“Are you crazy, nobody uses ChatGPT like that, you know that right?” Mark said as we sat in the pub. I took a sip of cold crisp beer, slightly puzzled. I thought about how I’d been using generative AI to change my life recently. How I had used it to build a zen-like focus-point around a huge Japanese Acer in my garden. “Surely everyone uses it like that, isn’t it obvious?” I said, feeling slightly self-conscious. I hesitated. Should I tell him how I had also spent all evening talking to ChatGPT coming up with ideas for a cottage garden? It had factored in thousands of different plants that would make even Monty Don drop his trowel in surprise. That was when it struck me that most people use ChatGPT like Google. They punch in a prompt pieced together from some cheat sheet they found on social media and watch it spew out information like a hot bubbling volcano of knowledge, before copying and pasting the answer and never to give it a second thought. For me, this was a bit like using an expensive Macbook Pro laptop as a tea tray. Sure, it works and it’s useful to some degree, but there is so much untapped power in this method of computing, and by method, I mean “conversational intelligence” computing.
Steve Jobs famously once compared the computer to a bicycle for the mind — for its efficiency in propelling humans to where they wanted to go. I would be so bold as to say, extending this analogy… That AI is a motorcycle for the mind.
I want to share with you the unique strategies and practical tips I’ve discovered along the way in areas such as coding, personal learning, creative design, and conflict resolution to name just a few. With such a vast scope of possibilities, there is a lot of ground to cover. So, through a series of blog posts, I intend to share these along with my thought processes that you can use today to elevate your ChatGPT experience and help you unlock its full potential. Get ready to take your AI “wrangling” skills to the next level and learn to drive the motorcycle for your mind. The only limit is your imagination… Oh and how many messages you’re allowed to send in an hour!
Exploring Tomorrow: The Potential of Conversational AI
Now, before I reveal how I designed my garden, first let’s start off with a vision of the future since every journey needs a destination, every adventure into the unknown needs a heading. I know many people have been making AI comparisons to the widely recognised award-winning film Her but there is an even more ambitious utopian vision of AI – Star Trek. Wait don’t tap back yet! Now I realise talking about Star Trek may turn many of you off but please bear with me for a moment before we get to the detail and a real-world example of how AI can improve your life. Star Trek was a TV show I think many people would agree predicted or even inspired tablets such as the iPad, personal communication devices such as the iPhone and I believe it equally predicted conversational computing.
Like many technology enthusiasts, I grew up watching Star Trek: The Next Generation. The ship’s computer, which responded to simple queries and facilitated complex problem-solving, always fascinated me. It responded to simple queries such as “Where is Captain Picard?” or “What program is currently running in the holodeck?” which could be compared to Siri or Alexa in today’s terms on levels of complexity and usability. However, the show also presented incredibly intricate demonstrations of the computer’s power and capabilities of human-level understanding, reasoning and problem solving. Often with the crew engaging in elaborate queries, making the computer meticulously cross-reference multiple aspects of data, building upon them, and manipulating the results – ruling out items based on specific criteria told to it, before finally arriving at the destination. Watch the clip to get a better understanding, you only need to watch 30 seconds of it but you should watch it for seeing where we’re trying to go!
The above video clip is perhaps a little cheesy out of context but a great example of conversational intelligence computing. Arguably it’s largely used as a plot device to progress the story and accelerate any science. Imagine two scientists working this using traditional office apps and GUIs (think endless clicking, dragging and typing). They’d likely spend a significant amount of time building pivot tables in spreadsheets, organizing data, creating formulas, and sorting through information. Then, just when they believe they’re on the brink of a breakthrough, they find themselves embroiled in a new battle: wrestling with PowerPoint in Microsoft Office to make that perfect graph legible on a slide! And let’s not even mention the ‘fun’ of deciding whether the XY scatter chart should be 3D or not. Meanwhile, in the episode, Geordie and the ship’s computer work out the answer they need in a mere 2 minutes of casual conversing.
The Pitch – Unleashing the Power of Conversational AI
Now speeding up the plot in a TV show is one thing but imagine the impact of conversational intelligence computing on real-world data calculations and problem-solving. The efficiency with which Geordie and the ship’s computer arrive at the answer demonstrates the transformative potential of such technology. In a traditional setting, the process might be a laborious time-consuming process. A conversational AI streamlines this process, enabling users to explore data, analyse information, and uncover insights through natural language. This however is only a limited way of solving a problem. I recently had the realisation that most problems humans encounter don’t require advanced calculations and formulas. The reality is most of our problems can be expressed in terms of objects (such as people and things) and how they interrelate (relationships, values and what they mean such as cause and effect etc). The unrealised benefit of conversational AIs is that they understand these ideas and concepts since many of them are built into what we consider as “natural language”. How do they do this? Well, just as we build mental models to make sense of the world, AI models create their own kind of “mental model” in the form of a high-dimensional space. In this space, each word or phrase the model knows is represented as a point. The closer together two points are, the more similar the model thinks those words or phrases are. This high-dimensional space is how the model ‘understands’ the relationships between different words and phrases. By ‘thinking’ in this space, the model can generate responses that make sense given the input it’s received.
The Theory – How Generative AIs Learn
Take for instance the phrase “turning the handle of the door” – you need to understand doors have handles and that handles turn. The AI does this too, by mapping these concepts in its high-dimensional space. You might not know how to turn it but if you read it enough times, you’d realise the relationship that exist between them. You might not even know what turning was or even what a door was, but you’d know it was a verb and a noun respectively from the sentence’s structure so that given a door, you could expect a handle and that handles turn. Then you might read about doors being on a house and associate a house with doors, etc. That’s largely how generative AIs “learn”. Not only that but they can be used to build up incredibly advanced queries though a simple back and forth in normal conversation. Advanced calculations, relationships and processes can be expressed in a couple of phrases significantly reducing the time and effort required. Now, what if I told you that the Star Trek vision is achievable today. That ChatGPT can be used in this way right now! It might sound like I’m indulging in the AI hyperbole, but trust me, we’re about to shift gears from theory to practicality.
A Complete Walkthrough: Designing a Garden with an AI Horticultural Expert
Please be aware, this walkthrough is very long and comprehensive. I’ve spared no detail, at the request of friends and people I’ve talked about AI to. Feel free to scroll past the chats but the end has some closing thoughts that may be worth reading.
- Detail – Being specific and asking for advice
- Filtering and Finding – Identifying Mystery Plants with AI
- Visualising – Painting a Picture with Words
- Wrangling – Getting the best out of generative AI
- Sparking Creativity – Feeding it stuff to riff on
- More Iterating – Refining until the job is done
- Distilling it Down – Getting Targeted Summaries
- Leveraging Context – 11th Hour Additions
- The Grand Finale – Unveiling the Completed Garden
- Closing Thoughts
It started simply enough. I fired up a web-browser one evening, navigated to ChatGPT. Anyone can sign-up for free to get started and in minutes you could be asking advanced generative AI questions about anything you want. That night, I stared at the screen and asked it a simple question. Was it up to the challenge?
I was using GPT-4, the paid version of ChatGPT and as I watched it plod along writing, it reminded me of some sage wizard. Slower than the GPT-3.5 version everyone else was using but somehow more knowledgeable and wiser. Perhaps I was just personifying it too much. I wasn’t expecting much, but my jaw soon hit the floor when I saw its response.
I’d half-expected ChatGPT to give me some generic boiler-plate response asking me to consult a gardening expert. So, this felt like stumbling upon buried treasure! As someone who grew up around plants having a florist for a mother, I knew my Geraniums from my Chrysanthemums. However, anyone who’s ever casually watched a gardening programme like Gardners’ World will know there is a lot to consider – soil, sun, flowering times, colours, and that’s even before considering watering requirements and where to put the things! The idea that someone could give me personalised advice for colour schemes, textures and more as easy as having a conversation and for as long as I wanted was somewhat mind-blowing.
Detail – Being specific and asking for advice
Best of yet, there were no silly questions. I could ask it anything and not have to worry about asking something obvious. I thought about the flower bed I wanted to work on. It already had some striking plants in it. I didn’t know what they were, but I was pretty sure I could find out! My general thought process was to just give ChatGPT all the information and let it figure out the rest so my detail may seem obsessive but often the devil is in the detail. I wanted to paint it a picture and let it do its work.
I watched as it reeled though all my points and more, in a matter of seconds. I realised some of the plants seemed a bit different from local plants I knew about so I thought it was best to tell ChatGPT where I’m located.
Filtering and Finding – Identifying Mystery Plants with AI
While I was there, I really wanted to know what the existing plant was that I couldn’t identify. As GPT-4 (the paid-for version of ChatGPT) was limited by how many queries I could ask it, I often “doubled up” with the statements and queries in one “prompt”.
I must admit I was surprised. Had the AI just filtered out and found plants for the lovely Blity martime climate so typical of rainy old England in a single line? I quickly Googled Brachyglottis and I was pleasantly surprised to see its familiar silvery leaves with the happy little yellow daisy-like flowers that had revealed themselves last summer. I smiled, it reminded me of that famous Sherlock Holmes quote – Once you eliminate the impossible, whatever remains, no matter how improbable, must be the truth. I’d found the plant and only in a couple of sentences. Awesome work SherlockGPT!
Visualising – Painting a Picture with Words
Being a visually oriented person, I reached a point where I needed a more viseral perspective. I wanted to remove a mental layer – the Latin names of the plants, while useful, were somewhat getting in the way of my ability to imagine the rich colours, textures and smells of the garden. I know this might sound unusual, considering we’re dealing with text. However, I wanted to envision what my visitors would experience as they walked past the flower bed by my front door.
Wow! I was taken back. With a simple question phrased in the right way, ChatGPT was describing in detail my future garden using all the alluring imagery of a descriptive writing author in the throes of a dream-like sensory experience. I had to admit, this got me excited for the possibilities. My mind danced with so many ideas, eager to get ChatGPT to play with items, swapping out plants, trying different pots and colour schemes… Essentially painting a scene with nothing but sweet, beautiful words. It was a book geeks paradise. It was also amazing, as a programmer I thought about the complexity needed to code something similar just to produce that response – the huge variation of words needed, all the different ways of making it flow naturally without making the generated text sound like it was produced from a template. Most programmers think writing relative date libraries (that annoying time thing on Instagram where it says 4 hours ago instead of the date) is a good challenge as it involves lots of time calculations and working out if the hour should be plurals or not, etc. This was a completely different level, it was simply staggering.
Wrangling – Getting the best out of generative AI
Eager to expand upon my initial results, I wanted to explore different options, I wanted to fine-tune the output and really unleash the creativity this AI was opening up for me. Best of yet I had a gardening expert I could bounce ideas off and query their thought process.
I love the smell of lemon balm, the citrusy green leaves that look like mint but I wondered if it was just sticking to that option as I had mentioned it earlier. I was also curious what style it was intending to use for my garden and the messages below are a great example of wrangling – finding out the details and iterating on them.
Perfect, this was when we’d locked down the initial style – a cottage garden. It’s also funny how it read my mind about the visual interest of the lemon balm. It made me consider what we really think of as “intelligence” is really perhaps just reactions to different input. Every time someone says “Good morning, how are you doing?”, how likely will the response be “I’m great thanks. How are you?”. The context of the conversation was about creating a colourful garden and here I am questioning the lemon balm. Did I really give it any other option that to work out what was different between to the two and suggest alternatives more in-line with my garden?
Always striving to bring my perfect garden vision to life, I was curious about potential color schemes. At the same time, I wanted to incorporate terracotta pots to evoke the feel of Ancient Rome or Italy – ever the romantic! I was eager to see how these elements could blend together and I would say is a key part of wrangling the AI.
So let’s just recap. I started with a very vague initial prompt about it’s capabilities and though building up the context and bouncing ideas around before wrangling the AI in the direction I wanted, I now have an over-arching theme for the garden, a calm cottage garden style with a slight Mediterranean feel. I also have a whole range of plants suited to the conditions factoring in colours, scents and more, which also incorporate a lot of the plants I already have.
Suddenly I snapped out of my lucid daydream. Mark was there staring at me, offering another pint of lager. The day before, he’d seen another engineer friend of his ask ChatGPT to come up with ideas for an iPhone game. The engineer was sorely disappointed with the bland list of suggestions, all drawn from the top games of 2021 – the cut-off date when ChatGPT stopped being fed new information. The AI churned out predictable suggestions: an endless runner game, a puzzle game, a tower defence game, etc. The list droned on, generic and stale. A far cry from the colourful, surprising and novel ideas the engineer had in mind. Yet, he laughed at the uninspiring results with a slightly satisfied smirk, seemingly blissfully ignorant. He declared, perhaps a touch complacently, that AI would never take his job.
Reflecting on this, I considered the stark contrast between the results I had obtained versus those the engineer had. A lot of what you read about generative AI is about crafting the perfect “prompt” – a silver bullet that will reach deep inside the AI’s model (essentially, the AI’s brain) and extracting the correct answer as precisely as a surgeon performing key-hole surgery. However, that approach often falls short of yielding impressive results. The key to leveraging conversational AI, surprisingly, lies in conversing with it! It shouldn’t be about one prompt; rather, it should be about an evolving idea that you iterate on through conversation with the AI, like I did with my garden. It’s similar to chatting with a friend in the pub. You build upon your initial query, and perhaps going off on a tangent when you aren’t getting the results you want, which in many ways is like wrangling – sometimes you need to jerk the AI in a particular direction to get around a particular dead-end or subtly nudge it to extract the results you want. Importantly, just as in human conversations, establishing a rich context is crucial. This approach is the secret to not just ‘wrangling’ the AI but it also adds a whole additional dimension to the prompts, providing the AI with a deeper understanding of your goals and allowing it to decide what’s relevant or not. Generative AI is built on the concept of probabilities and each word, sentence and idea is used to determine what response you’ll get. Trust in this process.
Sparking Creativity – Feeding it stuff to riff on
Further exploring the concepts of wrangling and creativity, let’s return to the engineer’s unsuccessful prompt. Part of the issue lies in the nature of how generative algorithms work and are trained, but there’s more to it. Think of an improv comedian at a comedy club for a moment. Improv is all about taking a given idea and running with it, and for that, the comedian needs something to riff on for the best results.
Suppose you ask a comedian to tell you a joke on the spot, and they might conjure up with something mildly funny and canned… if you’re lucky. But share an anecdote about that time you fell out with your partner while attempting to assemble IKEA flat-pack furniture, and you’ve probably given them enough fodder for a Netflix comedy special. Generative AI functions in a similar way — it reacts based on the information you provide, giving it options. So, feed it as much detail as possible, or steer it by saying ‘from the perspective of…’ and you’ll get more creative results.
More Iterating – Refining until the job is done
Now I feel I’ve mentioned about wrangling, iterating and bouncing ideas around to death but I’ve included the rest of the chat for brevity. Feel free to jump to the next section on how I got ChatGPT to create a shopping list below if it’s too much detail for you.
I essentially consider and query some points. I think this is healthy in any creative brainstorming conversation – making sure the choices are made for the right reason and can yield some interesting insights.
Distilling it Down – Getting Targeted Summaries
Ok, finally I’m happy with what we’ve talked about but a powerful technique with the AI is to ask it to sum everything up or describe what you have so far. In the case of designing my garden I actually wanted a shopping list – a distillation of everything we talked about refined down to a key list of plants that I can pick up at my local garden centre.
Never feel that once you’ve covered something, you can’t go back and refine it. Here I decided I wanted to clarify if the plants are the strongest smelling ones. I build upon it and ask for a new shopping list. ChatGPT handles it swiftly and without batting a virtual eye-lid!
Leveraging Context – 11th Hour Additions
I believe there is a lot of power in building up the context of a conversation and with a few sentences taking everything you’ve said and applying it to something new. Here I remembered I had a hanging basket above the bed. ChatGPT factors in everything we discussed before and comes up with some great suggestions for plants for the hanging basket.
There is an old comic strip (2010 I think) by The Oatmeal, about a client asking to make a design “pop”. The designer can’t comprehend what the client means by “pop”. I couldn’t resist to see what ChatGPT would do. Don’t feel afraid of using everyday phrases. It runs counter to everything we’ve been told about using computers and how to “Google” but ChatGPT is intuitive like that and I think you’ll be surprised with the results.
The Grand Finale – Unveiling the Completed Garden
So that brings us to the end of our garden AI journey but what did I end up choosing? And no, that header image isn’t a stock image. It’s the real thing! I realise a garden is a personal choice and that it might not be to everyones tastes but considering where we’ve come from and my experience at designing gardens being zero, I think it’s a pretty good result.
Lying beneath my array of pots and planters are Cotswold stone chippings, their bright texture inviting a warmth that varies with the day’s light. With the dawn, they embrace a soft glow, growing to a cheerful radiance at noon, and mellowing to a gentle moon-light shimmer as dusk settles in. This delicate shift in light plays upon the variety of plants, creating a rich backdrop against the earthy terracotta pots scattered about.
In the hanging baskets, Trailing Fuchsias and Bullseye Cherry Geraniums are poised to cascade a riot of colours, their buds promising an imminent spectacle in the coming months. Their yet-to-bloom potential stirs anticipation, an unspoken promise of colour just waiting to unfurl.
The bed is a treasure trove of colours and fragrances. Pots of Lavender, with their intoxicating scent, sit alongside a fragrant Rosemary Standard which adds a sense of place and purpose. The New Guinea Impatiens, clothed in shades of Salmon Pink and Orchid Flame, add an exotic flair. A Peony Lactiflora stands proud, its fresh vibrant leaves stretching openly to the sky, its large pink blooms about to burst, adding a romantic touch to the bed. A pot of Pink Sorrel is a delightful surprise, its lovely pink flowers and clover-like leaves charmingly unconventional but adding to that rustic cottage garden feel as they open and close their petals to greet you daily with the sun.
There’s also a big pot of Lemon Balm, two Festuca glaucas that add a surprising spiky texture, and Brachyglottis, with it’s muted silver leaves both in the ground and in a pot. Watching over these marvels is a sweet smelling Tree Germander with delicate blue flowers that attract dozens of bees, its stature lending a stately presence.
At every turn, there are sights to behold and fragrances to savour, even tastes to behold, with each plant and pot playing its part. The Cotswold stone chippings beneath, meanwhile, subtly enhance each detail, casting the entire scene in a warm and inviting light. It’s a personal project, a testament to patience and attention, and it’s a joy to watch it grow and evolve. It’s far from finished but all good art is never truly finished!
Throughout this blog post, we’ve delved into the untapped power of conversational AI, with a particular focus on ChatGPT. We’ve demonstrated that it’s much more than a straightforward prompt-based system; rather, it can be an impactful tool for problem-solving and complex data analysis. We’ve seen how ChatGPT can understand and analyse natural language and concepts, showcasing its potential in a real-life scenario — designing a beautiful cottage garden. Something far removed from 24th century spaceships, science labs or end-of-the-world scenarios!
Indeed, it seems we’ve only just embarked on a fascinating journey into the realm of AI. It’s time to truly get acquainted with ChatGPT, and start our conversation with the AI, not just asking it to complete tasks or answer questions but leveraging it for creative design and new ways of thinking. The ability of ChatGPT to understand and respond to complex queries in natural language is a game-changer in so many ways.
I stand by my earlier statement that AI is the motorcycle for the mind. Just as a motorcycle accelerates your physical journey, AI can speed up your intellectual and creative journeys, helping you arrive at solutions quicker and with less effort. Of course, I haven’t done anything that couldn’t be accomplished through weeks of library research, scouring gardening magazines, or making a dozen phone calls to that sage gardening expert affectionately known as ‘Granny.’ Yet, consider how much time I’ve saved. Reflect on the simplicity of swapping out plants and considering different options. Think about how easy it was to incorporate hanging baskets at the eleventh hour. And finally, imagine the convenience of generating a neat, comprehensive shopping list. This doesn’t even take into account the vast information you can query about any particular plant – how to repot it, potential problems, and so forth. It genuinely feels like having a helpful sidekick at my disposal, ready to assist with any need. I believe the secret lies not in treating it as an inanimate tool, but in embracing it as a dynamic partner or co-pilot in your thought process. Picture it as your very own R2D2 from Star Wars except with less beeping!
So, get set to explore, experiment, and evolve with ChatGPT. Whether you’re a professional aiming to harness AI in your work, or a curious individual keen on exploring the possibilities, the future of conversational AI is yours to discover. After all, isn’t it time to unleash the ‘motorcycle of your mind’ and journey to the places you’ve only dreamt about if you had more mental energy?
Sharing the Knowledge
Imagine a friend or colleague reading this article, discovering the world of AI and ChatGPT. Perhaps they’re tech enthusiasts, lifelong learners or even keen gardeners. Could this insight into AI be a valuable resource for them? Consider sharing this article – whether by hitting the ‘Share’ button on LinkedIn or sending it directly. You’re not just sharing an article, but contributing to their knowledge about this topical intriguing subject. Also if you want to keep updated, follow me on LinkedIn or Twitter. I’ll be posting more thought provoking articles on practical AI soon.