OSP Services
ISP Services
Project Management
Telecom Wireless
DAS Micro CellThe AI revolution may take longer than you think
The AI revolution may take longer than you think
Avi Goldfarb, one-third of the writing trio that has just published Power and Prediction, says: “The benefits of AI [artificial intelligence] come when we use it to do things differently. Too many AI applications are what we call ‘point solutions’, where you start with an existing workflow, and insert a machine to replace a task done by a human.” The problem with this, says Goldfarb, is that it’s the same old anchor that’s weighed down so many companies looking for change: we’ve always done it this way.
By inserting AI into an existing system, unless you’re extraordinarily lucky, “the best you can do is what you’ve always done, but a little bit better”. What you need is a ‘system solution’, in which “AI enables new workflows and new ways to deliver value”. The concept of adding value rather than integrating new technology for its own sake is crucial because, as the book’s subtitle – The disruptive economics of artificial intelligence – tells us, Power and Prediction is as much about the commercial future of the organisation installing AI as it is about the technology itself (which was neatly summarised in their earlier Prediction Machines.)
Prediction perspectives
It is this focus on economics that is a differentiating characteristic of yet another volume on AI. To write about AI, you need a unique selling point to stand out in a crowded market: Goldfarb believes his to be the disruptive economics stance. Referring to his co-authors, he says: “We’re economists. That gives us a different perspective. We view today’s AI as prediction technology, and over the past decade machine prediction has gotten much cheaper. By prediction, we mean in the statistical sense: the process of filling in missing information.” As the cost of prediction falls, says Goldfarb, we use it more. “This enables new opportunities. We draw on research in decision theory and the economics of technology to understand the implications.”
Going back to Prediction Machines (reviewed in E+T, April 2018), Goldfarb says that he “expected a technology revolution to happen. As time passed, we started to wonder what was taking so long.” Discussing the time-slip with Stanford economist Tim Bresnahan, Goldfarb came to realise that “the most impactful technologies in history have required decades of innovations in order to have a meaningful impact on the way we live and work”.
In the 1990s, Bresnahan coined the term ‘general-purpose technology’ to describe electricity, the steam engine, the computer and similarly transformative technologies, all of which required system-level change (see extract, right). At that point, continues Goldfarb, “we realised we had something important to say – describing how economic history could explain the current moment, while providing a guide for organisations to make the necessary changes”.
Resistance to disruption
Goldfarb thinks that there will come a time when AI takes its place on the list of cardinal general-purpose technologies, and fulfil its purpose of transforming the way we live and work. The problem is that such transformation takes time: “It took 40 years from the widespread recognition of electricity’s potential in the 1880s until it appeared in most US homes and factories in the 1920s.” These four decades witnessed innovation-led system-level change in factory design, electricity generation and distribution, electric home appliances, safety improvements and a variety of other aspects of home and work. But in the case of AI, the future could arrive sooner.
“It might not be 40 years,” says Goldfarb, “but such colossal transformation takes time. We need to figure out what an AI-centred organisation looks like in many industries, and what the AI-centred home involves. We need to determine what the rules should be, weighing important concerns about privacy and bias against the benefits of a tool that might enable better healthcare, education, and overall standard of living.”
As with all transformative technology, there will be winners and losers. The biggest winners, says Goldfarb, “will be those that figure out new systems for delivering value to their stakeholders. The biggest losers will be the incumbents in those industries that are disrupted.”
But often, he says, the industries with the most potential to benefit from AI are those hardest to change. While there is “enormous potential” for AI to transform healthcare – improving patient care and enabling healthier populations – such transformation “will encounter resistance from those who benefit from the current healthcare system”.
If we had regarded managing social interaction during the pandemic as a prediction rather than health problem, “if we knew who was infectious, then we’d keep them at home and the rest of us could go about our business. It is only the lack of information that makes a pandemic consequential.”
Taking the next step towards an intelligently automated business might sound like a technology challenge, but “the AI isn’t the hard part. The challenge is figuring out how to get all the other pieces aligned in order to do something differently. A system solution will often change who has control over decision making. It is that change in power that is disruptive and often leads to resistance.”
Meanwhile the economics are relatively simple: “When the price of machine prediction falls, people do more machine prediction and less human prediction, particularly with respect to decision making.”
Extract
When an AI-driven decision is part of a system, adopting AI can necessitate an organisational redesign with a new system. One difficulty existing organisations face in creating new systems is that they have been optimised to garner high performance from existing technologies, whereas adopting AI can necessitate a change in focus. In some cases, AI drives the organisation to become more modular, while in others, it can drive it to have greater coordination among the parts. The challenge is to recognise that the current focus is the problem, and widespread change is needed.
When top management understands that a new organisational design is needed in order to adopt and integrate an AI prediction to one or more key decision areas, a further challenge arises. This is because organisational design invariably involves a change in the value and, hence, power of the suppliers of different resources within the organisation. Those who expect to lose in the expected reallocation of power will resist change. Organisations rarely operate as a textbook dictatorship where what the CEO says goes and change happens. Instead, those expecting to have their power diminished resist change. In the process, they can undertake actions that at best reduce the ease by which change can be implemented. At worst, the anticipation of those actions may cause an organisational redesign to be curtailed completely or reversed.
Consider the example of Blockbuster Video. Blockbuster was the market leader in video-tape rentals throughout the 1990s and 2000s. The commonly known narrative of Blockbuster’s demise is that it was felled by Netflix and the rise of on-demand video. But in fact, Blockbuster did not passively succumb to the new ways. It understood what was coming but ultimately failed to adjust to it.
Edited extract from Power and Prediction by Ajay Agrawal, Joshua Gans and Avi Goldfarb, reproduced with permission.
Power and Prediction by Ajay Agrawal, Joshua Gans and Avi Goldfarb is published by Harvard Business Review Press, £22.00.
Avi Goldfarb, one-third of the writing trio that has just published Power and Prediction, says: “The benefits of AI [artificial intelligence] come when we use it to do things differently. Too many AI applications are what we call ‘point solutions’, where you start with an existing workflow, and insert a machine to replace a task done by a human.” The problem with this, says Goldfarb, is that it’s the same old anchor that’s weighed down so many companies looking for change: we’ve always done it this way.
By inserting AI into an existing system, unless you’re extraordinarily lucky, “the best you can do is what you’ve always done, but a little bit better”. What you need is a ‘system solution’, in which “AI enables new workflows and new ways to deliver value”. The concept of adding value rather than integrating new technology for its own sake is crucial because, as the book’s subtitle – The disruptive economics of artificial intelligence – tells us, Power and Prediction is as much about the commercial future of the organisation installing AI as it is about the technology itself (which was neatly summarised in their earlier Prediction Machines.)
Prediction perspectives
It is this focus on economics that is a differentiating characteristic of yet another volume on AI. To write about AI, you need a unique selling point to stand out in a crowded market: Goldfarb believes his to be the disruptive economics stance. Referring to his co-authors, he says: “We’re economists. That gives us a different perspective. We view today’s AI as prediction technology, and over the past decade machine prediction has gotten much cheaper. By prediction, we mean in the statistical sense: the process of filling in missing information.” As the cost of prediction falls, says Goldfarb, we use it more. “This enables new opportunities. We draw on research in decision theory and the economics of technology to understand the implications.”
Going back to Prediction Machines (reviewed in E+T, April 2018), Goldfarb says that he “expected a technology revolution to happen. As time passed, we started to wonder what was taking so long.” Discussing the time-slip with Stanford economist Tim Bresnahan, Goldfarb came to realise that “the most impactful technologies in history have required decades of innovations in order to have a meaningful impact on the way we live and work”.
In the 1990s, Bresnahan coined the term ‘general-purpose technology’ to describe electricity, the steam engine, the computer and similarly transformative technologies, all of which required system-level change (see extract, right). At that point, continues Goldfarb, “we realised we had something important to say – describing how economic history could explain the current moment, while providing a guide for organisations to make the necessary changes”.
Resistance to disruption
Goldfarb thinks that there will come a time when AI takes its place on the list of cardinal general-purpose technologies, and fulfil its purpose of transforming the way we live and work. The problem is that such transformation takes time: “It took 40 years from the widespread recognition of electricity’s potential in the 1880s until it appeared in most US homes and factories in the 1920s.” These four decades witnessed innovation-led system-level change in factory design, electricity generation and distribution, electric home appliances, safety improvements and a variety of other aspects of home and work. But in the case of AI, the future could arrive sooner.
“It might not be 40 years,” says Goldfarb, “but such colossal transformation takes time. We need to figure out what an AI-centred organisation looks like in many industries, and what the AI-centred home involves. We need to determine what the rules should be, weighing important concerns about privacy and bias against the benefits of a tool that might enable better healthcare, education, and overall standard of living.”
As with all transformative technology, there will be winners and losers. The biggest winners, says Goldfarb, “will be those that figure out new systems for delivering value to their stakeholders. The biggest losers will be the incumbents in those industries that are disrupted.”
But often, he says, the industries with the most potential to benefit from AI are those hardest to change. While there is “enormous potential” for AI to transform healthcare – improving patient care and enabling healthier populations – such transformation “will encounter resistance from those who benefit from the current healthcare system”.
If we had regarded managing social interaction during the pandemic as a prediction rather than health problem, “if we knew who was infectious, then we’d keep them at home and the rest of us could go about our business. It is only the lack of information that makes a pandemic consequential.”
Taking the next step towards an intelligently automated business might sound like a technology challenge, but “the AI isn’t the hard part. The challenge is figuring out how to get all the other pieces aligned in order to do something differently. A system solution will often change who has control over decision making. It is that change in power that is disruptive and often leads to resistance.”
Meanwhile the economics are relatively simple: “When the price of machine prediction falls, people do more machine prediction and less human prediction, particularly with respect to decision making.”
Extract
When an AI-driven decision is part of a system, adopting AI can necessitate an organisational redesign with a new system. One difficulty existing organisations face in creating new systems is that they have been optimised to garner high performance from existing technologies, whereas adopting AI can necessitate a change in focus. In some cases, AI drives the organisation to become more modular, while in others, it can drive it to have greater coordination among the parts. The challenge is to recognise that the current focus is the problem, and widespread change is needed.
When top management understands that a new organisational design is needed in order to adopt and integrate an AI prediction to one or more key decision areas, a further challenge arises. This is because organisational design invariably involves a change in the value and, hence, power of the suppliers of different resources within the organisation. Those who expect to lose in the expected reallocation of power will resist change. Organisations rarely operate as a textbook dictatorship where what the CEO says goes and change happens. Instead, those expecting to have their power diminished resist change. In the process, they can undertake actions that at best reduce the ease by which change can be implemented. At worst, the anticipation of those actions may cause an organisational redesign to be curtailed completely or reversed.
Consider the example of Blockbuster Video. Blockbuster was the market leader in video-tape rentals throughout the 1990s and 2000s. The commonly known narrative of Blockbuster’s demise is that it was felled by Netflix and the rise of on-demand video. But in fact, Blockbuster did not passively succumb to the new ways. It understood what was coming but ultimately failed to adjust to it.
Edited extract from Power and Prediction by Ajay Agrawal, Joshua Gans and Avi Goldfarb, reproduced with permission.
Power and Prediction by Ajay Agrawal, Joshua Gans and Avi Goldfarb is published by Harvard Business Review Press, £22.00.
Nick Smithhttps://eandt.theiet.org/rss
https://eandt.theiet.org/content/articles/2023/10/the-ai-revolution-may-take-longer-than-you-think/
Powered by WPeMatico
