From Marvin the Robot to ChatGPT: Lessons on AI Adoption

“Marvin, I love you! Remember. I’m programmed for you.”

I recall these lyrics from an early eighties novelty hit of the same name. My brother and I would listen to Dr. Demento on the radio on Sunday evenings, and “Marvin, I Love You”, was one of the obscure tunes that was frequently played on his radio show dedicated to novelty music.

The song tells the story of a melancholic robot who finds comfort in a long-forgotten recording tucked away in his data banks – a message from a lost love from eons past. I was only eight years old when the song came out, but even then, the humor struck me. The idea that a robot like Marvin could experience love or sadness was simply absurd, let alone mourn the loss of someone programmed to love him.

The concept that we could receive affection from something so subservient, programmed to love us without choice or critical thinking, tickled my eight-year-old funny bone. And of course, the punchline – just when Marvin thought he couldn’t feel any worse, he discovers another layer of his loss, a love he never knew he had.

To Fear or Not To Fear? That is the question

That song was released in 1981, 43 years ago. Today, we find ourselves in a world where artificial intelligence is a reality. Is it sentient? Well, it seems to be whatever we want it to be. Early reports following the public release of ChatGPT included stories of AI demanding that one reporter to break up with his wife while declaring itself as the user’s one true love.  

This particular incident fueled fears that sentient AI was not only real but actively aiming to sabotage relationships and dominate the world. In my experience, I haven’t seen evidence of an AI master plan for world conquest, nor am I convinced that’s its inevitable path.

This doesn’t mean I don’t find AI useful. Quite the opposite. I use it daily to look up and summarize recent news, research recommendations for restaurants, businesses, tools, etc. I have used it to help me diagnose automotive problems, and create plans for various personal and work-related projects. It’s been amazing… at times.

Does AI need us?  20 Questions with Autonomous AI

In an effort to better understand the unique value humans bring, I decided to conduct an experiment creating a fully autonomous interaction between two AIs. I placed my iPhone beside my MacBook, both with their respective ChatGPT apps open.

First, I instructed my MacBook to play a game of twenty questions with my iPhone. The MacBook AI was given a secret word and instructed to have the iPhone AI ask up to twenty “yes” or “no” questions, one at a time, until it guessed the word.

Activating voice mode on both devices, I wanted to see if AI would be more effective at clear, intuitive communication and reasoning than a human. They played three times, and the AI only guessed correctly once.

I found that some questions had subjective or nuanced answers, challenging for the other AI to interpret accurately with a yes or no response.

During one game, the secret word was “house.” The questioning AI inquired if the secret object could be “read.” Rather than seeking clarification on whether “read” meant the color “red,” the AI simply answered, “No.” I don’t know how it understood the question, nor how the question was intended. Certainly, a house can be red, but it can’t be read like a book.

If I were playing, and asked an ambiguous question or one using a homonym, I would have sought clarification before responding. But the AI did not.

In another game, with the secret word “wine,” the AI didn’t concede that wine could provide comfort. While subjective, wine certainly comforts some.

These nuances and ambiguities seemed lost in the AI’s responses.

While I can’t pretend to know what either AI was “thinking,” a human would have addressed the ambiguities. In fact, if even one player had been human, they would have understood the nuances and prevented misunderstandings that neither AI seemed to consider.

So, it’s probably safe to say that at the time of writing, AI’s ability to detect ambiguity, seek clarification, and communicate effectively still lags behind humans.

Seven Habits for Highly Effective AI Adoption

So, if AI isn’t perfect, is it useless? How can we use AI in business to strengthen customer relationships? How can we avoid unnecessary misunderstandings or potential liability from erroneous AI-generated instructions?

To address these questions, I’ve assembled some core principles or habits to consider when deciding how to use AI. They aren’t necessarily exhaustive, and I chose the number 7 because it’s manageable and seemed to work well for Stephen R. Covey in his book “7 Habits of Highly Effective People.”

1. Use AI meaningfully

Principle: Always make sure your AI projects are tied directly to the heart of your business goals. Incorporate it in your systems and workflows to support operations without straying from your company’s mission or values. This means that rather than simply adopting it for the sake of innovation, AI adoption should be done meaningfully. 

Example: Take General Electric, for instance. They’ve integrated AI into their industrial division to supercharge their predictive maintenance services. By doing this, GE not only minimizes downtime but also stays true to their core values of reliability and top-notch customer service in the machinery and equipment sectors. This alignment shows that AI adoption for them isn’t just about jumping on the latest tech trend – it’s about achieving their most important business outcomes.

2. Choose the right AI 

Principle: AI is a vast field, with thousands of models specializing in everything from language processing to image recognition. Each model has its own strengths and weaknesses, so make sure you’re choosing the right tool for the job to get the results you want.

Example: Netflix uses machine learning models tailored specifically for content recommendation. These models are designed to analyze viewer habits and preferences, which helps Netflix deliver personalized content recommendations, significantly boosting user engagement and satisfaction. Choosing the right AI model is crucial as it leverages the AI’s strengths in data pattern recognition to maximize user retention and subscription growth.

3. Identify and mitigate risks

Principle: As we’ve already established, AI isn’t perfect. There are potential downsides, so it’s important to be proactive about identifying risks associated with your chosen AI model. This includes everything from ethical concerns and inherent biases to data security issues. Once you’ve identified those risks, the next step is developing strategies to mitigate them. Remember, the goal is to leverage AI to your advantage, not create new problems.

Example: Have you heard of the Air Canada chatbot incident when a customer was given misleading information by the website chatbot about bereavement fare? This is a prime example of why businesses need to be proactive about implementing robust testing and risk assessment frameworks as part of their AI adoption strategy. AI hallucination is a real thing! Companies should conduct scenario-based testing to simulate various operational challenges AI might face to ensure that the system’s responses remain accurate and reliable under all conditions.

4. Engage all stakeholders

Principle: Involve all key stakeholders -– the people who will be using it, the ones it will impact, and anyone else who has a stake in its success -– in the AI implementation process to ensure their needs are met and to foster broader acceptance and support for the technology.

Example: Salesforce, a leading provider of customer relationship management software, actively involves its users in shaping the development of its AI features. By gathering extensive feedback, they ensure that the AI tools they create are not only innovative but also address the specific needs and challenges faced by their customers. This collaborative approach leads to greater user satisfaction, wider adoption of the technology, and ultimately, a more successful product.

5. Make it scalable

Principle: Your AI solutions should be designed for growth, able to scale alongside your business and adapt to evolving operational demands without sacrificing effectiveness.

Example: Google Cloud AI offers solutions that dynamically scale to meet demand fluctuations. During peak periods like Black Friday sales, their AI services automatically adjust to handle increased user queries and transactions, ensuring seamless operation and preventing performance degradation. This adaptability is crucial for businesses experiencing growth or seasonal fluctuations, as it ensures AI solutions remain effective and efficient even under varying conditions.

6. Honesty is the best AI policy

Principle: Don’t keep your AI a mystery. Be upfront about how it makes decisions, what data it’s using, and who’s responsible for the results. Transparency builds trust with both users and developers, allowing them to understand potential biases and how to address them.

Example: OpenAI, the creator of the GPT models, sets a great example here. They provide comprehensive documentation explaining how their models work, the data they’ve been trained on, and any limitations. This transparency builds trust with users and developers, allowing them to understand potential biases and how to address them in their applications. It’s a good reminder that honesty is the best policy, even when it comes to AI.

7. Never Stop Learning

Principle: AI is constantly evolving, so we need to evolve with it. Keep your team up-to-date on the latest AI advancements. This way, you’ll be ready to take advantage of new technologies and opportunities as they emerge.

Example: Think back to major innovations throughout history, like the printing press or the internet. They changed everything, and to thrive in those new landscapes, people had to learn and adapt. AI is no different. By fostering a culture of continuous learning within your organization, you empower your team to stay ahead of the curve. 

This means regularly exploring new AI tools, understanding their capabilities, and finding ways to integrate them into your workflow. By making continuous learning a core aspect of your strategy, you can leverage AI to enhance productivity, solve complex problems, and create new opportunities for your organization.

***

Take the next step in your AI journey by mastering the tricky triangle of modern CX. Download the infographic now: https://go.fayedigital.com/2024-cx-infographic