Welcome to the second article in my "Seeing the Elephant" series that explores at the implications and opportunities of ChatGPT and Generative AI.

In early January of 2023, I had a chat with ChatGPT that shook me. During that chat, I realized that the latest generation of AI tools has taken us to a critical inflection point.

A common mischaracterization of ChatGPT is that "under the covers," ChatGPT is simply using math and algorithms to predict the most suitable next word in a sentence. Basically, it's "auto-complete on steroids."

First, when it comes to ChatGPT, that's not technically correct, and more importantly, it misses the point. That's like saying that playing the piano is just pressing specific keys in a sequence with varying degrees of force. Technically correct, but vastly incomplete. 

Because the ChatGPT agent is explicitly designed for conversation rather than just question-answering, I wanted to test that.  To evaluate ChatGPT on a more complex and extended information task, I asked it to help me plan an long-distance sailing trip.

I picked that task because a few years ago, I sailed a 42-foot boat with some friends from San Francisco down the coast of California and then to Panama and out into the Pacific to French Polynesia. So that would make it easy for me to evaluate the accuracy and usefulness of the chat.

One of the first things I discovered during the chat was that ChatGPT shines when you start:

• Drilling down: What does "x" mean?

• Following up: Why do you think "y"?

• Being curious - Tell me more about "z."

Try it! You will find much more rewarding and informative conversations. 

The overall discussion was around 9,300 words, and the body of the transcript is about 35 pages. And importantly, the flow of the conversation felt very natural.

I started the conversation by setting the context: "Help me plan a sailing trip from San Francisco to Panama and then to French Polynesia. I'll be in a 45-foot sloop. "

It started out good but relatively superficial. However, the value of the discussion increased as I asked follow-up questions and requested more detail.   You could not do that with either Google or Bing in 2022.

While the full transcript of the chat might be an interesting read, today, I will focus on one seemingly simple question that exposed ChatGPT's superpower and made that whole conversation extraordinary to me.

A seemingly simple question

I asked,  "How many miles per day should I estimate for planning purposes?" 

Consider the parameters needed to understand the question and provide a useful response.

• We're talking about a boat, specifically a sailboat

• The length of the boat is 45 feet

• It's a monohull (not a catamaran, for example.)

• The route will dictate prevailing wind speed and direction, etc.

The answer started out strong, with important caveats about estimating the speed for sailing:

"It can be difficult to estimate how many miles you will be able to sail per day, as it will depend on a variety of factors such as the wind, currents, and the condition of your boat. "

But then it jumped straight to the *RIGHT* answer, having incorporated the conversational context:

"However, as a rough estimate, you could aim to sail around 200 nautical miles per day."

Wow. That answer was correct and included all the relevant parameters from an earlier conversation. ( 200 knots seems optimistic but in the right range. ) As I read that response, my jaw dropped in amazement, knowing how hard that would be to have programmed.

Then it went further...

ChatGPT's response did not stop at that point. It added further helpuful information, using the full context of the conversation, by interpreting that speed information specifically into what it meant for my trip:

"This would allow you to complete the trip from San Francisco to Panama in approximately 17 days, and the trip from Panama to French Polynesia in approximately 27 days. "

My experience in having this long chat convinced me that we are at the edge of a new era with AI -- the future has arrived. And we are in for a lot of change and disruption- good and not-so-good. 

Why it Changes The Web:

When I had my long detailed conversation with ChatGPT about planning my sailing trip, I got 100% customized, relevant, and focused answers to my questions without visiting any websites. 

And it was much easier and more natural than using Google orBing search to get the same information.

In my planning session with ChatGPT there were no opportunities to sell me sailing gear, training classes, gadgets, or books. Of course, that's likely to change as the chatbot/assistant starts to make suggestions for products and services.

As of March 2023, both Bing and Google are figuring out how to adapt to this change - and all the content developer and websites will also have to get ready for the disruption in the attention economy.

If you want to see the full transcript of my chat you can find it here. If not, drop me a note, and I'll send it to you.

FYI: I also shot a video about ChatGPT’s Superpower- you can find it here.

(click to open)

Next Time:  Generative AI and the Content Tsunami Subscribe below so you don’t miss it!

This is the first in a series,"Seeing the Elephant: Understanding AI technology", exploring strategies for thriving in a world disrupted by AI technology in 2023.

There is an ancient parable which goes like this:

Several blind men who have never encountered an elephant meet one for the first time. Each is asked to describe the elephant – but they are only allowed to learn about it by touching the part of the animal directly in front of them. 

The first touches the trunk and describes a powerful serpent. The second feels the tusk and describes something like a giant spear. The third holds one of the huge floppy ears and describes a bat-like creature.

Just like these blind men, many of those writing about ChatGPT are missing the big picture.

Before we can think clearly about the strategic implications of AI technology, we need to make sure we have an accurate understanding of what it is, and what it can do. Right now, many of us are getting that wrong.

A common criticism is that "ChatGPT Can't Do X" , where X is something humans do well. Often, like the blind men in the parable, this is the result of not understanding the bigger picture.

Some common themes:

1. It writes robotically, without any style (wrong)

2. It has no sense of humor (wrong)

3. It's not human and doesn't understand us (true, but…)

Let's take a look at each of these:

1. ChatGPT is not human and doesn't understand us

Ironically, Ian Bogost from the Atlantic got ChatGPT itself to write this:

"It is simply trained to generate words based on a given input, but it does not have the ability to truly comprehend the meaning behind those words. This means that any responses it generates are likely to be shallow and lacking in depth and insight." (written by ChatGPT) - The Atlantic, December 7, 2022 

Or as singer-songwriter, Nick Cave wrote:

 [ChatGPT's attempt to imitate my work] was "a grotesque mockery of what it is to be human,"

* * *

 ChatGPT has no inner being, it has been nowhere, it has endured nothing, it has not had the audacity to reach beyond its limitations, and hence it doesn't have the capacity for a shared transcendent experience, as it has no limitations from which to transcend." – Nick Cave, The Red Hand Files, January 23, 2023

Of course ChatGPT isn’t human.  Why do we need it to be?  Because it's not human, it has infinite patience and capabilities beyond what humans can do. My desk calculator isn’t human, but it can do amazing math well beyond my own abilities.

The mistake is anthropomorphizing ChatGPT -- judging the technology as if it were a co-worker down the hall. This is something I’ll explore more in an upcoming post. For now, let's just embrace that ChatGPT is not human and should not be judged as if it were.

2. ChatGPT can't write with style (Wrong!)

Tony Ho Tran of The Daily Beast recently wrote:

"...ask ChatGPT to write you an essay or a story. Do it a few times. You’ll quickly find that the writing is sloppy… There’s no style or flair." -- Daily Beast, January 17th, 2023

It's true that ChatGPT's default style is bland – but only if you let it.

As Tony Robbins is fond of pointing out,

"The quality of your questions determines the quality of your life."

With ChatGPT the quality of the question determines the quality of your answer.

When I asked ChatGPT to write a description of an encounter with an elephant on a trip to the zoo, in seconds, it produced this:

One of the elephants approached me, its trunk reaching out to sniff at my hands. I felt a thrill of excitement, for this was a once-in-a-life time encounter with an elephant's eyes. I was filled with a sense of wonder and respect for all that is wild and free.

I left the zoo that day with a new appreciation for the beauty and majesty of the elephants, and with a greater understanding of the importance of preserving their habitats and ensuring their survival for future generations.

The secret? I prompted it by asking for a description of the encounter "in the style of E.B.White."

Digging Deeper: I asked ChatGPT to create the same story in a variety of different writing styles including Jerry Seinfeld, William Shakespeare, and Mickey Spillane. It's an eye opening read, that you can find here, "ChatGPT Meets an Elephant" For fun, I also added AI generated images to spice it up.

3. ChatGPT has no sense of humor (Wrong!)

Another common misconception. Properly used ChatGPT can write with style and with humor – and do a lot of other things many people think it can’t do.

For example, Ben Eisen of the Wall Street Journal recently had the courage to let ChatGPT write him jokes for a comedy competition (well … a pun competition anyway…) The result?

I decided to use the robot’s best pun: “What’s the state where common sense is in short supply? Flori-duh.” The crowd loved it. --Wall Street Journal,December 12, 2022

To be fair, the Ben also delivered quite a few total bombs thanks to ChatGPT, but the “Flori-duh” pun got a real laugh from real humans.

So why do people think ChatGPT can’t write with humor? Simple: they don’t understand how to use it. It’s like compaining that your sports car won’t go fast enough because you don’t know how to shift out of first gear. To use ChatGPT properly you need to learn to drive it like a sports car -- shifting, turning, accelerating and braking constantly to get the best performance.

Prompt engineering 101

Prompt engineering is the skill of creating richer questions to get better results from generative AI like ChatGPT. 

Prompt engineering is a big topic I explore in an upcoming post but here are a couple quick prompt engineering tips to improve your Chat GPT results almost immediately. First, when you ask ChatGPT to explain a complex problem try adding the phrase, "explain step by step…"

For example, instead of asking "Explain how to make an omelet."

ask: "Explain step by step how to make an omelet" 

Second, if your responses are bland, stiff, or humorless, tweak your question like this:

• "Describe X in an informal and humorous style, and include a few jokes."

• "Explain X as a Norm MacDonald joke."

• "Rewrite that in an informal and conversational tone, and use humor."

So when you hear that ChatGPT can't do something, or does it badly, ask how could a better question to fix that?

The Bigger Picture: Seeing the Elephant

The technology behind ChatGPT is a big deal, and no one really understands the bigger picture yet. We’re all still blind ment, touching different parts of the elephant, and trying to fully understand what’s in front of us. Failing to acknowledge that is getting it wrong.

In this series, "Seeing the Elephant: Understanding AI technology" I'll continue to unpack the implications and the misunderstandings of this exciting technology.

JOIN THE CONVERSATION 💬

I'd love to hear your feedback, thoughts and insight about Generative AI and ChatGPT – please comment below. Also, please do me a huge favor and share it with your friends and colleagues, give it a thumbs up.👍

Next time: ChatGPT's Superpower

ChatGPT can provide competent, useful and detailed advice on nearly any topic, but you have to know how to ask it. In my next post, I walk through an extraordinary experience as ChatGPT gives me surprisingly accurate (well mostly accurate) and detailed advice for undertaking a three month voyage in a small sailboat from San Francisco to the South Pacific ⛵️

The wide availability of LLMs and related AI is a major inflection point in technology that will impact all of our lives in ways we are not yet able to anticipate.

“Useful Thoughts from UsefulBots” is where I will be sharing my thoughts, concerns, ideas and experiments as I try to understand this incredible technology, and it’s impacts.