Treat artificial intelligence (AI) like a human or alien being

But you tell it what kind of a person or alien it is.

In this publication, we explore the practical use of Generative Artificial Intelligence (Gen AI or simply AI) and how we can harness its ‘unique form of intelligence’ to drive our personal and organisational growth. With a focus on co-intelligence, we look at how we can innovate and enhance our knowledge in collaboration with AI.

To delve deeper into the topic, we turn to two leading sources:

1. Rajeev Kapur, an entrepreneur, speaker and author who focuses on the intersection of technology and business. In his book, AI made simple (Kapur, Rajeev. AI Made Simple: A Beginner’s Guide to Generative Intelligence — 2nd Edition. Rinity Media, 2024), Kapur highlights the ability of artificial intelligence to improve processes, make more informed decisions and create new business opportunities. He also emphasises the importance of “generative agents” in the effective use of AI.
2. Ethan Mollick, a Wharton professor and expert on innovation and entrepreneurship, is known for his work on the impact of artificial intelligence. In “Co-Intelligence: Living and Working with AI” Mollick, Ethan. Co-Intelligence: Living and Working with AI. USA. Penguin Publishing Group, 2024), Mollick explores the possibilities of AI and human creativity working together. He argues that this co-intelligence, based on a number of key principles, can lead to innovative and win-win solutions.

Generative Artificial Intelligence has the potential to revolutionise various fields by providing answers, solutions, insights and relevant information with unprecedented versatility and breadth. However, it is important to remember that the quality of its output is directly dependent on the accuracy and relevance of the data and parameters it is fed.

To improve the relationship and accuracy between humans and AI, Ethan Mollick stresses the importance of adhering to these key principles:

• First, “Always invite AI to the table”: involve AI in various activities, not only out of necessity, but also out of curiosity and to understand its capabilities and limitations.
• Second, “Be the human in the loop”: humans must actively participate in the process of learning and using AI models to improve their performance and ensure relevant results. This means recognising that AI is not infallible and can make significant mistakes, so it is crucial that humans monitor and correct the decisions and results generated.
• Finally, “Treat AI like a person”: interacting effectively with AI requires configuring it, giving it a personality and treating it as if it were an autonomous entity, an “alien being”. It is essential to communicate with it in a natural and personalised way, without forgetting the limits and the artificial nature of this technology.

The third principle, “Treat AI like a person”, suggests giving AI models a unique, anthropomorphic identity that is tailored to each use case. In other words, when we use an AI model that has been prepared with a specific context and profile, we are interacting with a personalised entity that provides us with relevant information and answers. We call this personalisation Sapiens_AI or SAI.

Context is important when working with IA. Context is like the background information the IA needs to fully understand a situation and respond appropriately. Without the right context, the AI may misinterpret things or give irrelevant answers. When we work with AI models without providing them with context and profile, we get generalities and generic answers. Unless you are using Apple’s new Personal Context Understanding capabilities, you need to provide some context to the AI you are using.

Apple Intelligence will know more about you than any other AI, because your phone knows more about you than any other device. That is why Apple’s privacy narrative is key. The service focuses on situations where knowing the user’s context is key, and secures the results (so the AI’s hallucinations don’t play tricks). In this way, it reduces the risk of failure and also places value on the user’s knowledge through the mobile phone.

Every AI model has an inner expert (thousands upon thousands) that can be brought to light with the right technique. The art of doing this is the key, and there are infinite ways to achieve this.

In this sense, prompts are the key mechanism for communicating with AI models and providing them with intentionality and context. Among other things, these snippets of text or information serve as instructions and context for the models, allowing them to understand our needs and goals. Prompts are used in the “text input area” or via conversational interfaces such as ChatGPT 4o.

It is therefore essential to use prompts in a systematic and structured way to give AI models entity (intentionality + context) and effectiveness. By creating specific prompts for each use case, called personalisation prompts or system prompts, we can generate SAIs (Sapiens Artificial Intelligence) that are tailored to our specific needs.

For example, we can create a personal trainer for people with reduced mobility, a paediatric dentist, a Python programmer in Arduino environments, a personal shopper specialising in weddings, a career coach for senior executives, a CEO in the electronic components sector, a product manager for circular financial services, and many more.

To give AI models a specific profile, we propose a structure of personalisation prompts that includes six key elements:

1. Identity: assigning a name to the model to create a more personal connection and facilitate interaction.
2. Role: define the specific function that the model should perform in a given task, adapting its behaviour and responses to the expectations and needs of the target audience.
3. Scope: define the application area in which the model will operate, including the type of task, the work environment or the relevant domain of knowledge. This will allow the model to tailor its responses and actions to be most valuable and useful in that specific context and role.
4. Constraints: define the constraints or rules that the model must follow during its operation, including ethical, legal or technical constraints. This ensures that the model operates safely and in accordance with established norms, and avoids undesirable or harmful behaviour.
5. Tone: defines the style of communication the model should adopt (formal, informal, friendly, professional, etc.).
6. Instructions: provide specific guidelines on how to carry out their tasks, including detailed steps, priorities or preferred methods. It allows you to assess the criteria and rationales involved in the process.

Although this structure may seem complex at first, it becomes simple and routine with minimal experience. By using these elements, we can create individual and effective AI models tailored to our specific needs.

This is where co-intelligence comes into play.

I hope you find this information useful and practical.

Now invite AI to your table and treat it with the respect it deserves!

I would love to hear your opinions and experiences on this topic, so feel free to leave a comment, or if there is a topic you would like us to discuss.

For now, we say goodbye with the promise of returning next week with more discoveries, reviews and opinions.

Best wishes and have a nice day.

Eduardo, GoodSapiens Team