How to Prepare for Tomorrow's AI
Not another article about AI!
But I promise you: this one is different.
There are people who think that AI is nothing to worry about. And there are people who think that AI will end civilization. (In a personal article I explain why people have such divergent views on the existential risks of AI.)
In between there are people who claim that AI will eliminate some jobs but create more, and there are others who claim that it will eliminate all jobs and crash the global economy!
This article explores what the period of transition, after the arrival of post human-like AI, will be like for businesses.
AI will very likely, at some point, surpass humans in cognitive abilities. Some people deny that, claiming that true intelligence requires sentience – awareness. But the evidence is otherwise. The jury is still out on this, but recent success with large language models seems to indicate that AI can “think” in ways that we previously thought were limited to humans, even though most researchers agree that today’s computer-based AI systems, which are actually simulations of neural networks, are not “aware”. (An intriguing thought is that if those systems were composed of actual silicon-based neural networks, might they be aware?)
So we should expect that AI will eventually surpass humans in cognitive abilities, and be surprised if it does not, regardless of the question of sentience. And it will likely happen sooner rather than later – I will get into that later in this article.
My main focus here is that when (if) that happens – that is, when AI becomes “smarter” than us – that change will trigger a socioeconomic transition that will impact every business. This article explores what the period of transition will be like, and how to position your business for it, and why it will be a matter of survival.
AI Has Not Plateaued
If you have used the latest “generative AI” tools like Bard and ChatGPT, you have likely been impressed but also reassured. For example, I have found that conversations with ChatGPT (the more powerful GPT-4 version) are insightful, but that it cannot do things like design an airplane wing for me. We still need engineers. And while GPT-4 can definitely write code, I would not trust it to take on a coding problem on its own: I would want an experienced programmer supervising.
AI not on a technology plateau: we cannot make three year plans based on today’s AI.
So we still need engineers and programmers, although maybe fewer of them. More than ever, we need the good ones, not masses of mediocre ones.
But things will change, again, and sooner than before. The arc of this technology is still on a steep ascent. You might have read that on Nov 23 there was an article in Reuters that claimed that,
“Ahead of OpenAI CEO Sam Altman’s four days in exile, several staff researchers wrote a letter to the board of directors warning of a powerful artificial intelligence discovery that they said could threaten humanity, two people familiar with the matter told Reuters.”
According to the article, this new and improved OpenAI technology is called “Q*”, and that “Some at OpenAI believe Q* (pronounced Q-Star) could be a breakthrough in the startup's search for what's known as artificial general intelligence (AGI)”.
Thus, we are not on a technology plateau with this stuff: we cannot assess what it can do and make three-year plans accordingly, because it is still improving rapidly.
In the past, when technology eliminated jobs, it created other new jobs. That is still the case today, but it will not stay like that. At some point the work economy will experience a rapid “phase change” (also known as a “state change”) in which jobs rapidly disappear – all jobs. Let me explain why.
Generative AI Is Not Like Other Tech
We are not there yet, but at some point soon, AI will pass a threshold at which it will be able to replace human thinking for most if not all things. People will trust AI to get the job done more than they trust people, no matter what it is: accounting, financial analysis, engineering, and even scientific research. You name it. Not to mention less cognitive labor like serving in a restaurant, deliveries, or construction.
That is, if you ask people, “Would you prefer to have an AI or a human oversee this task?” most will answer “an AI”. Again, we are not there yet, but it is coming.
Also, AI will be mobile: it will be wirelessly connected to autonomous robots, making them mobile and smart.
When we cross this threshold – the threshold at which people trust AI to get a job done more than they trust other people – then companies will not need employees anymore.
This is what is known in physics as a “phase change”. A phase change is when the interactions within a system change such that the overall behavior of the system suddenly changes. For example, ice melts when the average energy of the molecules exceeds a threshold at which point the molecules are no longer confined to a rigid structure. At that point, the overall behavior of the water changes from being solid to being fluid. This phase change behavior is explained by theories in the field of statistical physics.
The analogy to AI is that a molecule’s average energy can be compared to the average confidence in AI compared to employees. At that point, there will be a phase change, in which companies stop hiring humans. Why would they hire humans when AI is better?
Throughout history, advancing technology has empowered humans. But at some point, the technology will cross a threshold, and humans will become irrelevant. There will be a phase change: instead of the past repeating itself to create new kinds of jobs, the new jobs will be given to AI instead of people, and suddenly human jobs will begin to decline. The overall behavior of the system as an innovation-driven job-creating machine will have passed an inflection point. The system will rapidly change to one in which there are no new jobs for people at all, and companies will start replacing the remaining people with AI as the companies learn how to incorporate the new AI.
At some point the work economy will experience a rapid “phase change” in which jobs that were plentiful rapidly disappear.
The period of “phase change” will not be instantaneous. I expect it will last a few years, perhaps even a decade. But once it begins, it will be irreversible, and it will be global.
During that time, business will be increasingly chaotic. There will be no new “normal” until the phase change is complete. Long-term planning will be impossible. And the new normal will look nothing like today’s “normal”.
In the new normal, there will be no jobs. What life will look like in a societal context is a matter of debate – there are utopian and dystopian views – but companies will not employ people. People might work, but it will be voluntary. And the business environment will continue to grow ever-more fast-paced and chaotic. The endpoint of that increasing pace is another topic that I won’t get into here.
Business During the Period of Transition
During the period of transition, products will change frequently. Scientific research will experience an AI-enabled renaissance. Companies will come and go with increasing frequency as new products become obsolete in record time.
Remember the crypto craze of a year ago? And remember how fast we created a vaccine for the SARS-COV-2 virus? And how quickly electric cars have taken over the market, with Tesla now outselling all other passenger cars? These changes will seem glacial compared to what is coming, when AI begins to power innovation.
In short, it will be the age of agility. Companies that can adapt rapidly and efficaciously – that is, make good decisions and execute quickly — will be the ones that survive the transition.
The profile of customers will fluctuate with increasing frequency. In the utopian view, the transition will end with everyone having a comfortable basic income, with enough left over to travel and experience their every desire. In the dystopian view, the wealthy will own all the natural resources and have immense buying power, while most people scrape by on a subsistence “dole”.
Today’s rapid pace of change will seem glacial compared to what is coming, when AI begins to power innovation.
Either way, the transition will be a race to build wealth and acquire durable resources, just as today companies try to secure limited supplies of lithium and rare earth metals. The race will be like a Tough Mudder event, slogging through ever-changing terrain. Laws and governments will not be able to keep up, because those systems were designed for an earlier, slow-moving era.
The debate about whether the future is utopian or dystopian is not relevant for this article, nor do I want to get into the existential risks of AI. The only conclusion that we need in order to examine what businesses will need going forward is the conclusion that there will be a rapid transitional period during which human labor – including intellectual labor – will become obsolete, and AI will come to dominate the planning and execution of business.
I am fairly certain that the transition will not begin next year. But I am also quite certain that it will begin earlier than 20 years from now. I give a very high probability that it will begin within the next five years. And this is important: it will take time to prepare.
What Is Needed to Survive the Transition
To survive and flourish through the transition, companies will need:
Great behavioral agility.
Knowledge of how to use AI effectively – and keep up with it as it changes. I.e. continuous experimentation and innovation at all levels – not just driven from the top.
Remember that during the transition, AI will be taking on new kinds of work, be it science or engineering or production.
Even leadership. It is hard to imagine it now, but if you have conversed with generative AI, you know that it displays apparent wisdom. It passes the Turing test – you cannot tell if you are talking to a person or an AI. Now imagine that it has advanced more, and you trust it to make good decisions. Then you will want it to lead new projects, once you get comfortable with the idea. And it will be leading other AIs – not emotional people.
There will be no new hires except for AI experts and the very top-most software engineers that are obtainable. The critical people needed will be the ones overseeing and implementing the gradual transition to an all-AI workforce. They will be putting themselves out of their own jobs as well, so they will need to be taken care of, long term, in order for them to feel fully committed to helping the company during the transition.
There will be no new hires except for AI experts and the very top-most software engineers that are obtainable.
During the transition, while the company is staffed by people, high-agility leadership will be key. Behavioral agility entails the ability to lead, and ask hard questions, and keep track of what is happening in real time. This cannot be only at the top levels of the company: it must be at all levels. The company culture must be agile, dynamic, smart, and it must move like a well-coordinated organism. Delay is death. Fast but correct response throughout the whole organism is the way to survive.
The company’s culture defines its behavior as an organism. It must be innately agile in its behavior. It’s DNA. In how people think, how they approach things, how they lead, how they make decisions.
The time to start creating that culture is now, from the top and through all levels.
Is This Truly Inevitable? Or Perhaps Far In the Future?
We need to give weight to the data. The progress of AI has been steady:
Experts who work on AI systems are reluctant to be too sanguine when predicting future progress. It is like asking medical researchers when they will cure a disease: they are always very conservative in their response. But the data indicates that AI continues to improve rapidly, and has already surpassed human capabilities in many cognitive areas.
Problem solving is really the only remaining cognitive domain in which humans still surpass AI, but according to recent revelations from OpenAI, that might fall soon as well. But we should not over-simplify this. Problem solving is actually a broad range of abilities.
Systems thinkers are holistic thinkers – they can see beyond what we have today, and imagine what will come tomorrow, integrating multiple trends.
For example, there are contextually bounded problems, such as the math word problems that we encountered in high school, and there are unbounded real world problems. Bounded problems require analytical skills. But unbounded problems require judgment. John Mackey, the founder of Whole Foods, contrasts these two forms of thinking as linear “analytical” thinking and “systems” thinking.*
Systems thinkers are holistic thinkers – they can see beyond what we have today, and imagine what will come tomorrow, integrating multiple trends. They are good at solving unbounded problems. The question is, will AI be an analytical thinker, or a systems thinker? Given that AI has already shown the ability to link contextually diverse things (see for example here and here), it is reasonable to expect that it will be an amazing systems thinker. If so, it will have judgment – something that we have assumed was uniquely human.
We have to be realistic. The data tells us that there is a clear trend.
One problem with ChatGPT and other similar systems is that their performance over time fluctuates. That is known as the “aging” problem. But that problem is increasingly well understood, and will be solved, as will other current limitations.
So we have to be realistic. The data tells us that there is a clear trend, and that we will cross a critical threshold sooner rather than later. And the trend seems to be that AI will surpass us in every way – not just some ways. And simple analysis based on comparison with models from statistical physics tells us that the threshold will represent a socioeconomic inflection point.
We must plan according to the best information we have, and not be ideological or willfully blind to what is coming, even though it is our tendency as humans to want to deny what is uncomfortable.
* Sisodia, Rajendra ; Rajendra. Conscious Capitalism, With a New Preface by the Authors: Liberating the Heroic Spirit of Business (p. 184). Harvard Business Review Press. Kindle Edition.
What To Do
When AI passes a certain threshold, it will be more useful than people in nearly every profession, and that changes everything. Even the relationship-oriented professions will eventually fall, when the decisionmaking entities transition to AIs instead of people. For example, sales depends on being able to talk to decisionmakers, but if decisionmakers are AIs, then the whole concept of sales kind of goes away.
During World War II, my mother’s family lived in Vienna, Austria. My mother’s grandfather was the ambassador from Turkey to Austria, stationed at the Turkish embassy in Vienna, so he was up on current events, and had insight about geopolitical trends. He foresaw that the German regime would invade Austria, and so he took the extreme action of moving his entire extended family away, to Brazil, ahead of time. Thus, when the invasion actually came, they were all safe.
The time to prepare for the AI transition is now, and large strokes are needed. The companies that survive will be the nimble ones that are immersed in AI. They will have leaders at all levels who are able to imagine what could be: “systems” thinkers instead of linear “analytical” thinkers, to use the words of John Mackey.
These behaviors are needed in all of the company’s leaders – not just the ones at the top.
These leaders must also be active learners: they don’t treat a new technology such as AI as something “for others to know about” – they explore it themselves and become competent enough to have intelligent discussions about it, and they form connections with experts and make sure that the experts are involved in discussions and decisions.
When Elon Musk started SpaceX, his first hire was Tom Mueller, a rocket engineer. But Musk did not say “build me a rocket engine” and then walk away. Instead, Musk learned all he could, and it was Musk who made the critical decision a year later to shift from an ablative design to an actively cooled design: that became the “Merlin” engine that powers today’s highly profitable Falcon rocket series. Musk did not take over engine design, but he learned, paid attention, and had a keen sense of when to involve himself.
All the above is behavioral: none of it comes from any management process or “Agile framework”. It is all about leadership style.
These behaviors are not needed just among the top leaders: the behaviors are needed in all of the company’s leaders. Otherwise, the agile parts of the organization will be held back by the rigid parts.
It takes time to change behavior and knowledge. And there is no time to waste.
