What are the 4 types of artificial intelligence?

Early AI algorithms had one thing in common; they lacked memory and were purely reactional. Given a specific input, the output would always be the same.

That is the case with many machine learning models. Stemming from statistical math, these models were able to consider huge chunks of data, then produce a seemingly intelligent output. For instance, it is extremely difficult (if not impossible) to write a math formula for movie recommendations. But machine learning models were able to yield great results by looking at the purchase history of other customers. Solving that problem became one of the factors behind Netflix's success.

The same mechanism works for spam filters, which can statistically determine if the presence and density of certain words should raise a red flag.

This kind of AI is known as "reactional" or "reactive AI," and it works great -- even performing beyond human capacity in certain domains. Most notably, it defeated chess Grandmaster Garry Kasparov in 1997. However, reactive AI is also extremely limited.

In real life, many of our actions are not reactive -- in the first place, we may not have all information at hand to react on. Yet, we are masters of anticipation and can prepare for the unexpected, even based on imperfect information. This "imperfect information" scenario has been one of the target milestones in the evolution of AI and is necessary for a range of use cases from natural language understanding to self-driving cars.

For that reason, researchers worked to develop the next level of AI, which had the ability to remember and learn.

As mentioned earlier, in 2012 we witnessed the deep learning revolution. Based on our understanding of the brain's inner mechanisms, an algorithm was developed which was able to imitate the way our neurons connect. One of the characteristics of deep learning is that it gets smarter the more data it is trained on.

Deep learning dramatically improved AI's image recognition capabilities, and soon other kinds of AI algorithms were born, such as deep reinforcement learning.

These AI models were much better at absorbing the characteristics of their training data, but more importantly, they were able to improve over time.

One notable example is Google's AlphaStar project, which managed to defeat top professional players at the real-time strategy game StarCraft 2. The models were developed to work with imperfect information and the AI repeatedly played against itself to learn new strategies and perfect its decisions.

In the StarCraft game, the decision a player makes early in the game may have decisive effects later. As such, the AI had to be able to predict the outcome of its actions well in advance.

We witness the same concept in self-driving cars, where the AI must predict the trajectory of nearby cars in order to avoid collisions. In these systems, the AI is basing its actions on historical data. Needless to say, reactive machines were incapable of dealing with situations like these.

Despite all these advancements, AI still lags behind human intelligence. Most notably, it requires huge amounts of data to learn simple tasks. While the models can be retrained to advance and improve, changes to the environment the AI was trained on would force it into full retraining from scratch. For instance, consider a language: Once we learn a second language, learning a third and fourth become proportionally easier. For AI, it makes no difference.

That is the limitation of narrow AI -- it can become perfect at doing a specific task but fails miserably with the slightest alterations.

Theory of mind capability refers to the AI machine's ability to attribute mental states to other entities. The term is derived from psychology and requires the AI to infer the motives and intents of entities (e.g., their beliefs, emotions, goals).

Emotion AI, currently under development, aims to recognize, simulate, monitor and respond appropriately to human emotion by analyzing voice, image and other kinds of data. But this capability, while potentially invaluable in advertising, customer service, healthcare and many other areas, is still far from being an AI possessing theory of mind: The latter is not only capable of varying its treatment of human beings based on its ability to detect their emotional state, but is also able to understand them.

Indeed, "understanding," as it is generally defined, is one of AI's huge barriers. The type of AI that can generate a masterpiece portrait still has no clue what it has painted. It can generate long essays without understanding a word of what it has said. An AI that has reached the theory of mind state would have overcome this limitation.

The types of AI discussed above are precursors to self-aware or conscious machines, i.e., systems that are aware of their own internal state as well as that of others.  This essentially means an AI that is on par with human intelligence and can mimic the same emotions, desires or needs.

This is a very long-shot goal, for which we possess neither the algorithms nor the hardware.

Whether artificial general intelligence and self-aware AI are correlative is to be seen in the far future. We still know too little about the human brain to build an artificial one that is nearly as intelligent.

The fast-evolving nature of AI has resulted in numerous terms for the various flavors of AI that humans have invented so far and strive to invent. Additionally, not everyone agrees on what these terms refer to, contributing to the difficulty of understanding what AI can and can't do.

The following terms are pretty commonly used and often associated with the four AI types described in this article: 

• Narrow AI or weak AI: This is the type of AI that exists today. It is called narrow because it is trained to perform a single or narrow task, often far faster and better than humans can. "Weak" refers to the fact that the AI does not possess human-level, i.e., general intelligence. Examples of narrow AI include: chatbots, autonomous vehicles, Siri and Alexa, recommendation engines.
• Artificial general intelligence (AGI): Sometimes referred to as "strong AI," AGI is a type of as-yet unrealized multifaceted machine intelligence that can learn and understand as well as a human can.
• Artificial superintelligence: This refers to AI that is self-aware, with cognitive abilities that surpass that of humans.

Efficiency and productivity gains are two of the most-often cited benefits of implementing AI within the enterprise. The technology handles tasks at a pace and scale that humans can't match. At the same time, by removing such tasks from human workers' responsibilities, AI allows those workers to move to higher-value tasks that technology can't do. This allows organizations to minimize the costs associated with performing mundane, repeatable tasks that can be performed by technology while maximizing the talent of their human capital.

"CIOs need to see where AI can help functions do more with less time and less resources, so they can [enhance] the experience for employees and users alike," said Beena Ammanath, executive director of Deloitte AI Institute.

As fast as business moves in this digital age, AI will help it move even faster, said Karen Panetta, a fellow with the technical professional organization IEEE and Tufts University professor of electrical and computer engineering. AI enables shorter development cycles and cuts the time it takes to move from design to commercialization, and that shortened timeline in turn delivers better, and more immediate, ROI on development dollars.

Executives can use AI for business model expansion, said Chris Brahm, senior partner at Bain & Company, and leader of the firm's global Advanced Analytics practice.

"As you deploy data and analytics into the enterprise, it opens up new opportunities for businesses to participate in different areas," he explained.

For example, autonomous vehicle companies, with the reams of data they're collecting, could identify new revenue streams related to insurance, while an insurance company could apply AI ton its vast data stores to get into fleet management.

Delivering a positive customer experience has become the price of doing business, said Seth Earley, author of The AI-Powered Enterprise and CEO of Earley Information Science.

"We're trying to embody everything we know about the customer, the customer's needs, our solutions and the competition and then present to the customer what they need when they need it," Earley said. "If we had a salesperson who could do that for everyone, that would be great, but we don't."

AI, however, can do all that and more, leading to more customized and personalized interactions between organizations and each individual customer.

AI's capacity to take in and process massive amounts of data in real time means organizations can implement near-instantaneous monitoring capabilities that have the capacity to alert them to issues, recommend action and, in some cases, to even initiate a response, Ammanath said.

For example, AI can take information gathered by devices on factory equipment to identify problems in those machines as well as predict what maintenance will be needed when, thereby preventing costly and disruptive breakdowns as well as the cost of maintenance work performed because it's scheduled rather than because it's clearly needed.

AI's monitoring capabilities can be similarly effective in other areas, such as in enterprise cybersecurity operations where large amounts of data needs to be analyzed and understood.

Organizations can expect a reduction of errors as well as stronger adherence to established standards when they add AI technologies to processes, according to Madhu Bhattacharyya, managing director and global leader of Protiviti's Enterprise Data and Analytics practice. When AI and machine learning are integrated with a technology like RPA, which automates repetitive, rules-based tasks, the combination not only speeds up processes and reduces errors but can also be trained to improve upon itself and take on broader tasks.

The use of AI in financial reconciliation, for example, would deliver error-free results whereas that same reconciliation when handled, even in part, by human employees is prone to mistakes. "Can you maintain better quality with AI? Yes, you can," Bhattacharyya said.

Companies are using AI to improve many aspects of talent management, from streamlining the hiring process to rooting out bias in corporate communications. Writing about the growing use of AI in recruitment, independent consultant Katherine Jones said AI-enabled processes not only can save companies in hiring costs but also impact workforce productivity by successfully sourcing, screening and identifying top-tier candidates. As natural language processing tools have improved, companies are also using chatbots to provide job candidates with a personalized experience and mentor employees. Additionally, AI tools are being used to gauge employee sentiment, identify and retain high-performers, and determine equitable pay.

In addition to the benefits listed above, AI can fuel numerous industry-specific improvements. Here are three examples, from Shervin Khodabandeh, a managing director and senior partner at Boston Consulting Group and co-leader of its AI business in North America:

• Retailers can use AI to better target their marketing efforts, develop a more efficient supply chain and better calculate pricing for optimal returns. At retail companies where humans do the majority of the work, AI will help predict customer requirements and appropriate staffing levels.
• The pharmaceutical sector can use the technology to perform drug-discovery data analysis and predictions that can't be done with conventional technologies.
• The financial industry can use AI to strengthen its fraud detection efforts.

It's important to remember that as companies find ways to use AI for competitive advantage, they are also grappling with challenges. Concerns include AI bias, government regulation of AI, managing the data required for machine learning projects and talent shortages. In addition, financial gains can be elusive if the talent and infrastructure for doing AI are not in place, according to research done by MIT Sloan Management Review and Boston Consulting Group.

Read about the major risks associated with AI in this article.

Practical conversations about AI require a basic understanding of how data powers the entire process. "Data fluency is a real and challenging barrier -- more than tools or technology combined," said Penny Wand, technology director at IT consultancy West Monroe. In a 2020 report, Forrester Research found that 90% of data and analytics decision-makers surveyed see increased use of data insights as a business priority, yet 91% admitted that using those insights is a challenge for their organizations. Forrester further reported that the gap between recognizing the importance of insights and actually applying them is largely due to a lack of the advanced analytics skills necessary to drive business outcomes. "Executive understanding and support," Wand noted, "will be required to understand this maturation process and drive sustained change."

"To successfully implement AI, it is critical to learn what others are doing inside and outside your industry to spark interest and inspire action," Wand explained. When devising an AI implementation, identify top use cases and assess their value and feasibility. In addition, consider your influencers and who should become champions of the project, identify external data sources, determine how you might monetize your own data externally, and create a backlog to ensure that the project's momentum is maintained.

Focus on business areas with high variability and significant payoff, advised Suketu Gandhi, a partner at digital transformation consultancy Kearney. Teams comprising business stakeholders who have technology and data expertise should use metrics to measure the impact of an AI implementation on the organization and its people.

Once use cases are identified and prioritized, business teams need to map out how these applications align with your company's existing technology and human resources. Education and training might help bridge the technical skills gap internally, while corporate partners can facilitate on-the-job training. Meantime, outside expertise could help accelerate promising AI applications.

It's important to narrow a broad opportunity to a practical AI deployment -- for example, invoice matching, IoT-based facial recognition, predictive maintenance on legacy systems or customer buying habits. "Be experimental," Carey said, "and include as many people [in the process] as you can."

To turn a candidate for AI implementation into an actual project, Gandhi believes a team of AI, data and business process experts is needed to gather data, develop algorithms, deploy scientifically controlled releases and measure impact and risk. 

The successes and failures of early AI projects can help increase understanding across the entire company. "Ensure you keep the humans in the loop to build trust, and engage your business and process experts with your data scientists," Wand said. Also recognize that the path to AI starts with understanding the data and good old-fashioned rearview mirror reporting to establish a baseline of understanding. Once a baseline is established, it's easier to see how the actual AI deployment proves or disproves the initial hypothesis.

The overall process of creating momentum for an AI deployment begins with achieving small victories, Carey reasoned. Incremental wins can help build confidence across the organization and inspire more stakeholders to pursue similar AI implementation experiments from a stronger, more established baseline. "Adjust algorithms and business processes for scaled release," Gandhi suggested. "Embed [them] into normal business and technical operations."

As AI projects scale, business teams need to improve the overall lifecycle of AI development, testing and deployment. To ensure sustained success, Wand offers three core practices for maturing overall project capabilities:

• Build a modern data platform that streamlines how to collect, store and structure data for reporting and analytical insights based on data source value and desired key performance indicators for businesses.
• Develop an organizational design that establishes business priorities and supports agile development of data governance and modern data platforms to drive business goals and decision-making.
• Create and build the overall management, ownership, processes and technology necessary to manage critical data elements focused on customers, suppliers and members.

Once the overall system is in place, business teams need to identify opportunities for continuous improvement in AI models and processes. AI models can degrade over time or in response to rapid changes caused by disruptions like the COVID-19 pandemic. Teams also need to monitor feedback and resistance to an AI deployment from employees, customers and partners.

During each step of the AI implementation process, problems will arise. "The harder challenges are the human ones, which has always been the case with technology," Wand said.

A steering committee vested in the outcome and representing the firm's primary functional areas should be established, she added. Instituting organizational change management techniques to encourage data literacy and trust among stakeholders can go a long way toward overcoming "human" challenges.

"AI capability can only mature as fast as your overall data management maturity," Wand advised, "so create and execute a roadmap to move these capabilities in parallel."