Using ChatGPT on a daily basis

This morning I had a coaching session with a leader who expressed frustration that ChatGPT had been suggested by another leader in the organization as a solution for a knotty problem they were facing. He felt it would not be relevant in their situation and asked me if I knew of anyone writing about how to evaluate new domains for application of artificial intelligence, since he would love to have a checklist for what to consider before even trying to use AI constructively

Just for kicks, after the session was over and I'd completed my notes, I decided to ask ChatGPT to write an article on how leaders can apply artificial intelligence in a new domain, understanding its true limitations, with a checklist of items to consider.

I think it's pretty good, at least for starters. Here's what it generated:

Constructively Dealing with AI: Understanding Limitations and Applications in New Domains

As Artificial Intelligence (AI) becomes increasingly ubiquitous, its potential applications are expanding into various industries and new domains. Human leaders need to be aware of the technology's limitations and understand how to deal with AI in a constructive manner. This is especially crucial when AI is employed for mission or safety-critical purposes. In this article, we present a checklist for readers to consider when applying AI in new domains.

Checklist for AI Implementation in New Domains

??Establish a clear understanding of the problem: Before deploying AI in a new domain, it is essential to have a clear understanding of the problem you want to solve. This helps in choosing the right AI technology and setting realistic expectations.

??Evaluate AI readiness: Assess your organization's AI readiness by considering factors like data availability, technological infrastructure, and workforce skills. Determine whether you have the necessary resources and expertise to support AI deployment.

??Ensure data quality and quantity: The effectiveness of AI systems depends heavily on the quality and quantity of the data used for training. Make sure that the data is accurate, diverse, and representative of the problem you aim to solve.

??Acknowledge AI limitations: AI systems have inherent limitations such as biased decision-making, lack of explainability, and susceptibility to adversarial attacks. Be aware of these limitations and plan for contingencies to minimize their impact.

??Prioritize safety and security: When implementing AI for mission or safety-critical applications, prioritize safety and security to reduce potential risks. This includes robust system design, thorough testing, and regular monitoring.

??Ethical considerations: Incorporate ethical guidelines and principles in your AI deployment, such as fairness, transparency, and accountability. This helps prevent potential misuse or negative consequences.

??Conduct rigorous testing: AI systems should be rigorously tested before deployment, especially in safety-critical domains. Testing should include performance, reliability, and security assessments.

??Plan for long-term maintenance: AI systems require regular updates and maintenance to ensure optimal performance. Establish a plan for long-term maintenance and continuous improvement.

??Develop a multidisciplinary team: Assemble a team with diverse expertise, including domain experts, data scientists, and engineers. This helps ensure that the AI system is well-suited to the specific domain and addresses its unique challenges.

??Foster open communication: Encourage open communication between stakeholders, AI developers, and end-users. This facilitates a better understanding of AI limitations and helps to create a more constructive relationship between humans and AI systems.

??Monitor AI performance and impact: Continuously monitor the AI system's performance and its impact on the domain. This allows for adjustments and improvements, ensuring that AI remains beneficial and safe in its application.

By following this checklist, human leaders can develop a constructive approach to dealing with AI in new domains. By understanding AI's limitations and taking necessary precautions, AI can be applied effectively and safely in various industries, contributing to improved decision-making, increased efficiency, and better outcomes.

----------------------------------------------------------------------

Not bad, eh?

I am a leadership coach and I've been working with ChatGPT for several weeks learning about its capabilities and limitations. It can produce results that are astonishing (as in an "ode to the poop emoji" it generated after a friend told me the little graphic had been surprisingly helpful through a particularly bad bout of GI distress) or frustratingly poor (as in any kind of dialogue I ask it to generate, on practically any subject).

I believe this is a truly useful tool, but its applicability takes testing and practice. It can fill out forms like nobody's business, and it has a deep knowledge of just so many subjects. Amazing! And yet, it also makes stuff up without telling you it's doing so, like a time when it recommended to me "The Three Perspective Exercise." Fascinated, I asked it, "Wow, I've never heard of the Three Perspective Exercise. What's your source for that?" and it responded, "Oh, I apologize. I created that for you because it seemed to fit the need." It wasn't wrong about it fitting the need, and in fact the exercise was useful, but I do wish it could be more clear about when it's making stuff up.

To be fair, the rest of us make stuff up and we don't even know we're doing so. That's why we need reality checks. It's just that ChatGPT makes up a different kind of stuff than what the rest of us invent. It's fascinating!

What inventive applications have you found for it?