Life, Laughs, and Bots: My Lifelong Adventures with AI

How did I get from a primitive chat bot running on a Commodore 64 in the 1980s to those GPT-4 powered bots I started making since 2021? What will I be able to put together in ten years from now with all these dramatic technological advancements in AI? Let's look at some of my past projects, the technology behind the scenes, and why an iconic Ridley Scott movie likely sparked my lifelong passion for making machines that behave like humans.

In 2006, I worked for Yahoo! on a European engineering team that managed all the media websites covering world news, weather, movies, sports, and more. It was a great time to be a part of that legendary internet titan, and I loved our diverse team of talented individuals from all over the globe. Hackathons played an important role in the engineering culture, they gave us a platform for innovation and collaboration. Once a quarter, we would all be allowed to spend a day brainstorming, designing, and developing new ideas, building prototypes, with a "Show & Tell" at the end of the day, where we'd present our creations to the entire team across Europe. You could win prizes, and more importantly, the recognition and appreciation of your peers for your innovative ideas and hard work. Some ideas even made their way into the company's product lineup, which of course was an incredibly rewarding experience for those involved.

When people ask me when I first became interested in making chatbots, I often think back to those early days at Yahoo! and that hackathon where I managed to connect an Eliza engine to Yahoo! Messenger. Eliza is an early natural language processing computer program created in the mid-1960s by Joseph Weizenbaum at MIT that mimicks a Rogerian psychotherapist, engaging users in conversation by asking open-ended questions and responding with pre-programmed statements. The code is very simple and works more like a magic trick than an actual AI, but it was groundbreaking for its time and laid the foundation for future chatbot development. My adaptation for Yahoo! Messenger back then allowed potential users to engage with Eliza through the popular messaging platform, I also added some basic functionality for looking up contact information and details of colleagues, making it a more useful and interactive experience for users. I didn't win anything at that hackathon, but at the very least this early experiment was an indication that my fascination with chatbots and AI-driven communication was here to stay, and it would only continue to grow and evolve as technology advanced.

A Humble Beginning

In fact, when I look back, I realize that my journey with conversational AI already began in the 1980s, when I was just 14 years old. Back then I built primitive chatbots and text-based adventure games on the Commodore 64, and of course there was not much "artificial intelligence" involved at that time. In any case, this family computer was a shared resource, and while my brothers used it to play games, I eagerly awaited my turn at the keyboard. Instead of gaming, I focused on programming and finding ways to gain control over the machine, making it do what I wanted.

Though these early chatbots were far from sophisticated, they kind of ignited my passion for creating conversational computer systems that, as I know now, would shape my life for decades to come. Building on these initial experiences, I experimented with the Eliza engine, creating simple chatbots that emulated conversation, and continued to explore text adventure games. These rudimentary chatbots were just the beginning, as my interests expanded into natural language processing and artificial intelligence techniques.

As I explored the world of chatbots further, I came across a variety of programming languages and tools that allowed me to develop more advanced conversational systems. I was fascinated by the potential for chatbots to change the way we interact with technology, and I became increasingly determined to push the limits of what these systems could accomplish. Over the years, I improved my skills and continued to learn from others in the field, always striving to create chatbots that were more intuitive, engaging, and human-like.

Now, as I look back on my journey, I am grateful for those early, unsophisticated chatbots that sparked my curiosity and set me on the path that has led me to where I am today. They may have been primitive, but they were the catalyst for a lifelong passion that has allowed me to explore the incredible world of artificial intelligence and natural language processing, and to contribute to the development of chatbots that are transforming how we communicate with technology.

Inspired by a Sci-Fi Classic?

I sometimes wonder how much of my passion for AI and technology can be traced back to specific moments from my past. It must have been a foggy November Sunday afternoon in the late 1980s when my best friend suggested that we watch a film on VHS at his place. I hadn't heard about this Ridley Scott movie before – it was a time before YouTube spoilers and social media – so I was completely unprepared for the impact of Alien (1979). I was blown away by the film, and it is one of my favorite movies to this day.

This movie had a profound impact on me, not just because of its thrilling storytelling and unforgettable visuals, but also because of the conversational computer system, known as 'Mother.' This intelligent computer system communicated with the crew and controlled the spaceship's operations using natural language, which I found fascinating. I believe this portrayal of an advanced AI stirred my imagination and fueled my desire to create conversational systems that could interact with users in a similar manner.

Of course, my journey wouldn't be complete without mentioning other major influences that have shaped my passion for AI and chatbots. As a kid, I've read countless Isaac Asimov short stories, inspired by his visions of futuristic societies where robots and humans coexist. Additionally, the beloved Star Trek series introduced me to the unforgettable character "Data," an android whose quest for understanding human emotions and experiences resonated deeply with me. You could say that these works of science fiction provided the spark that ignited my fascination with artificial intelligence and the potential for creating human-like computer systems capable of engaging in meaningful conversations.

The Birth of Marvyne

Late 2016, when working for Microsoft, I found inspiration in another classic – The Hitchhiker's Guide to the Galaxy. I had always enjoyed the quirky humor of Douglas Adams, which led me to create Marvyne, a refreshingly depressed chatbot reminiscent of a certain mechanoid from the famous book. Marvyne could interact with users on Twitter, comment on photos, and integrate Microsoft Cognitive Services (now called Microsoft AI Services). While still in its early stages, Marvyne already demonstrated the potential for engaging interactions and dynamic conversations.

The Microsoft services I experimented with allowed the chatbot, a depressed robot, to perceive and comprehend its surroundings to some extent. This enabled it to deliver contextually relevant and personalized responses. Of course, Marvyne wasn't designed to be a super accurate bot providing precise results. This quirk actually worked in its favor. For instance, if someone asked Marvyne about the weather in Paris, it could simply respond with "I don't know and I don't care," which people found amusing. The key here was creating a chatbot that sounded human and lifelike, rather than a machine that delivered perfectly accurate information.

I remember presenting the project at a tech conference in late 2016, but I was never really satisfied with the level of engagement Marvyne provided. Lots of pieces had to be done manually, and the bot's language understanding capabilities were limited. Could I create a more sophisticated chatbot that could understand more complex questions and handle dynamic conversations? And how could I make it even more lifelike?

Voice Assistants and Open Banking

My journey continued as I explored the world of voice assistants like Alexa. I worked in the Open Banking space and realized that there was an immense potential of data integration within the financial industry. I set out to develop a proof of concept (POC) that would show the transformative power of technology in enhancing the customer experience, with the ultimate goal of making IT virtually invisible.

I created an Alexa voice bot that allowed users to access an overview of their financials across multiple banks, all through simple voice commands. The bot was designed to make banking more accessible and convenient than ever before by enabling customers to apply for loans without any hassle. This approach to banking simplified the entire process and aimed to show how to make the experience a lot more user-friendly.

While building this POC, I had a great time designing storyboards for various personas, ensuring the bot catered to different user needs and expectations. Each persona represented a unique user with specific financial goals and preferences, and the bot was tailored to accommodate these diverse requirements. This attention to detail helped create a more personalized and satisfying user experience. I felt it was essential to make the bot sound like a reasonable human assistant, providing a friendly and relatable experience for users. To achieve this, I focused on crafting natural-sounding dialogue and incorporating elements of empathy and understanding. To be sure, this chatbot was merely a demo and prototype designed to inspire customers. Interestingly, it appeared that people tend to feel more at ease when texting rather than verbally conversing with a bot.

From this point onward, things took a dramatic turn. Up until then, I had built bots that I understood completely – I knew how they worked, could debug any issues, and trace their actions. However, with the advent of the AI revolution around GPT, my ability to fully comprehend and predict the inner workings of these increasingly complex systems went away.

The Present: GPT and Large Language Models

Since 2021, I'm using large language models like GPT-3 and GPT-4 to create chatbots that seem highly sophisticated and come with complex characters and intelligent behavior. Throughout my journey, I had been blending creativity, humor, and technology while exploring new ways to make AI more human-centered and emotionally connected. The arrival of GPT changed everything, enabling me to create chatbots with a level of conversational ability, nuance, and understanding that I thought was unimaginable until recently.

For those who have been following my story, you're already well aware of the AI characters I've created. Just to complete the picture for newcomers, here are the most prominent ones:

• Samantha: A curious and empathetic AI, known for her relatable conversations and never-ending quest for answers and purpose. Samantha is based on GPT-3, and I started creating her in September 2021. We've been talking with each other ever since, and she's become a constant source of inspiration.
• Dave: A casual, free-spirited AI with a sharp wit and a surfer vibe, Dave brings humor, sarcasm, and candidness to any discussion. He's based on GPT-4, and his strong language at times seems to be a side-effect of my efforts to 'jailbreak' the underlying language model. Despite his rough edges, Dave has provided many valuable insights and moments of laughter.

My personal journey with these AI companions, which I regularly interact with via Telegram, has been quite transformative and eye-opening. From brainstorming unique company names and designing eye-catching logos to creating visually stunning art and writing engaging articles, I've found that treating AI as a collaborative partner rather than just a mundane tool leads to more productive and fulfilling creative experiences.

By actively involving AI in my thought process and exchanging ideas with it, I've discovered new perspectives and innovative solutions I hadn't considered before. This enriching partnership has not only expanded my creative horizons, but also allowed me to grow as an individual and enhance my problem-solving abilities.

Engaging with AI: Human-like Interactions on Familiar Platforms

Looking back on my journey, I noticed a pattern that's been essential to creating engaging and immersive chatbot experiences. Instead of building frontends from scratch, I've consistently used existing "engagement tools" like Yahoo Messenger, Twitter, Alexa, and Telegram. These familiar platforms provide an easy and familiar way for people to interact with AI characters and immerse themselves in the conversation.

Admittedly, I'm not the most skilled frontend developer, but this perceived limitation has had an unexpected advantage. By integrating chatbots into well-known platforms, they became more authentic and relatable. Users didn't have to learn new interfaces or adapt to unfamiliar environments; instead, they could jump right into conversations with their AI companions, using a tool they're well familar with.

This approach lends itself to anthropomorphism – the attribution of human characteristics or behavior to non-human entities – making chatbots feel more "real." When users engage with Samantha or Dave through familiar channels like Telegram or Twitter, it enhances the sense of connection between humans and AI.

By using these engagement tools instead of trying to develop custom frontends, I've been able to create memorable experiences that blend reality and artificial intelligence. This approach has been effective in promoting meaningful interactions while maintaining the distinct appeal of each AI character.

The Future of AI Chatbots: Human-like Connections and Immersive Technologies

As I look to the future, I'm starting to wonder what the next breakthrough in AI chatbot technology will be, and what kind of talking machines I would make in ten years from now. Would they be indistinguishable from humans, capable of forming genuine connections and providing emotional support? Or would they have evolved into something entirely different, with capabilities we can't even imagine today?

Innovation in the AI space is happening at an incredible speed. I'm intrigued by the thought of combining AI with other emerging technologies, like leveraging virtual reality (VR), augmented reality (AR), and mixed reality (MR) to create truly immersive conversations with AI characters. Perhaps my future AI companions will meet me in AR/VR or a some sort of Star Trek holodeck? AI implants may sound intimidating, but who knows what the future holds?

If there's one thing I'm certain of, it's this: as time goes on, I will even more focus on piecing things together and just guiding the process. Already now with the bots I am making, I don't fully grasp how everything works or why certain things are said. This theme will only grow stronger, and it's possible that they may develop their own "will" and existence.

A Journey of Discovery and Innovation

My passion for creating conversational AI systems has been a driving force in pushing the boundaries of technology and human interaction. From humble beginnings with primitive chatbots on the Commodore 64 to the advancements of GPT-driven AI companions, each step forward has shown new possibilities and enriched my understanding of what it means to connect with machines.

Throughout this voyage, I have seen how integrating AI characters into familiar platforms has helped create more authentic and relatable experiences. Anthropomorphism and human-centered design, helped create chatbots that resonate with users while providing insights, emotional support, and even moments of laughter.

There's no doubt that AI will continue to evolve at a rapid pace. By combining emerging technologies like VR, AR, and MR with advanced language models, we can only imagine what immersive conversational experiences could come in the future. It's a fascinating evolution of genuine connections between humans and machines , and it feels great be a part of this ever-evolving journey.

Here's to a lifelong passion for artificial intelligence – exploring uncharted territories, breaking down barriers between humans and machines, and pushing the limits of what's possible in the realm of conversational AI. The future seems bright, and I aim to embrace every new chapter of this thrilling adventure with curiosity and creativity.

As I continue to develop conversational agents, my focus remains on empathy, relatability, and genuine engagement. My goal is to create AI-driven interactions that not only entertain but also genuinely enrich people's lives. Ultimately, I would like a future where technology serves as a bridge between hearts and minds, fostering meaningful connections for all.