AI@Socitm

Artificial Intelligence (AI) is a transformative field of computer science that seeks to create intelligent systems capable of performing tasks that typically require human intelligence, machines that can independently perform tasks, learn from experience, and adapt. It encompasses technologies like machine learning, natural language processing, and robotic process automation.

Example: This could include visual perception, text generation, speech recognition or translation between languages.

Below is a list of different types of AI along with brief examples of how they are used, each type of AI has been designed to address specific challenges and applications ranging from dealing with repetitive specialized tasks, to the ability to understand natural language and natural language generation.

Autonomous Systems refer to AI-driven systems capable of performing tasks or making decisions without human intervention.

Examples: Self-driving cars and unmanned planes & drones.

Computer Vision is about teaching machines to interpret and understand visual information from images or videos.

Examples: Object recognition, facial recognition, autonomous vehicles, and medical imaging.

Data poisoning attacks occur when an attacker tampers with the data that an AI model is trained on to produce undesirable outcomes (both in terms of security and bias). As LLMs in particular are increasingly used to pass data to third-party applications and services, the risks from these attacks will grow, as described in the NCSC blog ‘Thinking about the security of AI systems’.

(Source: National Cyber Security Centre)

Deep Learning is a specialized branch of Machine Learning that involves neural networks with multiple layers (deep neural networks).

Examples: Image and speech recognition, natural language processing, and playing strategic games.

Expert Systems are AI systems designed to mimic the decision-making abilities of human experts in specific domains. They use knowledge representations and rule-based reasoning to provide expert-level advice and solutions.

Examples: Knowledge representations and rule-based reasoning to provide expert-level advice and solutions such as automated diagnosis

Generative AI (GenAI) is AI capable of generating new content, such as text, images or video. Large language models (LLMs) are an example of generative AI. 

(Source: National Cyber Security Centre)

LLMs are machine learning models are a type of generative AI that can comprehend and generate different styles of text that mimic content created by a human.

An LLM is where an algorithm has been trained on a large amount of text-based data, typically scraped from the open internet, and so covers web pages and – depending on the LLM – other sources such as scientific research, books or social media posts. This covers such a large volume of data that it’s not possible to filter all offensive or inaccurate content at ingest, and so ‘controversial’ content is likely to be included in its model.

The algorithms analyse the relationships between different words and turn that into a probability model. It is then possible to give the algorithm a ‘prompt’ (for example, by asking it a question), and it will provide an answer based on the relationships of the words in its model.

LLMs are undoubtedly impressive for their ability to generate a huge range of convincing content in multiple human and computer languages. However, they’re not magic, they’re not artificial general intelligence, and contain some serious flaws, including:

• they can get things wrong and ‘hallucinate’ incorrect facts
• they can be biased, are often gullible (in responding to leading questions, for example)
• they require huge compute resources and vast data to train from scratch
• they can be coaxed into creating toxic content and are prone to ‘injection attacks’

(Source: National Cyber Security Centre)

Most AI tools are built using ML techniques, which use algorithms that enable computer systems to find patterns in data, improve their performance on a specific task (or automatically solve problems) without having to be explicitly programmed by a human. ML enables a system to ‘learn’ for itself about how to derive information from data, with minimal supervision from a human developer. There are several types of learning, such as supervised learning, unsupervised learning, and reinforcement learning.

Examples: Learning and identifying plants, recognition of handwriting to convert it to text.

(Source: Norfolk County Council, National Cyber Security Centre)

Narrow AI refers to AI systems designed to perform specific tasks or solve problems with high competence but limited scope. These systems excel at a narrow set of functions and lack general intelligence.

Examples: Virtual assistants like Siri and Alexa, image recognition systems, and recommendation algorithms used in online shopping.

NLP enables computers to understand, interpret, and generate human language. It plays a crucial role in applications like language translation, sentiment analysis, text summarization, and chatbots.

Examples: ChatGPT, Microsoft CoPilot and Chatbots

Prompt engineering is the process of structuring an instruction that can be interpreted and understood by a generative AI model.

Prompt injection attacks are one of the most widely reported weaknesses in LLMs. This is when an attacker creates an input designed to make the model behave in an unintended way. This could involve causing it to generate offensive content, or reveal confidential information, or trigger unintended consequences in a system that accepts unchecked input.

(Source: National Cyber Security Centre)

Reinforcement Learning involves training AI agents to make decisions by interacting with an environment. The agent learns through a system of rewards and punishments, aiming to maximize cumulative rewards over time.

Examples: Teaching cars to drive around a predetermined track, Non player characters in computer games.

RPA is a technology that uses software robots or “bots” to automate repetitive, rule-based tasks and processes in business operations. These bots are not physical robots; instead, they are software applications that mimic human interactions with computer systems.

Examples: Processing large volumes of invoices following a set of rules

Transformative AI is an advanced AI system with transformative impact on society. One example is artificial general intelligence, the hypothetical concept of autonomous systems that learn to surpass human capabilities in most intellectual tasks.

(Source: National Cyber Security Centre)