If AI is the hammer, every problem is a nail

Mark Piper, Principal Explorer

30 April, 2024

On Hype Cycles

If you work in technology long enough, you get to witness hype cycles and watch in slow motion as those cycles leave burn marks across industries.

We don’t have to look hard to find some examples in recent history. Whether it’s 3D printing, Web3, the Internet of Things, Virtual Reality or wearable technology - history is littered with periods where vendors told us that the future is here, and we’ll all be changing the way we work for decades to come.

However, the majority of these hype cycles haven’t really done much to change the way we work. Money is spent, profits are made and quarterly results are smashed but the way we work has not fundamentally changed in 20 years.

Take augmented or virtual reality as an example. For decades we have been promised a new, evolving work experience where we can meet face to face via the metaverse.

I’m not sure who’s sitting around today as a cartoon avatar discussing April's P&L numbers, but I’m guessing that those who are, won’t hit their targets.

The only significant exception to the failed hype is remote work, which worked rather well for a percentage of the population - and even that is currently being reversed in many organisations in order to keep the old ways of working alive.

As a general rule, organisations stick to what they know. Most organisations resist dramatic change.

So when it comes to the current Artificial Intelligence hype cycle (current, as there have been many over the last 20 years), one can be forgiven for thinking that this is going to be the same trajectory of other recent ‘game changers’ such as web3 - straight off the cliff, no brakes applied.

The problem is, this cycle is different.

Commonly during cycles, vendors are quick to state the benefits of a technology, why you should buy it and how it might be integrated into your business. Like a lawyer in front of a jury, they state their case. They convince you that it’s worth exploring. They tell you of the possibilities. They sell a dream. A narrative.

Same Same, But Different

With this hype cycle however, the message from vendors is unified, and a lot more direct.

They are not asking if you would like some AI with that enterprise burger, they are saying that you should take the AI or else you won’t be in business by 2027.

This is not a new tactic, but it is rare. I believe the last time we saw such an aggressive stance was the migration of 32-bit to 64-bit processing. Then, similar to now, the vendors offered little choice and made it clear, the way we compute was changing. 64-bit is here to stay. Get spending.

We are at a time where the vendors are stating that inference computing is here to stay. Transformer architecture is coming and organisations must get ready now to integrate AI across their landscape in the coming weeks, months and years.

And they are not mucking around. As NVIDIA’s CEO Jensen Huang recently said at the 2024 GTC event:

“We need another way of doing computing — so that we can continue to scale so that we can continue to drive down the cost of computing, so that we can continue to consume more and more computing while being sustainable. Accelerated computing is a dramatic speedup over general-purpose computing, in every single industry.”

So any decision maker can be forgiven for suddenly thinking that they must not only understand AI as a capability, but also get working now on how to implement it.

It’s no surprise that they find themselves wondering “What is actually happening with AI and how can it actually help?”.

This is the core of what we have been looking into over the last few months in the lab.

Starting with building a timeline of AI, understanding the different types of AI and identifying the recent advancements in AI - we put on our safety goggles and waded into the hype to try and understand what is the actual state of AI and what can it actually do?

And we worked to make sure we understood the big picture beyond the tech.

What’s the impact on humans and teams? What’s the impact on society? What are the ethical considerations? What does it take to make it work? How can it be integrated? What’s on the market now?

The answers are as complex and nuanced as you might expect for such broad questions across such new technology and furthermore, it varies across industry verticals.

After conducting our initial research, we ended up splitting problem atoms and focusing on two key hemispheres.

The first, understanding what is possible with narrow (weak) AI. That is, AI applied to a specific problem. How can we solve for X?

The second, looking at Large Language Models and Generative AI to understand how they augment human workflows?

In Wielding the AI Hammer, I hope we have provided some high-level insights for those looking at the first question. We look at what needs to be considered when throwing AI at problems - like darts to the kanban.

We will touch on the second question in another post (although you can get a taste of our thinking in our recent post The Productivity Paradox).

Why Now?

Covering a subject as complex as AI is tough.

Figuring what is relevant today vs what might be relevant tomorrow, or worse, not relevant at all has been the subject of much discussion during our work.

One of the most fascinating observations from our research is that everyone is suddenly expected to act like they knew about AI all along.

Play it cool. Everyone knows about this AI thing right?

Models?
“Please, we’ve been working on them for years”.
Training?
“That’s a Tuesday around here, mate”.
Data provenance?
“We have the best data. The most clean data. Bias free since ‘03”.

But the reality is, this is new.

It’s new to us and it’s new to most enterprises. We are right now implementing new solutions largely based on the last 6 years of development and significant investment. On the timeline of computers, that’s not very old at all.

The fact is, no-one seems to want to admit that they do not have a long heritage in AI because it might make them look bad. The hype cycle is real and it influences perception.

I even had a conversation recently where it was made clear that if you are a business seeking investment and you don’t mention AI - you’re not getting funding. End of discussion.

Maybe this explains how Logitech, a hardware company who make keyboards, mice and webcams recently released their AI Prompt Builder - software that, with a single binding, lets you be a pro at ChatGPT.

We did mention that we are in the middle of a hype cycle.

It’s for these sorts of reasons we decided to cut through some bullshit and publish this report.

We don’t expect everyone to be an expert. Our goal here is not to cover every aspect of what’s required to successfully implement AI across the enterprise. We simply want to help those making decisions, to be informed as they do so.

What is a model? How is it trained? Why does data matter? What are the risks? How about the ethics of it all? How much is my budget going to sink after hitting the inevitable TCO iceberg?

These are, we believe, the questions keeping leadership and teams awake at night.

We sincerely hope that this work is useful to you, your leadership and the ongoing work to implement AI in your organisation in the months and years ahead.

If you have any feedback, or would like a briefing for you and your team, please let us know.