Planning for the unimaginable

We are in the Era of AI. It will change our economy and society. But how quickly and how hard will this change actually hit? Perhaps it will bring modest change that simply empowers existing people and organizations to be stronger, better, and faster. In this scenario, the faster and more adaptable will simply outperform the slow movers.

But I think it’s far more likely that we’re in at least a period of transformative disruption, where the rules are fundamentally different. In this scenario, many companies are headed for a precipice, and don’t yet realize it. They’ve been lulled by an extended period of incremental innovation, and lack the agility and foresight to adapt to the new reality. 

As I argued in late 2022, there’s also a non-zero chance we’re talking about a post work economy. The very nature of firms and why they’re needed might erode. We might be headed for radical change.

Most organizations plant their heads in the sand when presented with such a frightening range of outcomes—the status quo is much easier than reimagination. The easiest thing to do is focus on what you already know, and build a single forecast based on it. 

But during times of significant disruption, a single forecast is nothing more than a false certainty. The best prepared organizations identify realistic scenarios (including low likelihood but high impact ones), and proactively prepare with an eye towards the assessment of risks and rewards.

Scenario Matrix

What scenarios should we be considering? I propose two vectors: level of disruption; and, speed of progress. Each vector has three distinct values.

Speed of progress

Speed of progress refers to how long (5, 10, or 20 years) it takes to realize the maximum attainable level of progress (defined below) within the next 20 years. It doesn’t imply how long it will take for that technology to diffuse into the market. And to be clear, this means the slowest time frame effectively means very little progress at all.

Level of disruption

Level of disruption refers to the maximum level of AI technology advancement that will be achieved within the next 20 years. It doesn’t necessarily imply that technology will diffuse quickly into the market (although it might).

Modest means we’ve seen most of what we’re going to get out of AI and automation. Further improvements will happen, but they won’t be particularly significant. This is a step change, but it’s about augmentation rather than replacement—for people, organizations, and systemic models.

• AGI (Artificial General Intelligence) is at least 20 years off, and skilled humans retain an edge in many knowledge tasks
• AI-powered physical agents are fairly limited in their capabilities and humans remain a better choice for most physical tasks
• AI augments R&D but does not transform it

Transformative means AI is capable of transforming most jobs and industries. It’s about more than augmentation; it will enable significant displacement and much of that implies replacement—for people, organizations, and systemic models. Many fewer people are required for most job roles. 

• Lightly supervised machines can perform at human levels at all knowledge tasks
• Lightly supervised machines can perform most simple physical labor tasks at human level (e.g., fold clothes, move boxes, drive a car, run a 5k city race)
• AI at least doubles the pace of R&D

Radical means the role of humans in our economy will fundamentally—and permanently—change. Surprising new paradigms will emerge, and many existing organizations will be replaced or at least radically altered. 

• Unaided machines can accomplish virtually any task (including physical ones) better and more cheaply than human workers
• AI drives at least an order of magnitude increase in the pace of R&D

Arguably, there is too much of a leap between modest and transformative. But I believe it’s appropriate because if AGI (Artificial General Intelligence) or its equivalent is achieved, I foresee the transformative scenario as becoming inevitable. Either we’re stuck within our current paradigm, or we break through to AGI. The question then comes down to the speed at which supporting technological advances support broader automation.

Radical change implies that our current exponential rate of technology advancement continues unabated. In this scenario, there are few limits to what AI and automation can accomplish. 

Usable AI scenarios

As usual, the full matrix of scenarios should be collapsed for simplicity, and based on the fact that some are either strongly correlated or inconsequentially different (or both).  

Radical change within ten years naturally implies transformative change within five. And given that there’s so much uncertainty accompanying radical change it probably makes sense to focus on handling transformation and preparing for further adaptation as feasible. In other words, “Radical Moderate” has such similar implications as “Transformative Fast” that they should be combined into “Fast.” Similarly, all of the “Modest” outcomes and “Transformative Slow” are effectively the same as “Slow.” 

Overall, I believe the meaningful scenarios are:

• Slow: Transformative change either never happens, or not within the next 10 years
• Moderate: Transformative change in the next 10 years
• Fast: Transformative change in the next 5 years (and maybe radical within 10)
• Radical: Radical change within the next 5 years

The Slow outcome appears unlikely given the exponential rate of progress in AI. However, it’s possible that limitations in models, data, or regulatory barriers to training effectively quash meaningful advances in the coming years.

Radical appears unlikely as well, although it’s hard to count it out entirely. As of 2023, a large survey of AI experts predicted a 50% chance of HLMI (High-Level Machine Intelligence, in other words, our Radical scenario) by 2047, and a 10% chance by 2033. The experts in the field think HLMI is likely to come within 23 years, and their timelines keep compressing (13 years earlier than 2022’s consensus). Given the compression in timeframes over time, and that even now there’s a consensus 10% likelihood of HLMI in under 10 years, the Radical outcome can’t be entirely discounted. 

The most likely scenarios for now are Moderate and Fast. I suspect most business leaders will land on Moderate as the most likely outcome. But don’t forget that we’re only talking about technological advancement, and two further overlays are required:

• Diffusion speed (how quickly the tech makes its way into the market)
• Implications by business model and industry

No matter how quickly the technology changes, diffusion into the market will take further time and will affect different businesses in different ways. Given that our scenarios focus merely on the timeline to technological capabilities, I think there’s a strong argument in favor of a Fast being at least as likely as Moderate. 

As a reminder, Fast would imply that within five years:

• Lightly supervised machines can perform at human levels at all knowledge tasks
• Lightly supervised machines can perform most simple physical labor tasks at human level (e.g., fold clothes, move boxes, drive a car, run a 5k city race)
• AI at least doubles the pace of R&D

I think the odds we’ll see most knowledge tasks and physical labor actually replaced by AI and robotics within 5 years are vanishingly small. But I think it’s likely to be proven technically and practically feasible within that timeframe, and the race will be on. Will you be ready?