This is the first in a series of articles helping to shed some light on the topic of Artificial Intelligence (commonly known as AI). This first article will focus on providing some insight into the different types of techniques that sit under the broader AI umbrella. We discuss the key differences between generative and non-generative AI and then explore a couple of examples of how partners and clients are using AI in real and applied use cases.
Subsequent articles will deep-dive into a few of our partner solutions. We also plan to discuss the ethics of AI within financial markets and services and then finally, in the last of the series, we will publish our thoughts on what is coming next and how AI will disrupt financial services in the future.
What do we mean by AI?
Let’s start here. Artificial Intelligence (AI) has recently become a seemingly inescapable buzzword dotted throughout corporate websites, emails, blogs and the news. As a general rule, when the mass media picks up on an idea or trend, a great deal of the content that is being pumped out tends to greatly simplify the idea's original message, summarising everything into one acronym whilst promoting the twin drivers of fear and greed in equal measure in order to maximise the impact and grab the reader’s attention. From the fear of losing our jobs, as machines learn at an extraordinary pace, to the creation of content in the blink of an eye and generating untold riches through complex trading programmes, AI suddenly appears to be everywhere.
The term 'AI' was originally conceptualised by John McCarthy, an American computer scientist, in 1956. The term broadly refers to the use of various techniques that mimic human intelligence and it is has the potential to increase efficiency and accuracy while reducing costs and time. However, AI as a broad term is very much a way of grouping a wide range of techniques and tools.
One of the drivers for establishing NextWave and focusing on financial technology was the exciting idea of using these emerging techniques to help our clients’ businesses succeed. Many of our alliance partners have been developing solutions with embedded AI for years – the difference now is that this technology is becoming much more visible and mainstream as Microsoft and Google, amongst others, launch solutions such as ChatGPT and CoPilot. These new platforms allow AI driven solutions and their (more limited) functionality to be made available immediately on your desktop.
What are the different types of AI?
When we mention AI, what are we really speaking about?
With a cursory scan, you will find that there are several different types of techniques that can be categorised under this heading. Among these are Machine Learning, Deep Learning, Natural Language Processing (NLP), Robotics, Knowledge Graphing, and Speech Recognition, to name only a few. We have identified at least 16 different methods and techniques that may be classified as AI.
We have included a short glossary at the end of this article for those of you searching for definitions and more understanding
Generative vs Non-Generative AI
Further complicating matters is the fact that there has also been a great deal of discussion regarding generative and non-generative AI.
Generative AI
This is a type of artificial intelligence within the machine learning (ML) category that can create new content including text, speech, imagery, audio and synthetic data. The rise in popularity of generative AI is due to its relatively simple interface and speed. It usually begins with a prompt from a human and then proceeds to generate content using various AI algorithms which are trained using "training data". The outputs these AI algorithms produce are new instances of this training data – for example, if you train a generative AI model using song lyrics from an artist such as Beyoncé, they should be capable of producing new, unique Beyoncé songs that are very similar to her style and lyrics but not identical to past songs. Simply put, the content produced by generative AI models can include text, images or sounds that are virtually indistinguishable from the training data provided.
Examples of where Generative AI is being used include:
It is largely the excitement around these new generative AI led solutions that is driving much of the interest in the topic of AI. Just take a quick look at the examples mentioned and ask yourselves if any of this is really new to you. Are you aware that AI is disrupting content generation (for example with Chat GPT)? Are you aware of the ability to now generate new music using generative AI? In most cases, you will have heard of a solution, and you will be aware (or at least have some awareness) of how it is being used and perhaps have an outline understanding of some of the impact that technology is now having on the workplace and life in general.
However, there are challenges with generative AI led solutions. Many of them draw on publicly available data, imagery or sounds and the models learn based on these inputs, which may or may not include copyrighted materials. The outputs of the models is also enhanced depending on the additional data it is fed, which again could be confidential or copyrighted. As such, at the end of June 2023, a court case against ChatGPT was launched that seeks to test out a novel legal theory – the case claims that OpenAI violated the rights of millions of internet users when it used copyrighted social media comments, blog posts, Wikipedia articles and family recipes.
In response to this we are now seeing a number of ‘’closed loop’’ Generative AI platforms being developed that only provide responses based on data within a company’s infrastructure. One of our partners, Squirro ,has recently released such a solution, SquirroGPT, which can be trialled for free by our clients.
You can read more about Squirro GPT and download a free 14-day trial here (note that after the 14 days a fee is payable).
Content Generation
ChatGPT is probably the best known tool. Download the OpenAI link, and you can either use a free version or an enhanced version for $20, then ask it specific questions and it will provide you with detailed answers. You can even create free, AI generated artwork using an art-generating AI model, DALL-E2, which creates illustrations based on your instructions.
Note Taking And Recording
Microsoft OneNote, Notion.AI, Firelflies.AI, Google Keep – these are examples of where AI is being deployed to support people when taking meeting notes for meetings, phone calls and the like.
Voice Generation
One of our favourites in this space is Synthesia, who have created a set of avatars that speak your text and create watchable content. This is useful for promoting your business, explaining your products and services, and supporting any training programs you may wish to run for your staff and clients. You can see how we have used Synthesia on our Resource Centre.
Non-generative AI
Non-generative AI refers to AI systems that are trained on existing data rather than generating new data or content. Within financial services, this type of AI has been used for many years in various use cases.
For example, our alliance partner, Quantexa, uses AI-led algorithms to analyse vast amounts of transaction data in real-time to detect unusual patterns or behaviour that may indicate fraudulent activities. AI helps financial institutions identify and block fraudulent transactions, detect money laundering, identify increased risk and even help organisations to better understand their supply chains – all with the aim of reducing financial losses and enhancing security. Watch this short film explaining how Quantexa is using AI to help financial services companies combat crime.
Combining the two
In a great example of where generative and non-generative AI are being combined, Quantexa has also recently announced Q Assist that uses Generative AI within their platform to help with investigations and even generate a SARS (Suspicious Activity Report).
Is It One Size Fits All?
It is important to note that both approaches have unique strengths and use cases, and they complement one another in various applications of artificial intelligence. It is very much a case of using the best technique for the best outcome, be it generative or non-generative, or even both together.
What’s Next?
In our next article, we will focus in a little more detail on our partners technologies and explain how different AI techniques are being used for some of the use cases we have mentioned.
Glossary: What are the different types of AI?
1, Machine Learning
Machine learning is a branch of AI and computer science which refers to the use and development of computer systems that are able to learn and adapt without following explicit instructions, by using algorithms and statistical models that imitate the way that humans learn to analyse and draw inferences from patterns in data.
2. Deep Learning
Deep learning is a type of machine learning based on artificial neural networks in which multiple layers of processing are used to extract progressively higher level features from data. Deep learning models can recognise complex patterns in pictures, text, sounds, and other data to produce accurate insights and predictions.
3. Natural Language Processing (NLP)
Natural Language Processing (NLP) refers to the branch of AI concerned with giving computers the ability to understand text and spoken words in a similar way to humans, giving computers the ability to interpret, manipulate and process human language in the form of text or voice data and to ‘understand’ its full meaning, complete with the speaker or writer’s intent and sentiment.
4.Robotics
Robotics is an interdisciplinary sector of science and engineering dedicated to the design, construction and use of machines called mechanical robots that replicate or substitute for human actions. Robotics can take on various forms – a robot may resemble a human or be in the form of a robotic application, such as robotic process automation, which simulates how humans engage with software to perform repetitive, rules-based tasks.
5. Expert systems
An expert system is a computer programme that uses AI to solve complex problems within a specialised domain that ordinarily requires human intelligence and to provide decision-making ability like a human expert. It performs this by extracting knowledge from its knowledge base using the reasoning and inference rules according to the user queries.
6. Knowledge Graphs
Knowledge graphs, also known as semantic networks, organise data from multiple sources, capture information about real world entities of interest (objects, events, situations, concepts) in a given domain or task and illustrate a connection or relationship between them. In AI and data science, knowledge graphs are used to facilitate access to and integration of data sources; add context and depth to other, more data-driven AI techniques such as machine learning; and serve as bridges between humans and systems, such as generating human-readable explanations, or, on a bigger scale, enabling intelligent systems for scientists and engineers.
7. Knowledge Representation & Reasoning
Knowledge representation and reasoning refers to the thinking of AI agents and how thinking contributes to intelligent behaviour of agents. It represents real world information in a way that a computer can understand and us this knowledge to solve complex problems such as medical diagnoses and communicating with people in natural language. It is also concerned with how we can represent knowledge in AI not just in terms of storing data in some databases, but also in terms of enabling intelligent machines to learn from that knowledge and experiences to behave as intelligently as a human being.8 (and 10). Speech Recognition
AI based speech recognition uses solutions like Natural Language Processing (NLP) to analyse human speech and then transform this vocal data into a digital software that can be processed by computer software. The digitised data is processed using NLP, Machine Learning (ML) and deep learning techniques. Examples of Speech recognition are in smartphones, smart homes and other voice-activated solutions which employ this digitised speech.
9. Computer Vision
Computer vision is a field of AI that enables computers and systems to derive meaningful information from digital images, videos and other visual inputs to take actions or make recommendations based on that information. AI enables computers to think and computer vision enables them to see, observe and understand. It is similar to human vision, however it can analyse thousands of products or processes per minute, including defects or issues, and can therefore quickly exceed human capabilities.
11. Virtual Agents & Chatbots
A virtual agent is a software programme that utilises AI to recognise human speech in the way it’s really used. They are used by customer support teams to automate repetitive task. A chatbot is a piece of software that has been programmed to recognise and respond to human speech, mimicking a conversation between two people. Chatbots are rules-based, meaning they are designed to understand and respond to selected keywords or phrases. When a person uses a keyword that is recognised, the chatbot replies with a preset answer. Both virtual agents and chatbots simulate human conversations and engage with people in a natural way.
However, virtual agents are enhanced with conversational AI including Natural Language Processing and natural language understanding to allow virtual agents to establish what a user is really asking (or their intent) regardless of the exact wording they use. They are programmed to understand the overall meaning of a text, rather than specific words, and therefore won’t get tripped up by synonyms, typos, or abbreviations. You can ask a virtual agent open-ended questions, and from the context of the conversation they’ll be able to give you a relevant answer.
12. Autonomous Applications
Autonomous applications are software systems that can perform tasks and make decisions without direct human control. This includes driverless vehicles, unmanned aerial drones, smart home systems and automated industrial processes.
13. Suggestor systems
Suggestor systems, or recommendation systems, are AI algorithms, commonly associated with machine learning, that use Big Data to suggest or recommend additional products to consumers, using past purchases, search history, demographic information, clicks, likes and other factors to do so. Suggestor systems help users discover products and services they might not have not found on their own.
14. Strategy/Game Playing
Game playing is an AI application that involves the development of computer programmes to play games such as chess, but they can be used for education and military training as well. The goal of simulating game playing scenarios is to develop algorithms that can learn how to play games and make decisions that will lead to winning outcomes. This will help develop effective, decision-making systems for real-world applications.
15. Swarm Intelligence
Swarm intelligence is an AI or Natural Intelligence technique that creates algorithms that use the cooperative and group behaviour of social organisms in nature (such as ants).It is based on studying collective behaviour in decentralised and self-organised systems and can be used in the acquisition of library items, communications, the categorisation of medical datasets, dynamic control, the planning of heating systems, and the tracking and prediction of moving objects.
16. Cognitive Computing
Cognitive computing is the use of computerised models to simulate human thought processes in complex situations where the answers may be ambiguous and uncertain. It attempts to make computers mimic the way the human brain works, such as understanding natural language and recognising objects in an image.
Moving forward...
The biggest challenge for FS firms is knowing how to effectively leverage technology and data at pace to stay relevant and to get ahead.
We would love to hear from you for an informal chat about your goals and to share some approaches to delivering enterprise business solutions and transformation faster, cheaper and more effectively.
