Issue 4: Understand the business side of AutoML
Is No-Code AI an evolving business model for AI?
NoCodeAI is one term that appears to not need an explanation. It calls you to build AI with zero coding. Who is this for and does it serve a business purpose?
Lets start with AutoML which is about Automating Machine Learning. Like all terms in the world of AI, its origins are disputed based on how to define it. If you are interested in its history there are two versions, one from Datarobot here and from Wikipedia showing its recent rebirth.
If you are interested in what it does for you as a business innovator, read on!
It automates the boring part of a data scientist’s work. AutoML covers the complete pipeline from the raw dataset to the deployable machine learning model. It is marketed as filling a gap because of the demand for data scientists.
I do not believe that AutoML will replace data scientists.
It sure makes the junior data scientist get to model fitting faster. But AutoML is marketed to the business user as No-Code AI. That is where it gets interesting.
I recently taught a Capstone AI Lab online for multi-disciplinary business teams using IBM’s AutoAI and a new AutoML startup Akk.io. AutoML Vision with Google’s teachablemachine has been my favorite to showcase machine learning in a visual way.
If you trained a binary model of dogs and cats and got a model that was 90% confident, and feed it an image of a donkey, what will it do?
I tested it in my recent Capstone AI Lab course, remember, its a datascience course for business users. I trained a binary model of dogs and mops (curious? google dogmop photos) and I got an algorithm that was 90% confident, and showed it my face on a webcam and see the results (image below) for yourself.
I wore glasses and the model was more confident that I was more dog than mop! lol
Well, we have lots to unpeel in this topic to understand why a business user should care about AutoML.
tl:dr: You need to goto the start of the business problem you want to solve. Do not start with a dataset assuming it is building a model that is going to solve your business problem.
Tomorrow at our weeklywed we have Vera Serdiukova, Edge AI Product Leader from LG Electronics talking to us about AutoML on the edge. Now thats another fork in the road of AutoML from the cloud to the edge.
Stay with me on the business model question. Tell me if you want to continue this conversation about AutoML and No-Code AI.
Till next time!
And be sure to, tell your friends!