Everyone is talking AI nowadays, which usually means machine learning and predictive analytics. The background is obvious.
With AI, we can develop self-learning and self-optimizing software, which automates data analysis and gives us autonomous decision making. We can, therefore, produce much smarter software.
Everyone is talking AI nowadays, which usually means machine learning and predictive analytics
The AI algorithms that machine learning uses are for predicting the most likely future behavior of someone or something. One can say that we derive, or train, the software’s behavior using old data.
The software thus learns to predict how the future will likely become, and can adjust its behavior by continuous re-training as new data with other data patterns becomes available.
The Data Scientist Shortage
One problem here is that mathematicians, statisticians, and data scientists are required to develop the machine learning models, which are then implemented in different types of software to make them smarter.
With AI, we can develop self-learning and self-optimizing software, which automates data analysis and gives us autonomous decision making.
There is currently a shortage of data scientists. It is simply difficult to get the skills needed to develop self-learning AI systems.
Thus, the development is inhibited – or at least its introduction on a larger scale in wider circles.
Citizen Data Scientists
However, with “Automated Machine Learning” (Auto-ML, or AML), this problem is solved elegantly, at least for some types of machine learning applications.
With automatic machine learning, AI is used to generate AI; that is, we can use AML software to create other machine learning software automatically.
Thus, data scientists with expert mathematical skills are no longer needed – at least not for all purposes. Instead, staff with domain expertise but less mathematical knowledge can use AML software to generate new machine learning systems automatically.
The AI algorithms that machine learning uses are for predicting the most likely future behavior of someone or something.
This type of staff is called “Citizen Data Scientists”, as they certainly do data science tasks, but have mathematical skills that more or less corresponds to any citizen.
I consider myself to be a “citizen data scientist”, although I arguably have higher mathematical education and machine learning knowledge than the average citizen.
However, I am not a mathematician or data scientist; but can nevertheless produce machine learning models using AML systems.
AML: Generate AI with AI
Auto-ML systems can be roughly divided into two groups; first, the AML systems that allow anyone to create a machine learning model with a few mouse clicks, and secondly the systems that allow much more settings and thus mainly facilitate and automate the work of well-trained data scientists.
In principle, historical data is imported into this type of system, which is then automatically trained to make future predictions on unseen data. The systems use AI to select the most accurate AI algorithm, and also to configure it in the most optimal way (algorithm selection, feature engineering, hyperparameters, etc.).
The systems use AI to select the most accurate AI algorithm
Then, a machine learning model is automatically generated, either as an automated cloud solution that can be called from your software, or by generating program code that can be integrated into your software.
Some solutions also have a graphical user interface that allows you to make predictions on new data by simply importing an Excel file or similar, whereby dropping the requirement to develop your software that uses the machine learning model to make predictions.
Some of the Auto-ML tool providers include Compellon, DataRobot, DMWay, H2O, MLJAR, PurePredictive, and Xpanse Analytics.
I am a project manager for an AI project where we use DMWay to generate AI models with AI automatically. If time and opportunity are given, I will write an article on this at a later stage.
Auto-ML Democratizes AI
I predict that automated machine learning with AML will see explosive growth in the business world over the next few years, as many AI solutions can thus be developed with little or no knowledge of the mathematics behind it.
Automated machine learning with AML will see explosive growth in the business world over the next few years
One can, therefore, say that Auto-ML democratizes AI and machine learning and that virtually anyone with enough data can produce their machine learning solutions using AML.
Are you interested in AML? Alternatively, do you need a “citizen data scientist”, an AI consultant or a project manager for an AI project? Contact me, and we can develop new machine learning solutions for you!
If you have more specific needs, you can use our specialized consulting services aimed at the development of AI and machine learning, where we use a network of experienced mathematicians and data scientists for algorithms and model development.