AI for everyone? It’s no alchemy run by wizards!

8 July 2020
Blog Image

Is AI something for you?

We should use AI to make our organization smarter! Chances are you’ve recently heard this statement and see organizations acting to it. Leading companies are using AI across departments to increase productivity. Customer care organizations are using chatbots and speech recognition in their customer contacts. Marketing departments are predicting churn and segmenting their customer base.

Venturing into the field of AI may seem daunting at first. While AI has been hyped immensely in the last decade, it is a deeply technical field. Gartner even made a prediction that 80% of AI will remain alchemy run by wizards whose talents won’t scale in the organization. Your organization probably doesn’t have the scale and technological know-how of Google, Apple or Facebook. This might lead you to conclude that developing AI models isn’t for you. You are right, partially.

The difference between developing AI models and using them

While developing AI models might require very specific knowledge and data, using AI models can actually be much easier. Let’s illustrate this by comparing a GPS and facial recognition. When you use your GPS, you provide an address as input, and get the best route to the address as output. We don’t know how the GPS actually calculates this route, but we know how to use it. The same holds for facial recognition. We provide a picture as input, and get people’s identity as output. Both are complex tasks for which models exist to solve them. The difference lies in the way the models are obtained.

While models to find the optimal routes are programmed by engineers, models to recognize faces are learned using AI. Techniques like deep learning are used by feeding them extensive data sets until they solve the task at hand. Although the way these models are developed differs drastically, using these models is very similar: both require input and generate an output.

The cloud vendors are offering AI services

One example in which AI has become more accessible is content moderation. Every company needs an online presence and many websites encourage their visitors to upload content like reviews, pictures and videos. While the vast majority of posts will be perfectly fine, you don’t want criminal or inappropriate content to appear on your website.

Content moderators review the content and make sure it’s appropriate for publication. Manually combing through every post is time consuming and error prone. All three large cloud providers (Azure, AWS, Google Cloud) have recognized this need and provide services to assist your content moderators through AI (Content Moderator, Rekognition and Video AI). You simply send your content to an API and it detects which content is inappropriate.

A second example lies in the area of customer care. Many companies operate call centers for their customers. On top of that, customers are also finding their way to social media and other digital channels to express their frustrations (and hopefully great experiences as well). Processing all this customer feedback can be invaluable. It allows companies to connect with their customers, know what they say about their brand and react accordingly.

Manually processing all these interactions can be expensive, tedious and error prone. Luckily the cloud providers provide valuable services that help us to automate this process. Speech to text algorithms automatically generate written text from audio fragments. Azure (Speech to Text), AWS (Transcribe) and Google Cloud (SPEECH-TO-TEXT) all provide services that execute this task. You don’t need any AI skills to be able to use them. Sentiment analysis algorithms help us to analyze text and determine whether it’s positive or negative. Again, we don’t need AI skills to execute them since the cloud vendors all provide services: Azure (Text Analytics), AWS(Comprehend) and Google Cloud (Natural Language).

AI for everyone!

While developing AI models requires large amounts of data and very specialized skills, using the models doesn’t. Many vendors are providing services that make most common AI tasks easily accessible, even for smaller organizations. If you want to solve a specific AI task, chances are it’s already solved for you, the challenge is finding the right solution. At Datashift, we help companies identifying which AI solutions will generate the most value and how to implement them as efficiently as possible.