In his recent Professional Adviser article, our Chief Commercial Officer Mike Morrow, goes back to basics on AI to help planners learn more about the technology and understand where it can augment advice business.
Read the full article below.
I know what you're thinking…not another artificial intelligence (AI) article. But bear with us… this one's different.
Rather than talk about efficiency, standardisation, streamlining, speed, summaries, security and all the rest, we are going back to basics. When it comes down to it, what actually is AI? Here goes…
Understanding the basics, beyond ChatGPT
In a nutshell, AI is where you simulate human intelligence via machines that are programmed to think and learn like humans. That means they can do tasks that usually require human intelligence - things like recognising voices or images, making decisions, and translating different languages.
This all began way before the introduction of ChatGPT. In many ways, AI's been part of our day-to-day lives for years. Every time you open Google Maps, a little bit of AI is helping you get to your destination. Do you use your face or your thumbprint to log in to your banking app? Yes, that's AI.
But there are different types of AI.
Narrow AI is what we're all most familiar with right now. It's designed to perform specific tasks at superhuman levels, but it can't do everything. For example, a system that recognises faces better than humans can't learn to drive a car.
General AI is what AI researchers are aiming for. This kind of AI will be able to perform a wide range of cognitive tasks and creative thinking as well as - or better than - humans. How or when we'll get there, and the potential ethical, safety and resource concerns are all part of an ongoing debate.
Super AI is, for now, purely theoretical. This is where AI is smarter than the smartest human, so could potentially lead to amazing advances in science, medicine and all kinds of other areas. This is uncharted territory, and the consequences are unknown.
Separating hype from reality
Since ChatGPT took the world by storm, it seems like every internet search leads to an answer from AI. But beware – there are lots of limitations to what AI is capable of:
- It won't do everything. Think about the narrow vs general AI capabilities
- It lacks human understanding and emotions. Think about the emotional side of advice
- It'll give you an answer, even if it's wrong. We've all seen and heard some funny hallucination stories. It's not foolproof
- There are very high costs to develop and maintain it
In this series, we'll help you learn more about AI and understand where it can augment advice businesses. By building the foundations in your business (like good data) and then starting to practice with small, easy win use cases (like meeting notes), you can begin using it to make your life easier - so it really is worth getting to grips with the terminology.
Here are a few key terms, with examples to help you recognise what they mean:
AI – This basically means teaching machines to act and learn like people. Example: Alexa or Siri on your iPhone – every time you talk to them, you're using (and teaching) AI.
Algorithm – A set of rules a computer follows to solve a problem. Example: A recipe is an algorithm for baking a cake.
Deep learning – This is a more advanced type of learning using brain-like layers to understand complex stuff. Example: Facebook recognising faces in your photos automatically.
Generative adversarial network (GAN) – Two machines: one creates fake things (like photos), the other checks if they look real. Example: Making realistic-looking fake faces that don't belong to any real person.
Machine learning (ML) – This is where we help machines learn from data, so they get better at tasks without someone having to programme every step. Example: Netflix learning what shows you like and recommending similar ones.
Natural language processing (NLP) – Helping computers understand and respond to human language. Example: ChatGPT having a conversation with you.
Neural network – A computer system that mimics how our brain works to find patterns. Example: A spam filter that learns to spot junk emails.
Reinforcement learning – Letting a machine learn by trial and error, rewarding it when it gets things right. Example: Teaching a robot to walk by giving it points when it stays upright.
Robotics bias – When the machine makes unfair decisions because the training data was one-sided. Example: A hiring algorithm that favours men because it learned from mostly male CVs.
Supervised learning – Teaching a machine with examples that already have correct answers. Example: Showing a computer pictures of cats and dogs, labelled as ‘cat' or ‘dog', so it learns the difference.
Transfer learning – Taking what a machine learned in one job and using it in a new one. Example: A model trained to recognise cars can be quickly adjusted to recognise trucks.
Unsupervised learning – The machine looks at data without any labels and tries to group it by patterns. Example: A music app grouping songs into genres without being told what each genre is.
Variance – How much a model's results change when using different data. Example: A recipe that gives very different cakes each time you follow it - it's not reliable.
Next up
In part two of Adventures in AI, we'll cover how to experiment safely with AI and keep your and your clients' data secure.
This article is for financial professionals only. Any information contained within is of a general nature and should not be construed as a form of personal recommendation or financial advice. Nor is the information to be considered an offer or solicitation to deal in any financial instrument or to engage in any investment service or activity.
Parmenion accepts no duty of care or liability for loss arising from any person acting, or refraining from acting, as a result of any information contained within this article. All investment carries risk. The value of investments, and the income from them, can go down as well as up and investors may get back less than they put in. Past performance is not a reliable indicator of future returns.
