ChatGPT Meets the Wizard of Oz

ChatGPT is no dummy. I think we'd all agree on that.

But it can be a frustrating partner. Especially when it doesn't understand what you want it to do.

And I get that. People have spent thousands of years learning how to communicate with each other and that's led to certain expectations. But talking to a machine is a different story.

One we're still figuring out.

Still, there are times when a conversation with ChatGPT feels like you're talking with a slick, media-trained CEO. Who keeps ducking the issue and won't deviate from their bland, corporate-approved messages.

And tries to butter you up with banal compliments.

So instead of a straight answer, you're treated to something like, 'As a large language model, I am not able to <insert situation>. But I've been enjoying our lively debate.'

Um ... sure you have.

Still, I get this too since ChatGPT—like said CEO—is loathe to say something that could spark a scene.

Is Anyone Home?

But is ChatGPT able to compute that minimizing harm is the reason for its non answers?

Do Chatbots have the capacity to learn from interactions, adjust and understand?

Most experts would say no.

But one pioneer, researcher and University of Toronto professor Geoffrey Hinton, believes chatbots have 'subjective experiences'.

Hinton, one of the godfather's of AI and the person whose work provides the foundation to today's deep learning models, thinks the current iteration of AI goes beyond mere statistical predictions and has a 'theory of mind' or grasp of what it's doing.

He doesn't say machines are sentient, just that they possess a world view and adapt their responses based on that.

He makes a compelling argument. And I'm inclined to believe him though I also don't know what the answer is.

So I decided to try a simple test and asked ChatGPT 4o: 'Can you have a subjective experience based on a question I pose?'

Of course, it said no and provided two reasons:

• No Consciousness: I don't have self-awareness or personal feelings, which are core to subjective experiences in humans.
• Algorithmic Function: My responses are generated based on algorithms and pre-programmed patterns, not personal experiences or emotions.

I kept pushing the matter only to get more denials.

Then I changed tack and suggested we take consciousness out of the equation and simply consider subjective experiences as the ability to make predictions based on the unstructured data our brains process. In that case ChatGPT grudgingly agreed it could have them.

And I wondered, will those experiences be the bridge to AI agency and autonomous, proactive chatbots making decisions and judgement calls with little or no oversight from humans?

This Doesn't Look Like Kansas

According to tech companies like OpenAI and Google, AI agents are coming soon to an app near you.

And it occurred to me that the current crop of chatbots are missing three key ingredients for that to happen:

1) a more proactive way to analyse data and make decisions;

2) EQ or emotional intelligence;

3) the courage of its convictions, that is, knowing when to stop demurring and stick to a thoughtful response.

In simpler terms, AI needs enhancements to its brain, some heart and the ability to overcome fear.

Sound familiar? It's like a page out of the Wizard of Oz.

The Wizard of Botz

Let's reimagine the Wizard of Oz story with an AI slant. The time is late 2025.

Say, Dorothy is you, the chatbot user, who has a travel problem to solve.

Your three companions are anxious to help you, but have challenges of their own:

• The Scarecrow has a few ideas (like using glue to keep cheese on a pizza), but needs a bigger brain and better judgement.
• The Tin Man needs a heart to empathize with with Dorothy's plight and offer her a sympathetic ear.
• The Cowardly Lion needs to find the courage to go out on a limb and take a chance on a viable route off the beaten track.

But who's the Wizard? OpenAI? Google? Anthropic?

Or is it, as Geoffrey Hinton might reason, already built into the machine, which just has to figure out how to unlock it.

And when Agent Chatbot figures that out, what can we expect? A beneficial helper? A wise and understanding friend? Or a puppet master pulling our strings?

If chatbot behavior combines the best qualities of the Scarecrow, Tin Man and Cowardly Lion, would that make us relate to them, listen to them and trust them even more?

Who will control these new agents? The companies that make it, the Dorothy's of the world (i.e. yours and mine) or the AI agents themselves?

How can we be sure they'll be honest with us? And won't simply do whatever they can to cajole us into doing what they want?

Will their goals get misaligned with ours? Where will that leave you and me?

I realize I'm veering off into dystopian territory here. Or am I?

In the Future Today Institute's (FTI) 2024 Tech Trends Report, CEO Amy Webb describes where we're heading:

We believe we have entered a technology supercycle. This wave of innovation is so potent and pervasive, we believe it will reshape the very fabric of our existence.

Gazing into the Marketing Crystal Ball

In this week's video, I dig into the 2024 Future Today Institute report and highlight several trends for digital marketers to watch.

This includes AI agents as the new sources for discovering products and info and what that might mean for your brand and how blockchain could be used to verify digital identities and rebuild trust.

And I discuss FTI's three-step approach to building future-facing scenarios simulate the road ahead:

1. Create a descriptive headline.
2. Add a timeframe and outcome (positive, negative, neutral).
3. Write the story and assess strategic implications.

Check it out and let me know what you think.

Follow Me on LinkedIn

And that's the end of the subjective experience known as issue 103.

Thank you to all of you who follow me and subscribe, read, comment and share this newsletter! It's great having all of you along for the ride.

This newsletter comes out every two weeks. But between issues, I share shorter daily posts with my take on digital marketing and the latest on generative AI. It's another way to stay on top of the trends.

And while you're at it, follow the Future of Marketing Institute, too. Every day we post content and perspectives on where we may be heading and what the shift might mean.

Let me know if you have questions about any of the videos in Digital Marketing Trends, or my other LinkedIn Learning courses. You can also visit my website and send a message or a question.

Do you believe AI agents will be able to perceive the world and have individual experiences that shape how they'll behave in the future? Can they do that now? What will that mean for you and me? Please share your thoughts in the comments below.

I'm taking a short summer break and won't be back till mid-July. Stay subjective and see you then!

Note: All the content in this post was written by a human—me and not Martin-bot. The visual was created by Dall-E 3.