Article: How the Cloud will Democratise AI
Written by Rebecca Vickery, Data Scientist, Holiday Extras
How the Cloud will democratise AI
An increasing number of businesses are looking to take advantage of Artificial Intelligence (AI) to fuel growth, and as a result, the demand for people with the right skills to build these systems grows. The leading cloud providers are fast developing tools that will lower the barrier to entry and allow more businesses to take advantage of intelligent systems. Over previous years leveraging the power of AI might have meant significant investment both in people and technology. We are now seeing many emerging cloud-based tools that will enable the democratisation of AI.
At the moment it is difficult and expensive to find data scientists or engineers skilled in building AI-based systems and products. In January this year Indeed released a report that stated that job postings for data scientists jumped 31% in December 2018 compared to the same period the previous year. However, at the same time, data science searches as a share of all searches grew at a much lower rate of 14%. It is clear that demand for data science talent is far outstripping supply.
If you evaluate the potential uses for AI products in pretty much any business you will see that they are vast and wide-ranging. For most businesses to take advantage of this, by building these models in house, would require a large data science team. Conversely, there is another problem, and that is although data scientists may have deep data expertise, they often lack the necessary software engineering skills required to put these models into production. This makes operationalizing AI systems both time consuming and expensive.
The leading cloud providers are developing solutions to these problems and more, by releasing tools that enable the democratisation of AI across businesses. Let’s look at three key areas where the cloud is providing solutions to these barriers, and how we might expect this to develop further over the next few years.
Pre-trained AI services
Amazon Cloud Services, Google Cloud Platform, and Microsoft Azure have all released pre-trained AI services that allow intelligent models to be implemented by non-machine learning experts. These pre-trained models provide intelligent systems for some common use cases such as image recognition, product recommendation, and advanced text analysis. Over the next few years I would expect the offerings for these pre-trained services to become more far-reaching in the problems that they solve, and the businesses they can be used in.
Deploying custom machine learning models
The cloud providers know that pre-trained models will not be able to provide a solution for all problems, and indeed work for all businesses, so they are also releasing tools that easily allow data scientists to build, train and deploy custom models in the cloud. These services reduce the need for in-depth software engineering skills and allow data scientists to focus on creating quality models, that can quickly be scaled and deployed for operational use. Over the next few years, we should see the capability for these tools expand further, supporting a greater variety of machine learning libraries and simplifying the deployment process further.
AI for non-programmers
The cloud is also making AI accessible to non-programmers by releasing tools that allow custom models to be built without writing any code beyond SQL. Google recently released built-in algorithms that allow a non-programmer to submit pre-formatted data to an algorithm to train a model without writing any code. Google also has another product, ‘BigQuery ML’, which provides the functionality to train a machine learning model via a SQL query. I would expect these tools to become more ubiquitous across cloud providers and to support a wider variety of libraries. Ultimately this should make machine learning accessible beyond the data science and engineering teams, to analysts and other business users.
The cloud provides vast computing power which has made large scale AI systems possible. It is clear that the leading cloud providers are now leveraging this power even further by releasing tools that allow AI to be used by a wider range of people and enterprises. This in the future will lead to AI being democratised in the same way the Business Intelligence tools propagated data-driven decision making across all roles in a business.
Rebecca will be speaking at Cloud & DevOps World Summit on June 12th, presenting a Holiday Extras case study on 'Employing Google Cloud Machine Learning Engine to develop models in production'.