Andrew Brookes, CTO and Co-founder
Once upon a time, in the world of aviation, a brilliant idea was born. Engineers envisioned a system that could assist pilots during their flights, reducing workload and increasing safety. This ingenious invention was aptly named “autopilot.” However, as time went on and technological advancements continued, the capabilities of autopilot systems grew, and so did the reliance of pilots on these systems.
As a consequence of this over-reliance, there have been instances where pilots’ situational awareness and decision-making skills were compromised, leading to incidents that could have been prevented with more active human involvement. While autopilot can be an invaluable tool, the reality is that it shouldn’t replace the human element in flying. Pilots must maintain a level of awareness and engagement to ensure the safety and success of each flight.
Now, let’s fast forward to the present day, where another innovative creation has captured our attention: large language models, like OpenAI’s GPT-4. These models have the incredible ability to generate human-like text, and their applications seem almost limitless. But, if we are to take lessons from autopilot, it is essential to understand when to use these AI models and when not to use them.
Enter the concept of the “co-pilot”, which is essentially what such tools are. A co-pilot is more than just an assistant; it is a partner, working alongside the pilot to offer insights, support and guidance. In the realm of AI, a co-pilot is an AI model that complements and enhances human capabilities, rather than attempting to replace them entirely. This distinction is crucial, as it helps us navigate the ever-evolving landscape of AI technology and its role in our lives.
So, when should we use our AI co-pilot, and when should we rely on our own expertise? Let’s explore a few scenarios to illustrate this delicate balance.
Scenario 1: Writing a Report
Imagine you’re tasked with writing a detailed report on a subject you’re not well-versed in. Your AI co-pilot can be an invaluable asset in this situation, providing you with relevant information, assisting with phrasing, and even generating an initial draft for you to work on. However, it’s essential to remember that the AI model may not always be accurate or fully understand the nuances of the subject. In this case, the AI co-pilot serves as a starting point, while you, the human expert, are responsible for fact-checking, refining, and polishing the final product.
Scenario 2: Brainstorming Creative Ideas
In a creative brainstorming session, your AI co-pilot can offer a plethora of ideas and suggestions to spark your imagination. It can help you explore different angles and perspectives that you may not have considered otherwise. However, it is crucial to recognise that AI-generated ideas may lack the depth and understanding of human emotions and experiences. As the human collaborator, you bring your unique perspective and emotional intelligence to the table, elevating the AI-generated ideas into something truly remarkable.
Scenario 3: Technical Troubleshooting
When faced with a complex technical problem, your AI co-pilot can provide you with potential solutions and even guide you through the troubleshooting process. However, the AI’s knowledge is limited to the data it has been trained on. In situations where the issue is unique or requires out-of-the-box thinking, human intuition and expertise become indispensable. The AI co-pilot can offer suggestions, but it is ultimately up to you, the human expert, to make the final decision and implement the solution.
Scenario 4: Dealing with Sensitive Topics
LLMs are trained on vast amounts of data from the internet, which means they may have been exposed to biassed, offensive or controversial content. When dealing with sensitive topics, LLMs might inadvertently generate inappropriate or offensive language. In such cases, it’s essential to rely on human judgement and empathy to ensure that the content created is respectful, accurate, and compassionate.
Scenario 5: Legal and Compliance Matters
When working on legal documents or navigating regulatory compliance, precision and accuracy are of utmost importance. LLMs are not always reliable in generating text that adheres to specific legal requirements or captures the nuances of complex regulations. In these situations, it’s crucial to rely on human experts with a deep understanding of the law and the ability to interpret and apply it correctly.
Scenario 6: Situations Requiring Personal Experiences and Emotions
Large language models, while impressive in their capabilities, do lack personal experiences and emotions. In scenarios where a deep understanding of human feelings or a personal touch is required, such as writing a heartfelt letter or creating a eulogy, LLMs may fall short. In these cases, it’s essential to rely on our own emotional intelligence and personal experiences to create meaningful and authentic content.
Scenario 7: Handling Confidential Information
When dealing with confidential or sensitive data, it’s essential to maintain strict security and privacy measures. LLMs, by their very nature, require input data to generate relevant output. If this input data contains sensitive information, there may be potential risks associated with data privacy and security. In situations where confidentiality is paramount, it’s best to rely on trusted human experts to handle the information with the necessary care and discretion.
Enhancing, not replacing
In each of these scenarios, the AI co-pilot serves as a supportive partner, enhancing our capabilities rather than replacing them. By understanding the strengths and limitations of AI technology, we can harness its power while maintaining our uniquely human attributes.
So, let us embrace the concept of the AI co-pilot and celebrate its potential to elevate our work, our creativity, and our problem-solving abilities. In doing so, we acknowledge that AI technology is not meant to replace us but to empower us, reminding us that the future of AI is not about automation, but collaboration.