 20th century with the explosion of quantum mechanics from the likes of Einstein, Dirac, and Bohr.
If you want to understand all the details behind this research, then reading journal papers is the most thorough way. However, this can be quite a time-consuming process.
So, what should you do? Let’s break it down into pros and cons.
Pros:
Gain an in-depth understanding of the topic area you are interested in
Greatly increase your theoretical knowledge
More aware of the latest research in space
Helps strike inspiration and potential solutions for your work
Cons:
May be hard to find papers directly relevant to you
The language can be complex and difficult to understand
Takes quite a bit of time to read and fully digest
Blogs, videos, or other media may explain it in simpler
In my opinion, reading research papers is useful, particularly if you are moving on the individual contributor track to senior positions. In this path, it is important to have a solid theoretical understanding of an area(s) that you specialize in.
I don’t think you should necessarily read several a week, although the more, the better. Maybe around two a month in the area of your expertise should be sufficient to keep you up to date with anything major.
Finding papers can be difficult. A great starter resource I recommend is the GitHub repo ML-Papers-of-the-Week. It does what it says on the tin. If you want to find more specialist papers, then from a quick Google search you can probably find journals in that space you are after.
My challenge to you, find a paper to read this week and share your thoughts with me! I am looking forward to it!
What’s Been Cooking 🥘
Some tasty stories this week:
New LLM Beats GPT-4 — AI company Anthropic has released their new LLM Claude 3 which is outperforming GPT-4 and Gemini across benchmarks.
OpenAI Fires Back At Musk — OpenAI has punched back against Elon’s suit by releasing a blog post where they claim he wanted a for-profit structure in the past.
Ex-Google Engineer Steals AI Secrets — An ex-Google employee has been charged with stealing 500 confidential AI files while working for two Chinese startups, one in Beijing and the other in Shanghai.
Weekly Favourites ❤️
🎙️ Podcast — Bill Ackman | Lex Fridman Podcast #413. Some people may not like Bill Ackman, but he is an interesting and intelligent person nonetheless. I particularly liked the advice he gave on investing.
🎙️ Podcast — Yann LeCun | Lex Fridman Podcast #416. If you are in the AI space, you know who Yann LeCun is. A terrific conversation around LLMs and AGI.
🎬 Video — Nvidia CUDA in 100 Seconds. If you have ever wanted to know what Nivida CUDA is, then this video is for you!
(PS: Some links are affiliate links that I get a kickback from with no extra cost to you 😎)