Artificial Intelligence (AI) is quickly growing in popularity and corporate use, with chatbots reaching the mainstream through companies like Casper with their Insomnobot 3000, and Tesla making self-driving cars a reality. AI functions by measuring scenarios it is programmed to analyse, which in turn obtains data that can be transformed into action. This is a simulation of human intelligence without the inherently human hurdles of emotion and fatigue. A major benefit of using AI for business is the ease at which an AI engine can constantly detect and analyse data without tiring. A multitude of companies is investing in AI including Amazon, Google, Microsoft, and IBM. Amazon has their voice-activated bot Alexa and opened up an AI supermarket called Amazon Go in Seattle, Google purchased AI startup Deep Mind, Microsoft Ventures launched an AI startup competition, and IBM has had their own interactive question answering computer system called Watson since 2010.
Companies working with both humans and AI such as Mighty AI, a Training Data as a Service™ company, and CloudMinds, a provider of cloud-based systems for AI bots, have honed in on these benefits of AI whilst acknowledging the importance of human supervision. Adjacent to their avid use of machine learning, both companies remain conscious of the human role in effectively programming AI and monitoring its accuracy. Mighty AI hires people to pinpoint content correctly and tag it accordingly, and from that, the machine learning technology is able to do the rest of the work. One of the individuals hired to carry out this task at Mighty AI explains her job as “teach[ing] machines to identify high heels on a photo” in a promotional video for the company. On their website CloudMinds explain that their employees “are vital to making the vision come alive”, and they “are world-class scientists, engineers, business leaders and other professionals, like medical doctors”. When working with AI, humans are responsible for training machines as well as ensuring the maintenance of the machines in order to keep standards high and the work carried out by the machines streamlined. The accuracy of data collected is improved with the large numbers of talented and vigilant people who are hired to pinpoint content correctly and tag it accordingly, and from that, the machine learning technology is able to do the rest of the work.
A major consideration for businesses surrounding the growth in implementation of AI technology is what jobs will be lost, what new job roles will be created as a result of the technology (e.g. the role of tagging content), and how staff will adapt to working with AI. With the deployment of AI across business comes the necessity for people to build, programme, and train AI bots and computer systems. AI cannot function properly without human intervention and training. Without this human element, the use of machine learning is known as unsupervised learning where no training data is used as a basis for the machine to learn from. This leaves the AI to fend for itself without the guidelines of the humans sourcing data and content for it to learn from. Unsupervised machine learning is of particular use when there is no data or when AI is being used for purely experimental purposes. For example, the 2012 Google Brain project, which consisted of the AI being confronted with millions of frames from YouTube videos without any annotation, by looking for trends and patterns, the bot taught itself to identify animal faces. Supervised machine learning is a safer alternative particularly for the development of such things as self-driving cars as lives may be on the line, so knowledge of environments based on human labelling could help correctly identify a threat that the AI could miss if left to its own devices.
Supervised machine learning algorithms rely upon training data to continuously learn from. There are different categories of algorithms: regression algorithms predict the output values based on input data, classification algorithms assign data to different categories, and anomaly detection identifies a typical pattern called outliers. Anomaly detection, for instance, can be used by companies to detect security breaches and can even identify atypical physical features on a human body such as a tumour through scans like MRIs. The human trainer of the AI is responsible for teaching the computer system how to identify these anomalies and what constitutes an anomaly.
The human role when working with AI technology is to provide a safety net and secondary source to resolve and monitor potential issues in the development and deployment of new and potentially hazardous technology. Additionally, the synergy of man and machine significantly aids productivity and efficiency as the human employees share the workload with their AI counterpart. The importance of the human worker must not be overlooked as the machine does not work without its trainer.