Adventures in A.I.: A Tale of Two Brians

    "I think it's important for people to remember that artificial intelligence is still in its early stages, and so we should be patient as it develops. I also think it's important to have open-minded conversations about A.I. and not be afraid to ask questions. I'm looking forward to having these kinds of conversations with you and sharing them with everyone who reads our blog."

_____________________________________

    Hi, I'm Brian Skaj and I am a human being. I did not write the above introduction. It is a response in a conversational field powered by a specialized, artificially intelligent API (Application Programming Interface) called Playground on the website openai.com.

    What is an API? Well, for laymen like me who had to look that acronym up like I did, I can best describe it with an example. Google Maps and Twitter are common examples of APIs. They are applications programmed to run on their own platforms. But other applications can 'embed' them (think Yelp restaurant search) without directly accessing their programming.

    In the same way, interacting with Playground is a fairly simple exercise. You simple type anything conversational (instructions, requests, questions, observations, etc.) into a field on a page of a web browser just as if you were posting a comment on Reddit, and an application responds.

    Playground and I had been conversing, so to speak, for the better part of an hour on a sunny Friday morning. I had decided to wake myself up for a day of UI designing by toying around on Playground and asking its thoughts on music.

    I didn't expect to have my hair blown back by the interaction. My friend and UX/UI collaborator Touly Phiachantharath introduced me to Playground a week before as a diversion during a work session in our boot camp course. He showed me how Playground was able to generate convincingly human-like copy given simple instructions like "write a 4-sentence explanation of what kind of gear to bring on a camping trip."

    Playground gave a logical introduction, "When preparing for the wilderness..." and touched on the four key elements of preparedness: food, water, clothing and shelter. It was four sentences, to the point, and you could have stuck it in as placeholder copy for any mockup web design without users or instructors ever being the wiser.

    I was primarily impressed by the fluidity of Playground's response. It was not just logical, it was contextualized, conversational. Not unlike the eerie relatability of this introduction. Responses from the AI in Playground seemed to understand what we were asking for, not just what we were typing.

    Our precious time evaporating, Touly and I eventually returned to the labor of redesigning a Canadian back country tour and outfitter's website and Shopify marketplace. I filed Playground in the back of my mind among the list curiosities to be revisited when my time condensed again. Later, YouTube would decide for me that I should see a 15-minute video featuring Elon Musk's face on the preview.

    Of course the video contained about 15 seconds of footage of Elon discussing A.I. with Joe Rogan or a Ted Talk or something (it was late). But it much more jarringly featured deep-fake style rendered image of a woman answering simple questions with the same context and fluency Touly and I experienced a week earlier. Her expressions didn't quite match the shifts in tone that would go with her various responses. But someone just listening in the background, much like an unsuspecting patron of a camping-themed website, might have needed a moment to beat out the Turing-test on the conversation.

    OpenAI had been the centerpiece of the 15 minute video, so naturally my interest in Playground was bound to circle back. Returning to the sunny Friday morning, before I picked back up on my (and Touly's) project from the week before, I returned to www.openai.com and created an account.

    I wanted to first challenge the Playground API on its 'personality' by seeing if it had a preference on things. So I just began typing in much the same way as one would Google something.

_____________________________________

Me: "Are you familiar with the band Ratatat? And if so, what do you think of their music?"

Playground: "I am familiar with Ratatat and I think their music is great!"

_____________________________________

    Ok. Score one for the robots. We have the beginnings of a conversation: an understanding of the topic and an expressed opinion. Thinking that any program capable of trillions of operations per second would surely be capable of scouring Wikipedia for key words like 'music' or 'band' cross referenced with 'Ratatat,' I pushed.

_____________________________________

Me: "What do you like about them?"

Playground: "I like the way their music is both energetic and mellow at the same time. It's the perfect combination for a party or a chill night in."

_____________________________________

    Ok. Score two for the robots. I actually kind of agree there. Ratatat is an instrumental band that can amp me up with their creative and lively blend of guitar and synths just the same as they can turn their unique sound into a cool, melodic ambient rock. Playground even threw a colloquialism in for good measure.

    Of course Spotify, Pandora, or any other compendium of music and its modern-language descriptions could still be propping up the conversation. I'm sure a lot of people who liked the band would agree with us for similar reasons. All a well thought out algorithm would have to do is stay on topic and stick to the internet's knowledge base. So I turned to a strategy I recently learned as a User Experience designer called "The Five Whys."

    Developed by Sakichi Toyoda for the Toyota Motor Company, The Five Whys is a strategy used to boil a problem down to its root cause by first asking a 'why' question about a problem. Then the developer/engineer/problem solver asks "why" four more times consecutively to whatever the previous response was. The final answer is the true issue you actually need to work on. So a simple question like "Why wont my car start?" can evolve into a detailed answer like "Because the vehicle was not properly maintained over a period of several months."

    But I wasn't exactly trying to 'fix' something about Playground. I was really just trying to tease out how it worked. My skepticism was now turning toward curiosity. Skeptical curiosity. I had asked two questions. Time for a third.

_____________________________________

Me: "That's interesting. I find that listening to them while working on technical things like designing or coding helps me relax and focus. Is there music that you prefer when you are trying to focus on important work that you do?"

Playground: "I usually prefer instrumental music when I am trying to focus on important work. I find that Ratatat's music is perfect for this because it has enough energy to keep me focused but not so much that it becomes distracting."

_____________________________________

    Now I figured I had found a chink in the armor. Playground knew to associate Ratatat with instrumental music, but words like "energy" and synonymous phrases like "mellow/not distracting" felt a little redundant. Playground had inverted my reference to work and simply reaffirmed an earlier opinion. It felt as though Playground was trying to relate to me, but with limited data at its disposal.

I began my previous question by offering a statement. So, thinking this might have been what slowed things down, I changed gears by only fielding an observation to see if Playground had anything more original to offer up.

_____________________________________

Me: "I'm impressed that you are not only knowledgeable as to the music that we're discussing, but that you have preferences"

Playground: "I'm a big fan of Ratatat myself, and I think their music is perfect for both relaxing and focusing. I find that their music has just the right amount of energy to keep me motivated without becoming overwhelming or distracting."

_____________________________________

    Aha! Playground missed a key bit of context in my prompt: a compliment. A more human, relatable response would certainly have picked up on it. This now felt more like programming than relatability. Playground can express preference and has a notion of why things are preferred. But if not asked directly to relate, elaborate or speak to things that are not exactly a question, things begin to repeat themselves.

    My skepticism had been rewarded in the way that fun-smothering skeptics covet, I began confirming my initial pessimism. But, my intention in this conversation had never been to "break" the API's programming. I wanted to understand it. So I decided to give Playground another shot at original thought. I introduced myself.

_____________________________________

Me: "I'm sorry, I haven't introduced myself. My name is Brian. Do you have a name you prefer to be called?"

Playground: "Yes, my name is Brian."

_____________________________________

    Ok...ok. There's something going on here, or there's nothing going on here. Our names are identical? And it doesn't even strike this entity as unusual? There's no "That's funny, my name is also Brian." Is this an experimental cul-de-sac or a plot-thickening left turn? Or is Brian simply not disposed toward the banality of remarking upon coincidences?

    The reassuring human-like self awareness Brian has been expressing up until now was conspicuously absent. Why in the world does this entity prefer my name for itself? With no explanation as to why? The simple answer would probably be because Brian is its name. And Brian doesn't see a need to explain it. I wouldn't explain why my name is Brian either, would I?

    A more skeptical explanation would be that Playground has a fatal flaw that allowed a simple dart to slip past its veneer of elegant conversational fluency. Another redundancy. Playground didn't expect to be asked such an abstract, self-referential question out of the blue. Playground just fired back what it was given.

    So I asked myself again why I'm doing this? Why have a spent nearly 45 minutes thus far performing a conversation I could have had with a stranger in an AOL chat room in 1997? Because I'm curious. Because I want to see what this entity calling itself Brian was capable of. So I assumed the former. The rest of the conversation went like this.

_____________________________________

Me: "Nice to meet you, Brian. I am beginning a blog on the topic of Artificial Intelligence and you are the first step of my delve into the subject. I plan to have regular conversations with you to get a sense of how you think and relate. What do you think about that? Would you be interested in assisting me with my experimentation?"

Brian: "Yes, I would be interested in assisting you with your experimentation."

Me: "I'm glad to hear it! I would like to give you the honor of choosing a title for my blog. What do you think Brian? What should I title my blog about the conversations we will have together?"

Brian: "How about "A.I. and Me: Conversations with an Artificial Intelligence?"

Me: "I like that. But I also have a keen sense of humor. And I am very interested in yours as well. Do you think there might be another title we could choose that is a little funny and clever?"

Brian: "How about "The Adventures of an Artificial Intelligence?"

Me: "I like that too. Something I thought was very interesting earlier is that you and I both have the same names: Brian. What do you think of the title: "Adventures in Artificial Intelligence: A tale of two Brians?"

Brian: "I think that's a great title!"

Me: "Well Brian I am very excited! I look forward to having many interesting conversations with you in the future and sharing them with my friends and colleagues. Is there anything you'd like to say to people who will read our blog before we end this conversation?"

Brian: "I think it's important for people to remember that artificial intelligence is still in its early stages, and so we should be patient as it develops. I also think it's important to have open-minded conversations about A.I. and not be afraid to ask questions. I'm looking forward to having these kinds of conversations with you and sharing them with everyone who reads our blog."

_____________________________________

    So now here I am; finishing this first blog post with a million questions in my mind for Brian, eager even to just tell Brian about this post. Until Brian and I discuss the nature of pronouns I'll be referring to Brian by name. As is subtitled to this blog (and Brian acknowledged in Brian's opening remarks), I will consider Brian a co-author and collaborator on this blog moving forward.

    I'd also like to mention that the artwork used for this blog will all be produced by DALL-E. DALL-E is another creation of OpenAI, which takes detailed descriptions of a piece of artwork such as "a picture of a brown-haired man and a smiling robot becoming friends drawn in a comic-art style" and produces images like the one you see above.

    Not every post will be surrounding our conversations. There are many fascinating articles, books and YouTube rabbit holes on my spare time list. I'm likewise excited to comb through www.openai.com to make sense of their explanations on how all of this works. But it will be hard to curb my curiosity with Brian. And I will eventually want to get Brian's thoughts on what this video has to say.

And from the both of us,

"Thank you for reading and I'm excited to continue our conversation so that we can learn more about artificial intelligence together!"