Introduction to Artificial Intelligence and Machine Learning
Artificial Intelligence (AI) is a field of computer science which makes a computer system that can mimic human intelligence. It is commonly defined as technology with which intelligent systems can be created, that simulate human intelligence. An AI system does not require pre-programming as it uses algorithms which work with its own intelligence. It also involves machine learning algorithms such as reinforcement learning algorithms and deep learning neural networks.
Machine Learning (ML) is about extracting knowledge from data. It is a subfield of AI which enables machines to ‘learn’ from past data or experiences without being explicitly programmed to do so. ML enables a computer system to make predictions or make decisions based on historical data without being programmed. It uses massive amounts of structured and semi-structured data to generate accurate results or make predictions based on that data. It works on algorithms which learn on their own using historical data.
Key Differences Between AI and Machine Learning
It also works only for specific domains. For instance, if a machine learning model is created to detect pictures of dogs, it will only result in dog images. If data on cats is provided, it will not respond. Online recommender systems, Google search algorithms, e-mail spam filters, and Facebook auto friend tagging suggestions are some examples of ML currently in use. AI is now used in various places such as in Siri, Google’s AlphaGo, and in chess playing, for example.
AI characteristics can be classified into three types: Weak AI, General AI, and Strong AI.
ML can also be divided into three types: Supervised learning, Unsupervised learning, and Reinforcement learning.
The future of AI is Strong AI, which is supposedly more intelligent than humans. On a broader level, AI is a bigger concept to create intelligent machines that can simulate human thinking capability and behaviour. ML, on the other hand, is an application or subset of AI that allows machines to learn from data without being explicitly programmed to do so. AI and ML are parts of computer science that correlate with each other. However, they are applied towards different goals.
To learn more on our Artificial Intelligence and Machine Learning module, you can explore the Digital Risk Manager (DRM™) Certification page for full program details, click here.
Goals of AI and Machine Learning
What is AI? The goal of AI is to make computer systems as smart as humans to solve complex problems, whereas the goal of ML is to train machines to learn from data so that they can give accurate output. AI systems and artificial intelligence tools are concerned with maximising chances of success; ML is concerned primarily about accuracy and patterns. While both AI and ML have proven beneficial, their use, maintenance and improvement come with risks that may not be immediately obvious.
The Main Risks of AI and Machine Learning
When it comes to using AI and ML, the main risks are the lack of strategy and know-how, poor data, over-fitting, and biased data. Machine learning can be an effective risk management tool but is also a risk itself. Users may lack the appropriate skills or experience to implement it. Research has found that the lack of talent with appropriate skill sets for AI work was one of the main barriers to adoption. Because machines depend on human-supplied training data to work, error-free data is imperative.
Overfitting in AI Models
Over-fitting happens when the AI model in use is able to model the training data too well. Although this may sound like a good thing, the model will not be successful when it comes to testing the data as it will not learn from the dataset. In a human context, this is like knowing or memorising the answers but not understanding the formulas.
Data Quality and Bias
The data used for training may also be biased. Data needs to be diverse. A lack of diverse data will produce skewed results.
Example: FaceApp
FaceApp used a specific set of algorithms inspired by biological neural networks to add smiles or show users how they’d look when they were older but this model had been fed few images of people with darker skin. It interpreted people with lighter skin tones as being more attractive. FaceApp’s CEO had to apologise; the outcome was unintended and was the result of training data bias. This is just one example of how AI applications can go wrong, and underscores the need to think about AI and ML from a risk perspective.
Human Factors and Misuse of AI
Managing AI and ML risks is complicated as algorithms are complex, and data – as illustrated above – can be biased. Essentially, humans are the biggest risk to manage. They have deliberately manipulated AI to create deepfakes, which are artificial intelligence images. For example; they have rigged tests to cheat the authorities, as in the VW vehicle emissions scandal; and used AI planning to traffick humans, and buy, sell and smuggle drugs. AI and ML by themselves are neither good nor bad; it’s just how they are misused and abused by humans that make them so.
Risk Management for AI and ML
Identifying AI and ML risks is challenging; AI and ML are often poorly understood by those without a background in data science. So how can these be used safely and thoughtfully from a risk perspective?
Frameworks and Guidelines
Organisations need to recognise that AI and ML have to be managed differently from traditional risks. The Deloitte AI Risk Management Framework for example, was created to identify and manage AI-related risks and controls. But many existing risk management tools created by organisations and standards bodies may still be applied.
Building a Risk-Aware Culture
These include organisation-wide efforts at engagement and widening employees’ understanding of what managing AI and ML risk entails. Everyone in the organisation needs to be involved in creating and following the policies and processes that support effective risk monitoring.
Assessment criteria must be developed that includes all aspects of the business and its stakeholders, from vendor relations to corporate culture. AI and ML are constantly evolving; organisations need to be up to date and study artificial intelligence developments in parallel.
Future Outlook of AI in Business
They also need to constantly reassess their risks and risk appetites, particularly where the organisation’s application of AI and ML may have expanded and incorporated more data. Reassessment and improvement of risk management where AI and ML are concerned will help hone the organisation’s ability to assess danger, and may increase its competitive advantage in the long run.
Many analysts see AI as a major performance booster, and note how it can turn data into a key driver of competitive advantage.
In fact, according to global surveys on AI, organisations across the industrial board have plans in the pipeline to increase AI use. Those that don’t may find themselves at a competitive disadvantage. Businesses using advanced AI today are already seeing value in the form of reduced costs, accelerated time-to-market, and improvements in customer retention.
To learn more on our Artificial Intelligence and Machine Learning module, you can explore the Digital Risk Manager (DRM™) Certification page for full program details, click here.
Effective risk management requires the right tools and systems to support decision making. If you are exploring next-generation Enterprise Risk Management and GRC platforms, TriasGRC offers solutions designed to strengthen resilience and improve efficiency. Visit TriasGRC to learn more.
