Navigating the AI Revolution: A practical guide for business

Introduction - Part 1 of 4

As a technology leader I play a key role in helping businesses and colleagues in the exec understand how to get to grips with new technologies and the capabilities they can provide for us. I have spent the last 15-20 years helping organisations in different industries not only understand this, but turn that understanding into practical plans that drive real value.

In that time I have never come across a more disruptive technology than AI.

When asked by my peers to explain AI and how we can use this tool to help solve business problems or drive opportunities I have had to break the challenge down and look at it through three lenses;

1. Preparation and Utilization: How do we prepare for and leverage AI to benefit not just customers but also colleagues, shareholders, and other stakeholders?
2. Internal Applications: How can we employ AI tools internally to enhance efficiency, security, creativity, and overall value?
3. Partnerships: How do we collaborate with partners who use AI and provide AI enabled solutions in a way that aligns with our ethical standards and business goals?

I intend to cover each of the points above over 4 articles, this being the first. Hopefully, they will serve as a starting point for other business and technology leaders embarking on this exciting new journey.

Firstly, in this article I want to provide a simple high level guide to what we mean by AI, where we are now and key considerations before we move to looking at how we can start to apply this in the following articles.

A brief overview of AI

Artificial Intelligence is a field of computer science focused on creating systems capable of performing tasks that would normally require human intelligence. These tasks include learning, reasoning, problem-solving, perception, and language understanding.

Types of AI

1. Reactionary AI: These are the most basic forms of AI and are designed to perform a single, narrow task. Examples include chatbots and automated customer service or process.
2. Limited Memory AI: This is where we are currently at. These AI systems can learn from historical data to make better decisions. For instance, machine learning algorithms used in predictive maintenance fall under this category. You can see the underlying types of learning algorithms.
3. Theory of Mind AI: These would be AI systems that understand human emotions, beliefs, and can interact socially. They are the subject of much debate and research, but as yet do not exist.
4. Self-Aware AI: This is the final stage of AI development, also referred to as Artificial General Intelligence and would involve machines that have their own consciousness. This is way more of a concept than a reality at this point.? Once these arrive humans are no longer the most intelligent species on the planet. A sobering thought.? Timelines and estimations vary wildly on when this will happen.? Given the rapid development to date, we really don’t know.

Where are we now

There is lots of reading and research around this but most agree the current tools, including generative large language models (ChatGPT et al), are at stage 2 - Limited Memory AI. Stage 3 is very much work in progress in terms of research with stage 4 still in the realm of sci-fi.

AI, true AI is not programmed in the traditional sense. It is taught, it can adjust itself to improve over time as it learns. It does this by adjusting 'parameters' in the algorithm, and decides how best to adjust these parameters to complete a given task or tasks to a suitable standard.? The more complex the task the more it needs data and time to learn which parameters to adjust.? To put this into context LLaMa, Facebooks AI can adjust 65 billion parameters in its model.? Chat GPT 3 175billion……ChatGPT 4 1.7 trillion.? The other key term to understand in this new world is ‘Tokens’ these are units of meaning such as a word that goes into the training data.? Different models have different volumes of training data, again we are talking billions or trillions of tokens.

AI Analogy

To make this easier to understand, let's use an analogy involving a child and wooden blocks. Imagine you're teaching a child how to fit different-shaped wooden blocks, which we'll refer to as the tokens, into corresponding holes. You don't need to explain the parameters of how to move their arm or grip their fingers to pick up a square token; they figure that out intuitively. Once the child learns that the square block fits into the square hole, they can apply the same logic to other shapes, like fitting a round block into a round hole.

AI, especially stage 2 or Limited Memory AI, functions in a similar way. It uses past experiences and data to refine the parameters of its performance, improving how it handles different tokens or tasks. However, there are some key nuances to keep in mind:

• Volume of Data: Unlike a child, who can learn from just one or a few experiences, AI usually requires a large volume of data to adjust its parameters effectively. The more data it has, the better it can optimise its performance.
• Task-Specific Learning: While a child can generalise, still a human USP, the concept of "fitting tokens into corresponding holes" to a wide range of activities, Limited Memory AI is often optimised for specific tasks. It adjusts its parameters to do one thing really well but may not be able to apply that learning to a broader set of problems.

By understanding these nuances, we can better appreciate both the capabilities and limitations of current AI technologies.

As we race to adopt these capabilities, and I will go into detail on how to do this for business in the subsequent articles, this does need to be balanced with some careful considerations.

Headline considerations for business

Ethical Considerations: While AI offers numerous advantages, it also raises ethical questions that organisations must address. These include data privacy, algorithmic bias, and the potential for misuse. As we integrate AI into our operations, it's crucial to consider these aspects and develop ethical guidelines that align with our business values and societal norms.

Regulatory Landscape: Understanding the regulatory landscape around AI is essential for businesses. Regulations like GDPR have implications for data privacy and AI. Being aware of these laws can help businesses stay compliant and avoid legal pitfalls.

Mis-selling: Last point is on snake oil, there is a real lack of understand about AI and some people, are going to use this current gap in knowledge to try and up sell their products and capabilities, we need to guard against that. Please involve your friendly tech team in any AI supplier led discussions. If you do find yourself alone keep an eye out for these red flags;

• Overpromising Results: If a vendor promises quick, extraordinary results without a clear explanation or proof of how their AI solution achieves this, be cautious.
• Lack of Transparency: A credible AI vendor should be able to explain how their model works, the data it was trained on, and its limitations. Be wary of those who can't or won't disclose this information.
• Ignoring Data Privacy: If the vendor dismisses concerns about data privacy and doesn't explain how the data will be used or stored, it's a major red flag.
• No Customisation: AI solutions often need to be tailored to specific business needs. A vendor who offers a one-size-fits-all solution without considering your unique requirements should be questioned.
• High-Pressure Sales Tactics: Be cautious if the vendor is pushing for a quick commitment without giving you adequate time to evaluate the solution or consult with your tech team.
• No Trial or Demo: A reputable AI company should offer a trial period or demo to let you evaluate the product. If this isn't available, it could be a sign that the product doesn't deliver on its promises.

Overall though I am on the sunny side of AI and what it will do for us. Today I am ploughing a field with a horse, one field a day and bloody hard work, tomorrow I am doing multiple fields a day in a big shiny tractor!

I wont be replaced by AI, however if I am not careful I will be replaced by a human using AI. In my next article, 2 of 4, I will cover how businesses can prepare for AI enabled humans in our world.

For complete transparency I did co-author this with an AI friend.