AI is here to stay

I don’t think of myself as much of a futurist. Predicting the future is a mugs game. You only need look at all the predictions, forecasts and prophecies that never come to pass to know that we are pretty terrible at that sort of guessing. Actually… that’s probably why we so rarely go back and check ‘experts’ when they make a prediction. We know it’s likely to be wrong and we blithely look for the next prediction to sooth our needs for certainty and security.?

Nonetheless, 8 years ago I made a presentation at IMARC titled “Geology, Innovation and Technology” and today I revisited those slides. It’s pretty tame as these things go. The audience was largely non-technical and I was trying to look towards big picture trends and industry needs. There’s a link to a pdf of the slide deck below so help yourselves.

What was I thinking in 2015 and how much has changed? Did I get things right or was I entirely off the mark? Those were the days late in the life of QG when the market was tough for us and the future of the organisation looked uncertain. We had been acquired by ARANZ Geo, the precursor to today’s Seequent, the home of LeapFrog Geo and other apps. I’m sure that coloured my thinking a bit. Circumstance always does.?

Here are the key messages from those old slides:

• The business of geology in all its wide variety is fundamentally about three things
• Identifying unique aspects of the earth’s crust that have some potential value
• Quantifying the nature of that value, be it tonnes and grade or some less standard value proposition
• Clearly communicating these facets to create opportunities?
• The central challenge behind those three items -?Identifying, Quantifying and Communicating - is the intersection of extreme uncertainty and natural variability. It is at that intersection that the true value lies if we are willing and able to see it.
• Realising that value requires understanding the properties of the rocks wherever there is a geological variable that affect the value chain. It is not just about exploration or resource estimation or grade control, in the resource industry the rocks are a material value driver and knowing the right things about the right rocks can enhance performance and mitigate risk

From that background I moved onto innovation in a geological context. I identified three high-level themes:

• Data creators
• Data receptacles
• Data Analysers

It was these three themes that needed most attention for anyone looking to innovate the geosciences.?

In the realm of data creation I foresaw an avalanche of new data, all derived from sensors and creating unique and new challenges. Multi-spectral, automatic, and more exhaustive. Removing the undependable human factor from our observations. More data with enhanced knowledge of the metadata - how the measures were influenced by external factors.

For the data receptacles my vision was for close-to-source acquisition with automated and contextual validation allowing realtime error management. I looked forward to benefits from in-process outlier identification and noted that unless the then current standard of data management was not up to the task much less capable of handling the needs of the future.

When it came to the data analysers theme the wish list started. I discussed concepts such as quick and robust decisions processes, exploiting ‘data exhaust’, dynamic and on-the-fly models, connecting decisions and execution with minimal human intervention, and rapid feedback loops.?

All of these hopes and predictions were presented in the clear understanding that the then current world was already struggling in many areas. We did not make the most of what technology was already available, so some fundamental change would be required to push us from our comfort zone into a brave new world. I talked about the need to shift from thinking about ‘averages’ and instead apply probabilistic thinking, considering the variance, prior probabilities and system constraints.?

I think I was cautiously hopeful about the future….?

Were my predictions good? I guess that depends on how you look at them and your thoughts about our current world.?

When I read through those old slides there is the feeling of a core idea. While it’s not explicit I can feel my younger self beginning to explore a world where machine learning and ‘artificial intelligence’ is common place. Where ML/AI work together with we humans to improve decisions, make faster decisions, enhance execution and learn from our performance.?

Which brings me to some new/old thinking around innovation, and AI.

I think the framework still works. We can still think of things in terms of how our data is created, stored and analysed. What’s more, I think we still have some big challenges in each space. We have not kept up with the demands (needs?) or the AI-centric paradigm.?

We still lack data, we still undervalue data and our thinking is mired in the cost of acquisition and valuing data in old ways. Today’s value of a drill hole, a sample, a map is fundamentally different to what it was 10 or 20 years ago. But we fail to recognise that change in value. That was obvious at the recent MREC 2023 where there were multiple presentation on drill hole spacing - all valuing data in terms of conventional estimation practices with all its limitations.?

We should be thinking about data in terms of its future value. It’s potential in a world of commonplace machine learning. That is, to me at least, a profound shift in mindset. Profound enough to make drilling more holes and collecting more data from those holes not only desirous but necessary. Is our industry ready for an explosion of drilling? Where will the rigs and the people come from? What technology can we bring to bear on drilling to make it faster, more effective and more socially acceptable??

Data management is still fractious. While there have been some improvements these are largely still proprietary systems. We need a strong platform with common standards not this world with constant import/export, worrying over formats, data types and other structural limitations. We need data stored alongside its metadata keeping us (and our AI) fully aware of quality, representivity and other vital knowledge. We need to be able to exploit what may seem to be completely unrelated observations, combining data sets in new ways to create new insights and identify new connections. Data in isolation is not really ‘data’ in today’s world. It is more akin to observations.

It is the realm of data analytics that has garnered the most interest. The explosion of ML and AI across multiple industries and in use cases we barely dreamed of 10 years go, is profound.?

If there is one prediction I am confident of making it is that AI is here to stay.?

What does that mean for us? We need to be prepared and to take hold of the technology. Those of us who are well prepared will see advantages the less prepared cannot fathom. I believe AI is a transformative technology for the resource industry. It has the potential to enable step-change improvements in productivity, countering the downward trend of the last 20+ years. It has the potential to reduce systemic waste and maximise the value of scarce resources.?

But that change will not come without pain, without winners and losers. It’s going to be a bumpy ride - what else is new?

So how do you get in position to take advantage of this sweeping change? My bet is on the data. It’s time to focus on the data. Everywhere. Quality, quantity, completeness, consistency. Remove human-derived error where possible. Collect at source, collect once and collect multiple variables at the same time.?

AI, impressive as it seems (at least to the media) is still an immature technology. We might be better calling it ‘artificial stupid’. In narrow fields it is powerful but it still takes a human to see the big picture, do some reality checking and make choices about the future.