Oliver Molander

Most popular programming languages 2000 - 2023 📊

🟣 JetBrains recently surveyed 29,269 developers, scattered across 187 countries. Devs are most likely to learn Go and Rust in 2023, the survey says: https://lnkd.in/dRXAJH4D

🟣 Upskilling continues to be important to developers, as the study further showed that half of all survey participants said they plan to adopt a new programming language. The top five are:

1️⃣ Go
2️⃣ Rust
3️⃣ Kotlin
4️⃣ Python
5️⃣ Typescript

Programming languages go in and out of favor. The data from the survey however underscores that Go and Rust are increasing in popularity.

What do you think?

#programming #rust #go

It's funny because it's true...

In the research and academic setting, the datasets one works with are often clean and well-formatted, freeing you to focus on developing and training models.

They are static by nature so the community can use them to benchmark new architectures and techniques.

This means that many people might have used and discussed the same datasets, and the quirks of the dataset are known.

You might even find open-source scripts to process and feed the data directly into your models.

➡️ In production and in the real world, data, if available, is messy. It’s noisy, possibly unstructured, constantly shifting, etc. Since data, in reality, is oftentimes produced and consumed over time, it is exposed to data drift causing the data to be outside the relevant range of the models.

As Chip Huyen mentions in the Stanford University Machine Learning Systems Design course:

🟣 Data in research is clean, static and mostly historical data

🟣 Data in production (e.g. the real world) is messy,
constantly shifting, historical + streaming data, biased (and you don't know how biased) and you have privacy + regulatory concerns

➡️ In academia and research, since you don’t serve your models to users, you mostly work with historical data, e.g. data that already exists and is stored somewhere. In production, most likely you’ll also have to e.g. work with data that is being constantly generated by users, systems, and third-party data.

As we’ve entered a new decade we’ve also started to realize (or accept) that data quality trumps any algorithm; as long as ML is the foundation of AI, data will be 10x more important than the algorithm.

☝️And a lesson for anyone coming from academia into the real dirty world: never trust a cute-looking dataset. Data in the wild is always hard to tame.

In the real world, your data is never perfect.

#dataengineering #machinelearning #dataquality