Is AutoML Good or Bad for AI Developers?
This post takes a look at the benefits and disadvantages of AutoML solutions for non-developers.
Join the DZone community and get the full member experience.Join For Free
There’s no doubt that AI is a hot topic right now with everything that entails and the demand for machine learning engineers and data scientists are surging. But are AI developers becoming obsolete already to the emergence of AutoML solutions right before their time to shine?
AutoML solutions such as Google AutoML lets non-developers train AI models without coding or even knowing much about data science. I’m also right now a private beta tester on a tool that lets you build deep learning models and it could honestly be taught to a 3rd grader in 20 minutes. The power and advances of these tools are amazing. I recently made potato chips recognizer with Google AutoML in just 3 hours and it worked great.
The AutoML solutions are very often based on Transfer Learning that lets the creator of the models use very little labeled data to make impressive results.
This is all great for the applied uses and the business cases of AI, but where does that leave the average data scientist or machine learning developer? Will doing actual data science and machine learning be a job for the few big tech employees at Google and OpenAI and everyone else are off to do drag and drop AI work?
I think there are arguments for and against. Looking at the developments in programming languages that went from all low level to new high-level languages that took away a lot of the difficult work with programming and lowered them to get started didn’t seem to hurt the demand for developers. Actually it will the lower cost to make useful applications added to the demand and both the very experts and new entry-level skilled developers had a spot in the market.
In general, I also guess that the amount of AI applications going up will require a massive streamlining of data and availability of data sources functioning as breeding grounds for new AI projects. Some build on AutoML solutions and some with more custom developments.
I might be going a bit far here, but I also think that the need for data scientists that can tell the actual story behind the data and not just turn data into mindless predictions will go up. It’s going to be the wild west for a while with thousands of new models being applied with no understanding about the data behind. Once the models are in the wild the demand to understand why they make the predictions that they do will be in demand.
Whatever happens, it will be very exciting to be a part of.
Opinions expressed by DZone contributors are their own.
Seven Steps To Deploy Kedro Pipelines on Amazon EMR
Revolutionizing Algorithmic Trading: The Power of Reinforcement Learning
Introduction to Domain-Driven Design
Mastering Go-Templates in Ansible With Jinja2