Digital Transformation Was Yesterday’s Adventure
What really is “digital transformation”?
It’s when you have automated everything that can be automated, so that everything is controlled by digital software.
In a digital environment, robots do all the moving of parts. Machines make product parts, controlled by digital designs. The designs are created by engineers using computers.
The entire process is orchestrated by digitally-defined programs. There are few human handoffs.
Information flows in and out of the organization digitally. Suppliers connect to a logistics system that is up to date on your needs, based on data that is automatically provided by your manufacturing systems.
There is little manual entry of data.
All that is the foundation for the next big shift.
What Comes After Digital Transformation
The next big shift is in the creation of the digital content.
The content that drives the factory. The content that drives your digital marketing. The content that updates your cloud based offerings.
All your digital artifacts. The things that, today, are created by people.
People will still be in the loop, but they will be directing, not creating.
At least for now. When people are no longer in the loop, we can all go home. (Oh wait – we are already working from home!)
There has been a lot of conflicting reporting about the latest large language model AI systems. Can they be trusted? Are they ready for prime time?
What we need is a rigorous study to find out. And Microsoft has done that. They report that,
“GPT-4 can solve novel and difficult tasks that span mathematics, coding, vision, medicine, law, psychology and more, without needing any special prompting. Moreover, in all of these tasks, GPT-4’s performance is strikingly close to human-level performance, and often vastly surpasses prior models such as ChatGPT. Given the breadth and depth of GPT-4’s capabilities, we believe that it could reasonably be viewed as an early (yet still incomplete) version of an artificial general intelligence (AGI) system.”
AI systems are not digital programs. This is not just more sophisticated software. It was not the software industry that created AI.
Recently Erik Brynjolfsson, a leading, technology-focused economist based at Stanford University, together with MIT economists Danielle Li and Lindsey R. Raymond, released an empirical study of the real-world economic effects of new AI systems. They looked at what happened to a company and its workers after it incorporated a version of ChatGPT into their workflows.
What happened is that the junior people became much more productive. For them, it was like having an experienced assistant next to them.
It’s Not Digital
This might seem like an inevitable extension of computing: first computers automated routine things, and now they are actually smart in some ways and can do some of the more creative things. But it’s not like that.
These AI systems are not computers. They are not digital programs. This is not just more sophisticated software. It was not the software industry that created AI.
Today most AI runs as digital software, but that is a peculiarity. It is a workaround. It is temporary.
Yes, today most AI runs as digital software, but that is a peculiarity. It is a workaround. It is temporary – an interim step. We use computers to run AI because computers are the tools we have, but doing that is massively inefficient. It is kind of like using horses to simulate a car.
Today’s AI consists of neural networks. A neural network is a connected system of “neurons”. These are idealized: they receive signals from adjacent neurons, and compute a function that determines if they will “fire” to signal other connected neurons.
They are not digital. We use computers to simulate them. But the actual neural networks that we are simulating are not digital. What we need is real neuronal devices that we can connect together to create real neural networks.
Guess what? We have that: they are called neuromorphic chips. IBM has been a pioneer in that, but there has been immense progress recently, creating chips that can adjust their basic architecture. That way, they can be repurposed for different kinds of neural networks.
The BrainChip – Part of the Next Generation of Neuromorphic Chips
In the not-too-distant future, neuromorphic systems will overtake digitally simulated neural networks. Why? Because they are much faster, and use thousands of times less power – and that’s important for self-driving cars, mobile devices, robots, and every kind of real world use of AI that is not situated in a data center.
There are even companies using real neurons – biological ones – to create neural networks. Cortical Labs is one. (And an intriguing question is, if such a system tells you that it is sentient, can you confidently say no? After all, it is actually alive, just like we are.)
So the future is not about digital transformation. That was yesterday.
Researcher at Cortical Labs, examining the electrode connections to real neurons in a living neural network
Digital is needed as a baseline for the AI to be able to do things, so that the AI can reach out and control things. But digital is now the assumption. It is very late in the game for digital transformation. People are now doing AI transformation. If you are still stuck in a digital transformation, you have a lot of catching up to do!
