Are we witnessing a new technology revolution?
Over the past decade there has been a steady increase in businesses diving into the AI, Data Science and Machine Learning hype doing what they can to hurry up and gain a competitive edge from what this technology has to offer.
But is it really having a massive impact on the world?
And what the heck is AI, Data Science and Machine Learning?
Thanks to movies and television shows there likely are a lot of misinformed people walking around with ideas in their head about what AI is or what they think it can do. And hyped up news reports don't exactly help clarify complicated subjects in short edited segments about the latest technology breakthroughs.
So what do you think AI, Data Science and Machine Learning are and what can they do to help your business and improve your life? This article will help to answer those questions.
Let's start with the term Artificial Intelligence.
Artificial Intelligence
The Oxford Languages dictionary defines artificial intelligence as:
"The theory and development of computer systems able to perform tasks that normally require human intelligence, such as visual perception, speech recognition, decision-making, and translation between languages."
There are various ways to implement artificial intelligence systems. Traditional computer programming (procedural, logical, etc.) has been used to implement what one could arguably call artificial intelligence. Executive decision support systems, business analytics/business intelligence systems have been used for decades, often created using traditional computer programming paradigms and statistical methods.
The most promising form of AI today comes in the form of machine learning algorithms (more on that below) that enable solving more complex problems than could realistically be solved with traditional computer programming paradigms in the past.
Artificial General Intelligence (AGI)
The term artificial general intelligence is more aligned with what we have seen in movies and television shows. Think human level capabilities of processing information, problem solving and pattern recognition applied to general situations.
While there is active research in this area that is has produced some quite impressive results, those results are likely more accurately classified as artificial narrow intelligence, such as robots that run, jump and talk and can carry on a basic conversation.
Artificial Narrow Intelligence (ANI)
Over the past decade there have been major breakthroughs in artificial narrow intelligence. This has been partly due to increases in processing power of computers, partly due to more efficient ways of implementing ANI systems, partly due to access to large amounts of data, and partly due to breakthroughs in new and innovative ways of implementing machine learning.
As the name suggests, ANI focuses on specific "narrow" tasks. And the capabilities of these systems are very impressive. Computer vision and natural language processing (among others) are two areas that have seen impressive advancements over the past decade. Some of the capabilities of these systems include:
- Robots that can run, jump walk and talk,
- Identify objects in a video feed at the pixel level, associating every pixel to an object,
- Summarize long text documents so well that it can be difficult to tell the summary was written by a computer algorithm,
- Consume a text body and then answer questions about that content when a person asks a question and the answers are so well formed they seem to come from an informed human familiar with the content of the text.
Machine Learning (ML)
The underlying technology that enables ANI is the machine learning algorithm. Machine learning algorithms are a paradigm for enabling computers to solve problems. Instead of providing a computer step-by-step instructions and logic on how to make decisions, like traditional computer programming, instead machine learning algorithms can learn by example.
The machine learning algorithms enable computers to identify patterns and then leverage them to make decisions. Fundamental truths are embedded in the big data assets that companies have access to. And machine learning algorithms can extract these truths and use them for autonomous decision making or as input into decision support systems that are the tools of business decision makers.
Data Science and The Data Scientist
Implementing Artificial Intelligence requires subject matter experts from many different domains: statistics, computer programming, cloud/system administration, dev ops, etc. These combined together form what is loosely called data science.
The team that implements and supports artificial intelligence systems contains many different roles. While the role data scientist does have some commonly accepted responsibilities, there often is a grey area where one might have the title data scientist but actually be in a role more accurately described as statistician, data engineer, data analyst, business analyst, machine learning engineer, computer programmer, systems engineer, or cloud engineer.
It has been said that a data scientist is one that knows more statistics than a typical computer/systems engineer and is a better computer/systems engineer than a typical statistician. The ideal data scientist for applied applications of AI and ML in business would likely be described as a critical thinking problem-solver well versed in statistics, an excellent computer/systems/dev-ops/security engineer, with a willing to learn, is open-minded, team-oriented, and has great communication and presentation skills.
The Citizen Data Scientist
Finding and hiring good data scientists is not an easy task. Artificial intelligence systems require subject matter experts with many different skillsets. Couple this with the fact that more and more companies have decided they need AI systems in their business. The shortage of subject matter experts capable of implementing AI systems has resulted in an increase of consulting services and products of varying capabilities coming to market that attempt to address demand.
This has simplified (in some cases) the requirements and roles necessary for an organization to implement some form of what could be considered artificial intelligence.
For example, there are out-of-the-box analytics products that promise artificial intelligence capabilities that can be used by non-technical users. The capabilities of these tools varies but some solutions promise users that don't have statistics and computer programming experience they will be able to mine data, implement machine learning pipelines and refine them over time. However, I personally have not seen any of these fully deliver on that promise. And such systems historically have not been very customizable.
A term for people in roles that can leverage the value of data and machine learning, without needing the formal training and skillsets typically required for AI/ML projects, is the citizen data scientist. While they don't have the formal training a data scientist might have, they learn some of the basics to be able to help their organization implement and benefit from artificial intelligence. Their primary role is often something else, such as marketing manager, business analyst, etc. But they are able to tap in to the data within their organizations in ways that adds value and extracts useful knowledge from the data.