Democratizing AI

By: Guest

8, May, 2019

Categories:

Artificial Intelligence - Deep Learning - Machine Learning -

  •  
  •  
  •  
  •  
  •  
  •  

By Jacob Wells – Technical Specialist for Wolfram Research

As modern machine learning has burst out of the research lab into everyday practical application, it seems to have created an instant elite of data scientists who know their neural networks from their gradient-boosted trees and can command the high salaries that back up that knowledge. However, technology always evolves to become more accessible, and as a result this has allowed AI to be practised and applied by a range of consumers, students, scholars and professionals.

Here at Wolfram we are trying accelerate this change by automating the harder parts of that process, allowing these non-experts to undertake an AI-based project with ease. By allowing us to provide easy-to-use neural networks, automated machine learning and a host of other tools used in AI development, you have more time to focus on the application and direction of the project in hand, rather than having to know and implement all of the computational details.

If you’re familiar with Amazon Web Services and other cloud-based computer services, you will know the average person’s hardware is no longer restricted to the confines of their personal computer, and now it is possible to tap into the seemingly infinite power of computers hosted elsewhere. This has allowed anyone to use the computing power comparative to huge organisations. This level of accessibility is reflected on the software side of the situation, through the Wolfram Language. Rather than reserving the power only for the experts, we allow anyone with elementary programming skills to begin building AI-based applications with ease. Additionally, we provide the complete pipeline allowing you to draft an idea, apply the correct tools and then publish that to a shareable document or app via the cloud.

Simple Supervised Machine Learning in Wolfram Language

Until recently, smaller businesses with limited budgets were at a disadvantage in terms of the application of AI-based projects; however, with the accessibility presented in the Wolfram Language and the presence of cloud-based computing services, it is now possible for smaller businesses to harness the power of AI—for example:

  1. Smarter marketing. Through accessible AI, a multitude of powerful marketing strategies can be deployed—for example, image recognition–based advertising, product categorisation, product pricing, sentiment analysis and many more.
  2. Detecting and analysing inefficiencies within a production process. By harnessing the power of AI, it is possible to have 24/7 analytics of a production process.
  3. Instant and accessible business analytics. Through AI it is possible to query the numbers in any which way to unlock insight that may not have been seen through more traditional statistical analysis.

These existing AI tools were created by pioneering data scientists; however, through the Wolfram Language, they are now accessible to all. With the complete symbolic representation of neural networks and our automatic algorithm selection within the machine learning framework, it couldn’t be easier. I’m not suggesting that it’s the end of the expert “data scientist,” as there is always room to develop and release groundbreaking neural networks and other powerful applications. Automation can come from two different sides, the first being from the ground up, where you have minimal knowledge and you’re looking for the ease of use and accessibility of AI within the Wolfram Language. The second use is for the data science expert who can use the Wolfram Language to make the process as fast as possible. The notebook system, ready-to-go neural networks, automated GPU training and easy function calls allow for both rapid prototyping and easy learning.


Some of the AI possibilities in Wolfram Language

Beyond the tools provided, the role of an expert data scientist is to apply imagination to the application. There’s no use in having easy-to-use tools if you don’t know what you want to infer from your data. Traditionally, we have measured purely statistical figures from business data, but with the power of AI data scientists are able to prod, poke and question the data in mostly any way they want, unveiling key insights which weren’t previously visible to the traditional practices.

To learn more about the democratization of AI, come along to Wolfram’s Mark Braithwaite’s talk “AI: Experts Not Required” at 12:30pm on the AI Technology Solutions Track at AI & Big Data Europe 2019 in Amsterdam. Wolfram will be at also be at stand 682, and we are bringing our in-house experts on data science, machine learning, AI and a bit of everything else. We would love for you to come along, and perhaps we can find a way you can harness the power of the Wolfram Language.