The world seems to be polarising on the AI debate. There are AI Evangelists, who see Gen AI as the dawn of a new era and the start of a phase of massive disruption to our work, way of life and culture.

Some of these AI Evangelists are pessimists, claiming AI as an existential threat, predicting job losses, economic break-down and worse. Others are more optimistic, seeing AI as ushering in a new era of opportunity and growth, albeit with a potentially painful phase of transition.

Both foresee the technology having a significant impact either way.
Then there are the AI Sceptics, people who see Gen AI as just another tech-driven hype bubble that is bound to burst.

I started off a sceptic. I thought ChatGPT was an interesting novelty but a fundamentally flawed party trick. As I got deeper into the underlying models and explored how they can be applied in more advanced ways I’ve come to realise we really could be at the start of a fundamental change.

The reasons for my change of heart are not necessarily that easy to see based on exposure to tools like ChatGPT alone. So I thought it might help to explain the logic behind my transition from sceptic to evangelist.

I’m calling it ‘LADDER LOGIC’ - the higher up the Gen AI ladder you climb and the more you see, the more apparent this fundamental change becomes.

I know I may not convince the skeptics out there, but at least you’ll get a sense of the logic behind my evangelistic views…

Here’s my ‘LADDER LOGIC’ for being an (optimistic) AI Evangelist…

(Anyone far more knowledgable than I am on the technical side of things will hopefully excuse how I have rationalised and described some of the principles here, but it helps me to navigate what I think is going on)

STEP 1 - Models

On this first step you know there’s such a thing as generative AI and large language models and you’ve heard people talking about how amazing or rubbish they are, but you haven’t tried them yourself. Based on anything else you’ve ever seen it’s hard to image something that will make a whole lot of difference to the world.

STEP 2 - Questions

You’ve jumped onto one of the tools and had a go. Maybe been on ChatGPT and asked some questions. It’s pretty cool that you get answers and some of them are impressive, but other less so, some completely wrong. It feels more like a novelty party tick - but not good enough or consistent enough to be relied on for anything serious.

STEP 3 - Prompting

You get more experienced and sophisticated with your prompting - maybe using personas to add context, guides for what you want you outputs to look like, examples of what good responses are. This makes the results more helpful - it can help boost your productivity a bit, but you can’t see how it’s going to change the world.

STEP 4 - Sequencing

Now you’ve discovered that prompts entered in sequence can link together to build on and evolve the responses that go before it. You see it’s possible to get to a better place following the right sequence. The value you see in Gen AI is growing.

STEP 5 - Data

You’re now giving the model specific data to look at and analyse when you’re prompting. The better and more focussed the data is, the greater the impact on the results you’re seeing. Gen AI is being a lot smarter.

By now you are probably starting to reach the limit of what’s possible using a chat-based system. You can use GPTs to save your prompts and data which is a shortcut, but there’s a lot of time and effort involved in doing what you need to do, and the chat interface isn’t always stable, reliable and repeatable.

At this point being able to work with the models direct through APIs helps.

STEP 6 - Frameworks

Working directly with the model via API allows you to overlay your own logic frameworks and models of thinking, tailoring the general models to make them better at tasks specific to your industry - the level of smarts you see improves dramatically depending on the nature of the frameworks you are trying to overlay.

STEP 7 - Fine-Tuning

You can now fine tune your models, taking output, correcting it and feeding it back so the results you see get smarter and smarter. The more repeatable your tasks, or generalised the learning from your feedback, the smarter the model will get.

STEP 8 - Automation

This is where the benefits compound. When complex steps of analysis and creativity, over dozens or even hundreds of steps, are combined and automated it’s possible to get really sophisticated deliverables from AI.

STEP 9 - Collaboration

Just as human teams often do their best work when they collaborate, the same can be true of AI models. A teams of models developed around different data and areas of expertise can interact and build on each others thinking and ideas, elevating what they are capable of.

STEP 10 - Reasoning

Early on in your journey up the ladder, the fact that Gen AI is ‘just a predictive model, anticipating the next word’ was a good reason to believe there were serious limits to its potential. As you climbed the ladder you saw how powerful that approach could be.

Now you start to understand that models of reasoning can be applied to improve performance further. Juts as you would read, pause, have a moment of reflection to adjust your thinking then move on to learn and build more, these models are capable of doing the same. It’s like adding an inner monologue to the models. Combining the next word predictive power with built-in reasoning creates a step change in what’s possible.

STEP 11 - Autonomy

Typically in your journey so far these models have been working under the direction of humans setting the agenda and lines of any enquiry. Increasingly it’s possible for Autonomous AI agents to take broad objectives and design their own approaches and processes for solving problems.

None of the improvements I describe rely on the underlying models created by the likes of Open AI and Google getting any better - they become more powerful by travelling up the ladder. But the models themselves will get smarter too, compounding the impact. Predictions of AGI (Artificial General Intelligence) currently range from 7 months to 5 years amongst those who believe it’s coming.

When I was on steps one or two I was an AI Sceptic. With each step up the ladder, seeing what’s possible, and knowing that AGI is potentially on the horizon in the base models, I have turned from an AI Sceptic into an AI Evangelist. And a positive one.

Whilst I believe disruption is coming I think there’s an opportunity to reinvent broken agency business models and supercharge the direction of the insights, strategy and innovation industry.
When I post on LinkedIn, some people think I’m over-stating the impact of AI. Maybe they’re right. But from where I stand on the ladder I don’t think so.

For those on lower rungs the scepticism is understandable. My goal when I write about AI isn’t to convince people that the future I talk about is true, it’s to encourage people to want to climb higher and see for themselves.

We’re going to continue to explore and push the boundaries 100% Virtual insight, strategy and innovation projects can deliver. The possibilities are so exciting.

If you want to know more about ONE Strategy Studio and our vision for AI driven virtual projects in the insight, brand strategy and innovation space then please do get in touch.