Are you falling into the "ML for everything" trap? 🧠 Senior Data Scientists know that engineering maturity means delivering simple, robust, and effective solutions, and that often means avoiding the complexity of Machine Learning. In the rush to adopt AI, many junior practitioners reach for a neural network when a simple SQL query would do the job better, faster, and cheaper. This video is about reaching Data Science maturity. We break down the costly pitfalls of over-engineering and show you ...
| 0:10 | In the rush to adopt AI, it's easy to |
| 0:12 | fall into a common trap, using machine |
| 0:14 | learning for everything. This is a |
| 0:16 | junior mindset. Senior data scientists |
| 0:18 | deliver robust, simple, and effective |
| 0:21 | solutions. And that often means avoiding |
| 0:23 | the complexity of ML. This is about |
| 0:26 | engineering maturity, avoiding costly |
| 0:28 | pitfalls, and choosing the simplest, |
| 0:31 | most robust solution for the problem at |
| 0:33 | hand. So, let's show three examples of |
| 0:36 | when it's better not to use machine |
| 0:38 | learning. Then, we'll talk about a |
| 0:39 | general framework for deciding whether |
| 0:41 | to use ML or not. Let's start with a |
| 0:44 | common feature, the most popular list. |
| 0:46 | You could spend weeks building a complex |
| 0:48 | personalization engine [music] complete |
| 0:50 | with feature stores and training |
| 0:52 | pipelines just to battle inference |
| 0:54 | latency for a realtime result. But what |
| 0:56 | if you just need to show what's trending |
| 0:58 | right now? The answer is often a simple |
| 1:00 | SQL query, a count of clicks over the |
| 1:03 | last 24 hours. It's fast, cheap, and |
| 1:06 | perfectly interpretable. If you need |
| 1:07 | even faster realtime updates, you can |
| 1:10 | use an in-memory store like Reddus with |
| 1:12 | a sorted set. Both are orders of |
| 1:14 | magnitude simpler than a full-blown ML |
Showing first 30 lines. Search the full transcript in our archive.
ThingsMrBeastSaid.com is an unofficial fan archive. Not affiliated with MrBeast or Beast Industries.