Artificial Intelligence Explained

AI is many things

AI is a set of computer software algorithms that process large volumes of data and derive conclusions independently of human intervention. The term AI describes various algorithms designed to address specific applications shown below.

AI is a name for many computer software algorithms[/caption]  

Neural Networks, Convolutional Neural Networks, Robotic Process Automation (RPA) and Self Learning are alternate names for some of the terms show in the diagram.

AI is not real intelligence

It is not intelligence in a human sense although much of it is modeled on a brain. Each application of AI is a computer software algorithm designed for a specific set of functions. There are algorithms that can exceed accuracy of doctors . Google’s DeepMind algorithm learned to play chess and beat the World Champion. There are pretty good systems that can drive a car to its destination although none are replacing humans yet. AI algorithms are used by hedge funds on Wall Street to pick investments. There is no system however that can do them all. 

AI is hardware intensive. For example, the Google DeepMind computer that beat the World chess champion uses over 1,200 computers (CPUs) and 176 Computer Graphics Processors plus countless other hardware (memory, etc…). That is a lot of hardware compared to the human brain and it only does one thing, plays chess. The Deep Mind can learn other games and play them better than a human but still we are a long way from humanoids walking the streets and competing with humans.

How AI works?

AI is an entity that remembers all the data it learned, can absorb huge amounts of new data and can analyze it within seconds. It derives conclusions and makes decisions or directs specific activities without a human. It has no emotions, no biases, does not get distracted and can work nonstop. 

An interesting and troubling aspect of AI was raised by MIT Technology Review recently. In some cases the designers of the AI algorithms called Deep Learning don’t actually understand why it makes certain decisions. For example, an AI system decides the credit worthiness of a loan applicant or whether a prisoner gets parole and no one understands why it decided what it did. This however, is no different from a human brain. We don’t know exactly how the brain decides. In a typical parole board hearing some board members will vote in favor and some against a parole, even though they all saw the same information about a prisoner.

Why is AI Important?

Simply stated, AI can replace humans in specific tasks, do the job better, do it more reliably and consistently and do it cheaper. If done properly AI can increase both the top and bottom lines of a business.

This is why Venture Capital firms invested over $60B between 2015 and 2019 in thousands of AI companies.


According to directory of high technology companies there are 15,900 companies Worldwide that offer AI based products. 90% of AI companies employ less than 50 people which means they are start-ups and that AI is in the early stages of innovation. 

The number of companies is not very accurate as according to Forbes half of AI start-ups aren’t really AI but label themselves as such.

AI uses

AI performs well in two extremes of complexity. It is great at replacing humans in repetitive tasks ranging from product inspection in manufacturing to back office administration tasks. It is also good at analyzing large amounts of data. In each case the reasons are different. Humans can get bored, distracted or sloppy in performing routine tasks but AI will not. Large amounts of data hide trends, correlations, similarities and relationships that humans are not likely to discover. The table below shows some applications of this technology.


AI Risks

  • It is not an off the shelf product: Unlike CRM or Cyber Security software, AI is a set of tools that you need to adapt to your environment. 
  • The technology is moving fast: What was good information yesterday is obsolete tomorrow. Keeping up to date is hard even for experts. 
  • Shortage of technical staff: Demand for AI engineers is high and pushed salaries to the $150,000 range with some in Silicon Valley earning over $300,000. 
  • Deployment times vary widely:  Surveys show a wide range as can be seen below.

Source: Algorithmia’s “2020 State of Enterprise ML”. This question about how long it takes to deploy an ML model into production was only asked to a subset of respondents at a company that has an ML model production.


Key considerations if you’re thinking of using AI in your business.

  • Best way to begin is to focus on one business problem. The narrower the scope the quicker the effect. 
  • Plan to experiment using a pilot. A third of projects can get stuck at the proof of concept stage Alegion survey indicates (see diagram below).
  • AI takes a lot of training data. Do you have it or can get it? For example, an image recognition project may require 25,000 training images to get accuracy of 98%.
  • What concrete measurements will you use to determine its ROI potential? Increasing revenue if you’re considering a product recommender, greater customer handling capacity if you’re planning a chatbot for your service desk, lower production costs if you’re using machine vision for QC.

Questions to ask vendors

The list of questions depends on your industry and application of AI you envision for your business. We provide more specific questions in tutorials covering these topics. Here are some that appear universally:

  • What business improvement can I expect? Do you have any data to back your claim?
  • How much data and in what format do I need to train your system?
  • Is my data suitable?
  • Is there training data I can buy?
  • How much time and cost to deploy a pilot?
  • What internal expertise do I need to deploy and maintain the system? Given that AI is new, updates and changes are likely to be frequent.

How to find the right vendor to help you.

Most vendors’ Web sites offer marketing fluff with few specifics. You have to register and talk to their sales staff to get information you need. This is understandable as the topic of AI is new, complex and they want to make sure their product fits your needs.

Given the large number of vendors this can be a lot of work. You can save a lot of time by using a vendor pre-qualification portal like