The Data Scientist made ChatGPT and ChatGPT undid the Data Scientist.
(This article is a transcript in English of one of my french LinkedIn posts)
The headline is deliberately provocative... Nevertheless, I observe that in recent years many jobs in Data have begun a "metamorphosis". Among them, the job of Data Scientist is undoubtedly the one that has changed the most.
Just a few years ago, a Data Scientist could, with his PC and Jupyter Notebook "contribute to science" with his knowledge of math, statistics and motivation! His main job was to design machine learning models and train them. If we want to simplify, we can say he was an "R&D Data Scientist".
Gradually, there was a shift from the "classic" DS because with the arrival of increasingly efficient and accessible machine learning tools and libraries, designing models became an increasingly uncomplicated task. On the other hand, industrializing models and monitoring their life cycle became critical issues. This led to the emergence of the ML Engineer and MLOps Engineer.
Their jobs were now to design systems and maintain them over time.
At the same time, the job of Data Scientist became more specialized, it is increasingly difficult to be competent in both tabular data processing, NLP and Computer Vision. Some Data Scientists therefore now focus on a domain of expertise.
Then, with the emergence of very large models such as LLMs, designing state-of-the-art models is becoming, in part, a luxury reserved for companies that have very large computing capabilities and a lot of money! Not just anyone can work on a daily basis designing and training models like GPT3.5 or GPT4.
So the DS job is changing again, it's less and less about training new models from scratch, but more about using pre-trained models, adjusting them, and assembling them to form a coherent system. In other words, the S in Data Scientist is starting to falter as the job shifts little by little towards Engineering. In fact, we should talk more about AI Engineer.
Today, some of the DS's prerogatives are still nibbled a little more by Generative AI by automating certain data collection, cleaning, labeling and code generation tasks.
Data Scientist is becoming a very (too) generic term that can refer to both the R&D DS, the ML Engineer, the MLOps Engineer and the AI Engineer (and sometimes even the Data Engineer).
The Data Scientist still retains all his relevance and expertise, even if his scope has shifted with technological developments. His role is not called into question but is transforming enormously.
Finally, what is a Data Scientist in 2023?