Artificial Intelligence (AI) is stirring up a storm in many sectors, in no other sector is artificial intelligence having more of an impact than on manufacturing.

What is artificial intelligence?

Artificial intelligence (AI) is a division of computer science focused on building smart machines capable of performing intelligent tasks that require human intervention. Artificial intelligence is the combination of machine learning, deep learning, neural networks and data mining - all seen championing the business world.

Artificial intelligence definition:

Artificial intelligence (AI),  the ability of a digital computer or computer-controlled robot to perform tasks associated with intelligent beings. Encyclopaedia Britannica

How does Artificial Intelligence work?

AI performs depending on the assigned task. This includes unique implementations and strategies, but isolated as two main approaches:   

Rules-based:

This approach uses algorithms — a series of direct instructions used by computers to solve problems. The chosen algorithm will determine and direct the AI on how to approach the problem it faces. Each defined algorithm will have different goals, strengths, weaknesses and approaches, depending on the desired result and problem at stake.

Examples based:

An examples-based approach creates models out of data by finding patterns. Models in this case can be examples of visualisations or predictions.  Some examples of the data used in these scenarios include: user profiles, data logs, transactions, weather reports, etc. The AI is "fed" data, ans using machine learning it gains a general understanding of the project by identifying patterns in the data set.

The types of artificial intelligence:

AI programming focuses on cognitive skills - characterised as four types of AI: reactive, limited memory, theory of mind and self-aware. The processes to achieve these include:

Learning processes: Acquiring data and creating rules which turn the data into actionable information. These provide the machine with actionable instructions to complete specific tasks.

Reasoning processes: Determining the correct algorithm (path) to achieve the desired outcome.

Self-correction processes: A process of continuous adjustment, fine-tuning and developing the algorithm to ensure the desired results.

Four types of artificial intelligence:

In a 2016 article Arend Hintze, an assistant professor of integrative biology and computer science and engineering at Michigan State University, explained that AI can categorise into the following four types: 

  • Reactive machines: These AI systems are task specific and have no memory. These systems cannot use past experiences to inform future ones. A notable example includes, IBM's Deep Blue - A chess program that beat Garry Kasparov in 1996 and 1997.
  • Limited memory: Systems that have partial memory to inform future decisions. Used in self driving cars, which learn though various data inputs such as LiDAR and cameras.
  • Theory of mind: AI systems which can identify human emotions and predict human behaviour. The social intelligence these systems achieve is pivotal for effective interaction with human teams.
  • Self-awareness: Yet to exist, these systems have a sense of consciousness which allows the machine to understand its current state.

History of artificial intelligence:

The following is some of the most important events in AI.

1943 

  • "A Logical Calculus of Ideas Immanent in Nervous Activity." A paper by Warren McCullough and Walter Pitts proposed the first mathematic model for building a neural network.

1949

  • The theory that neural pathways generate from experiences and that connections between neurons become stronger the more frequently used, proposed by Donald Hebb.

1952

  • Self-learning program to play checkers, developed by Arthur Samuel.

1954

  • IBM machine automatically translates 60 Russian sentences into English.

1956

  • Artificial Intelligence (AI) is first introduced in the Dartmouth Summit Research Project on Artificial intelligence where the goals of AI are outlined.

1972

  • The inception of the programming language PROLOG.

1985

  • More than a billion dollars a year is being spent on expert systems which forms the industry known as the Lisp machine market.

1987-1993

  • Cheaper alternatives emerged and the Lisp machine market collapsed in 1987.
  • The end of the DARPA led Strategic Computing Initiative in 1993 after almost $1 billion of spending and falling far short of expectations.

1991

  • Deployment of DART by U.S forces, an automated logistics planning and scheduling tool, during the Gulf War.

1997

  • World chess champion Gary Kasparov, beaten by IBM's Deep Blue.

2005

  • Investment in autonomous robots like Boston Dynamic's "Big Dog" and iRobot's "PackBot." is increased by he U.S. military.

2008

  • Breakthroughs in speech recognition applications such as, google speech.

2014

  • Google's self-driving car passes a human driving test.

2016

  • Google DeepMind's AlphaGo defeats world champion Go player Lee Sedol.

Weak AI versus strong AI:

Weak AI, also known as narrow AI, excels on preforming one task extremely well, such as answering questions or playing chess. Strong AI, can preform a range of tasks, and eventually teach itself to solve new problems.

Weak AI requires human input and relies on it to define its working parameter. Strong AI in comparison, does not require human intervention, thus over time, strong AI develops consciousness to improve its learning. 

What are the applications of AI technology? 

  • Automation: Automation tools can expand the yield and types of tasks performed. Robotic process automation automates rules-based data processing tasks that are often repetitive.
  • Machine learning: The methodology of computers to perform without programming. Implied as the automation of predictive analytics. There are three types of machine learning algorithms which include:
  • Supervised learning: Labelled datasets to detect patterns in new data sets.
  • Unsupervised learning: Datasets sorted by similarities and differences.
  • Reinforcement learning: Datasets are not labelled, but feedback presented to the AI after actions performed.
  • Machine vision: Machine vision captures and analyses visual information using a camera. This includes analog-to-digital conversion and digital signal processing.
  • Natural language processing: The processing of human language by a computer program. For example, spam detection in email inboxes or text/speech translation systems.
  • Robotics: Robots that can perform tasks that are too difficult or dangerous for humans to perform. Use cases include, manufacturing assembly lines - robots used to spot-weld, paint and manoeuvre new vehicles.
  • Autonomous vehicles: A combination of computer vision, deep learning and recognition, these vehicles build the skills required to keep a vehicle piloted on public roads while avoiding obstacles.

Artificial intelligence examples:

  • AI in healthcare: Application of machine learning to make better and faster diagnoses than humans - the IBM Watson. Questions taken by the machine, the system mines patient data and other available data sources to form a hypothesis, which it then presents. A more accessible technology is AI chat-bots which can help patients find medical information, schedule appointments and reduce pressure on healthcare professionals.
  • AI in business: ML algorithms can integrate into analytics to uncover information to serve clients better. For example, chat-bots incorporated on websites to serve customers, on queries and problems.
  • AI in education: AI can automate grading which could reduce the burden on teachers, giving them more time to teach. AI as a tool for additional tutoring to provide additional support to students.
  • AI in finance: Programs that can collect data and generate informed advice reports. This could reduce the need for accountants in the future. Currently, the majority of financial trading is algorithmic.
  • AI in law: Research through documentation automated by AI to help this labour-intensive task in the law industry to better client experience and outcome. Some law firms have used machine learning to predict outcomes and extract information out of documents.
  • AI in manufacturing: Robots programmed to perform repetitive, highly skilled tasks at 24/7, whilst maintaining efficiency and accuracy. Robots are becoming multi-skilled and programmed to interact with their human colleagues.
  • AI in banking: The use of chat-bots to serve customers, allows banks to cut costs while maintaining a good level of customer service. Banks are also using AI to improve decision-making on loans, and identifying investment opportunities through AI models.
  • AI in transportation: AI used to manage traffic, improve cargo efficiencies, improve safety and the foundation of operation in autonomous vehicles.

How important is artificial intelligence?

  • Automation of repetitive learning through data: Algorithms can perform high volume, technical computer tasks without fatigue at a large scale. Similar to how robots automate manual tasks.
  • Adding and increasing intelligence: Products will become "smart" with the adoption of AI. Many of these algorithms combined with a large amount of data to provide automation, intelligence and analysis to improve their capabilities. Examples include Siri and Alexa smart AI's.
  • Adaption through progressive learning algorithms: The AI identifies the structure and patterns of the data which it can develop a skill from. Irregularities and patterns could help the AI teach itself new skills - the YouTube recommendation algorithm works in this way. These models continue to adapt as they gather more data.
  • Analysis of deep data: Using neural networks, skills can be developed by the AI to identify outliers in data. Bigger datasets fed into the algorithm improve their accuracy. This method is often used to build fraud detection systems. This analysis deep data can help with problem solving and one of the cornerstones of AI research.

What are the advantages of artificial intelligence?

  • AI reduces the time taken to perform a task. It enables multi-tasking and eases the workload for existing resources. This can lead to reduced costs and greater efficiency.
  • AI  can operate 24x7 without interruption.
  • AI is industry agnostic with mass market potential - deployed across industries.
  • AI facilitates decision-making by making the process faster and smarter through modelled decisions.

What are the disadvantages of artificial intelligence?

  • High initial cost  
  • No human replication  of tasks
  • Little to no improvement with experience  
  • The lack of creativity in the AI  
  • Unemployment as jobs replaced by AI

Artificial intelligence in manufacturing process:

The advent of artificial intelligence for manufacturing will revolutionise the entire sector. This includes adding value intelligent supply chains through: mass customisation, efficiency, predictive maintenance and raw material stock analysis to name a few.

The pandemic has allowed a hard reset and a chance for manufacturers to implement advanced artificial intelligence into their supply chain. What are the use cases of AI in manufacturing?

Applying AI to manufacturing can reduce running costs significantly.

Predictive maintenance allows manufacturers to predict when machines need maintenance with a high degree of accuracy. Advanced sensors and analytics embedded into the machinery will allow manufacturers to prevent unplanned downtime by using machine learning. For example, predicting the optimal time to replace replace or resharpen tools in the manufacturing chain.

Supply chain management will be able to benefit from using big data. Manufacturers collect a whole host of data, but often struggle with making sense of it all. With intelligent AI, manufacturers will be able to unlock the insights that were previously unreachable.

Improvements in robotics, from traditional pre-programmed robots to AI-powered robots which can interpret data efficiently and make decisions on their own. This alone will not only save time, but manufacture to a better standard. This is because AI powered robots can interpret data from CAD models and schematics to identify the most efficient and appropriate way to complete the task without any pre-programming.

Mass customisation is a manufacturing technique that enables personalisation of custom-made products at scale. AI provides the possibility of creating customised solutions for development and production through greater efficiency in the form of modular production. AI algorithms can formulate estimations and implement changes by looking at patterns in consumer behaviour, socio-economic and macroeconomic factors, location and more. This not only means a more personalised solution for the customer, but optimisations and adaptations made independently by the AI.

The manufacturing sector is the perfect candidate for artificial intelligence. As the technology matures and costs drop, expect to see more manufacturers utilising AI as it becomes more accessible. The future of AI is certain in an industry known for embracing new and emerging technologies. Manufacturers are consistently looking to minimise cost, make data-driven decisions, optimise operations and deliver a better product to their consumers.

How close are we to artificial intelligence?

How far away are we from a future surrounded by artificial intelligence? Most of the methodologies and applications of AI mentioned are possible today, the algorithms just have to be applied. What we are yet to see is the successful execution of  artificial general intelligence.

The mammoth task of replicating the processing power of a human brain to achieve artificial general intelligence is pending and quantum computing is likely to be our portal. It is likely we will begin to see the first AGI systems in 2030 and could take up to 2060 to see AI with the consciousness of a human.