A Non-Technical Guide to Building an AI System—From an AI Engineer

Written by Jerry Buaba

Over the decades, start-up companies and individuals taking on personal projects have adopted the habit of integrating Artificial Intelligence and Machine Learning into almost every project. AI and ML are meant to make things efficient and work faster—but you might be doing harm than good if you have a faulty AI system? In this story, we will discuss the benefits of AI in organizations and individual projects, reasons why AI should not be in every project, and how to build a solid AI system.

As AI is being democratized, more and more organizations are looking towards building models and making data predictions, but few are doing what it takes to implement a proper AI system.

Gartner predicts that through 2022, 85% of AI projects will deliver erroneous outcomes due to bias in data, algorithms, or the teams responsible for managing them.

Just like a design system for products or brand books for messaging, there needs to be components and a process that guides consistency in how you collect data and build AI models.

What is an AI System?

An AI system is process infrastructure to use and scale your AI in an efficient way. Much like a design system, an AI system is a whole process that is built and maintained by an organization. A solid AI system is accurate, unbiased, secure, and saves time from collecting data to deploying algorithms.

4 Things to Address Before Building an AI System

Artificial Intelligence and Machine Learning can accomplish so many complex tasks but there are many concerns that organizations and individuals should know about before jumping into implementing AI and ML in their products and services.

  • Use Cases: No system, no matter how intelligent, can do everything. AI should be directed at solving case-specific problems and not general problems. In every scenario, before one attempts to build Artificial Intelligent systems, one must make sure there is a problem that is to be solved and later work towards solving that particular problem. It is absolutely impractical to attempt to solve all the problems of the world using AI.
  • Privacy: At all times, data privacy laws should be obeyed. In the attempt to build robust intelligent systems, organizations and individuals may tend to require lots of data from clients, customers, employees and others but not every data is accessible for use. Some data is protected by a significant number of laws called the data privacy laws and infringement on these laws have serious legal ramifications. Before attempting to collect or use any sort of data from clients or employees, make sure you have the permission to do so. If you do not have the required permissions to collect and use certain data, ask your clients or employees to sign a contract explicitly giving you the permission to collect or use their data.
  • Bias And Inequality: Any organization or individual that seeks to develop products or provide services that are powered by Artificial Intelligence or Machine Learning has the sole responsibility of making sure that the product or service they are developing provides the required services to people of all kinds that have the legal rights to access that system. The system should at no point in time be biased with respect to gender, skin color, race or ethnicity and services provided should be equally provided in deserving proportions. An Intelligent system is only as good as the data used to train it. To make sure bias and inequality in intelligent systems is well taken care of, make sure to train it with good data containing all required labels.
  • Security: Security in AI involves the process of leveraging AI to identify and stop cyber threats with less human intervention than is typically expected or needed with traditional security approaches. The importance of this kind of security however is to protect the confidential data of anyone using that system. If any intelligent system rather compromises the security of users’ confidential data, that system needs to be checked.

How to Know If You Should Use AI

Artificial Intelligence and Machine Learning have been around for quite some time and I know everyone by now knows the potential and capabilities of these technologies. Okay, I’ll throw in some stats just in case! 😛

  • 54% of executives say AI solutions implemented in their businesses have already increased productivity.
  • 61% of business executives with an innovation strategy say they are using AI to identify opportunities in data that would otherwise be missed.
  • 36% of executives say their primary goal for AI is to free up workers to be more creative by automating tasks.

(Source)

The rapid growth of AI has created a yearning for developers and employers to implement Artificial Intelligent systems in their own projects in an attempt to achieve great things. Amazing feats have been achieved by doing this in the sense that, they can now do away with human-related errors, get better precision and accuracy at certain tasks, automate many jobs, predict future values and detect fraudulent activities and potential problems.

However, people forget that not all problems can be solved by AI and that some things are best left without the interference of AI.

Don’t Succumb to Machine Learning Madness

Image for post
Artificial intelligence will not destroy this planet, irresponsible human intelligence will. ― Abhijit Naskar

Artificial Intelligence and Machine learning are very useful, but they should not be in every project. Using Machine Learning to solve simple problems is like using a chainsaw to slice a loaf of bread. Some problems require simple programming logic to solve and others require a much complex solution. If a simple alternative exists to solve your problem, go for it. Do not waste time and so much effort to build an intelligent system to solve a problem you could easily solve with some OOP.

Also, before starting an AI project, make sure you have carefully gone through the concerns and potential risks that come with AI. Make sure you have a defined path around the concerns and risks and be sure to keep an open mind about the work you are doing.

But, If You Do Decide to Use AI, Follow These Rules

  • Define a path for implementing an AI system: Any organization or individual seeking to build or implement an AI system into a new or existing project should first and foremost identify the problem they are willing to solve with AI. AI is meant for problem solving, it’s not just any cool toy that you can slip into your project for aesthetic purposes. The next thing would be to get familiarized with the concepts of the specific branch of AI you will have to use to get your work done. when this is all done, plans for data collection and retention have to be put in place. It has been already established that any intelligent system can only be as intelligent as the data used to train it, your data collection and retention strategy should be top notch and very robust. Finally, whatever system you’re building or implementing should be taken slowly and carefully. The system should be well balanced to learn and grow with data inputs and solve problems its was built to solve.
  • Address the ethics of AI: With the many concerns about the use of AI, the most practical approach will be to adopt a set of principles generally called Ethical AI. These ethics are a standardized set of principles that ensure that whatever we use the power of AI for is generally acceptable and obeys all privacy laws regarding the use of AI. The adherence to ethics in AI also makes you stand out as a professional developer or organization.
  • Know what a good AI system looks like: Since the primary purpose of AI is to make work easier and faster, a good AI system would be one which meets the criteria of making work easier and faster, accomplishes complex tasks with impressive accuracy or precision and finally increases productivity in any given organization. This can be simply put by saying, a good AI system is any system that effortlessly accomplishes tasks it has been assigned and primarily solves the problems it was built to solve.

A Mediocre AI System Won’t Help You

So much has been said about the pros and cons of using AI in organizations and individual projects in this story and conclusions can be drawn that, although AI is an amazing aspect of technology, it can sometimes do more harm than good if not executed properly.

With all being said about AI, it should be noted that Artificial Intelligence is a really powerful technology and if its power is harnessed properly, lots of complex tasks that cannot be efficiently done by man can be carried out with ease by AI systems.

Image for post


Exclusive datasets, guides, and insights to your inbox.

Join 3,000 subscribers. GDPR and CCPA compliant.