Top scientists in artificial intelligence (AI) have warned that deep learning will put radiologists out of work. Healthcare professionals have said that AI will change the way doctors and patients interact.
Tech executives have said that fully self-driving cars are just around the corner. All of these predictions have been wrong.
AI has come a long way, but it still hasn’t changed many industries in the way it could. But when you think about the steam engine, electricity, and the internal combustion engine, it’s not surprising that AI isn’t being used as quickly as those.
Ajay Agrawal, Joshua Gans, and Avi Goldfarb are professors at Toronto University and the authors of the new book Power and Prediction. They think that we are at a point where AI’s power is clear, but it hasn’t been widely used yet. And if we want to use the power of AI, we need to be able to deal with the problems that get in the way. To do this, we need to know not only how AI is used but also how it works.
Point solutions and Systems
In Power and Prediction, the authors explain AI as software that can predict things like whether a customer buys a recommended product or a financial transaction that turns out to be fake.
Machine learning (ML) models have reached the point where, with the right training data, they can make impressive predictions.
But when it comes to putting the power of prediction machines into applications and products, organizations have to deal with different levels of challenges.
“The technical progress of AI was and is very impressive. So it makes sense to think that their apps might grow at the same rate,” Agrawal, Gans, and Goldfarb told VentureBeat.
“They haven’t, and we wanted to find out why in our research. People were focusing on AI point solutions, but we ended up thinking about how to get value from AI in the systems we already have.
It was obvious that something was wrong. To really use AI, you need to be open to a wider range of actions, but many companies aren’t ready for that.
AI’s easy-to-get solutions are point solutions. These are all places where businesses already use prediction. One example the authors gave was the Canadian company Verafin, which uses AI to predict fraud.
Verafin, which is based in St. John’s, Newfoundland, became Canada’s first AI unicorn when it was bought by Nasdaq in 2020 for $2.75 billion. Analysts who made predictions about commercial AI in Canada in the past didn’t think about Verafin or St. John’s.
Verafin’s success is due to the fact that it used an important AI point solution. Predicting fraud has always been an important part of the work of financial institutions.
Replacing their old systems with an AI-powered solution that makes better predictions required minimal changes to their organizational structure.
In other areas, adopting AI means not only making changes to the technology, but also rethinking the whole system, including the product, organizational structure, company goals, alignment of incentives, and other parts of a business. This makes it much harder for businesses to use AI to its fullest capacity.
The authors of Power and Prediction say, “Our focus on what prediction machines could do had made us blind to the likelihood that they would be used in the real world.”
“We had been focusing on the economic properties of AI itself, like lowering the cost of making predictions, but we didn’t take into account how expensive it would be to build the new systems that AIs need to be a part of.”
AI adoption in the “between times”
Agrawal, Gans, and Goldfarb call the current state of artificial intelligence (AI) the “Between Times” of AI. This means that we are between showing that the technology can do what it says it can do and realizing its promise through widespread use.
This has been done before. In the 1890s, when people thought of systems from the point of view of the steam engine, the main benefit of electricity for manufacturers was that it saved them money on fuel. Electricity wasn’t just a cheaper version of a steam engine, though.
Its main use was to separate energy from where it came from. You didn’t have to put a steam engine next to your factory anymore. But most factories were built this way, and it wasn’t until the 1920s that this potential was fully realized.
By that time, new factories were built with the idea that the power generator could be miles away, and electricity could be brought to any part of the building with a cable or a power outlet.
Andrew Ng, a scientist who works on AI, has said that AI is the “new electricity.” Sundar Pichai, the CEO of Google, has said that AI is “more fundamental than electricity.” Most likely, they are right.
But in Between Times, most people are using point solutions like fraud prediction using machine learning, video transcription, image classification, etc.
“We are at the point where, if AI is going to be a game-changer, we will soon start to see the first signs of that change. Agrawal, Gans, and Goldfarb all said that it is likely to come first from new businesses that use AI to start completely new ways of doing business.
At the moment, point solutions favor those who are already in power. But history shows that organizations that have been around for a long time take a long time to make the system changes that new technological revolutions require.
“New businesses have an advantage because they don’t have to change what’s already there. The authors said, “They can start from scratch.”
“But at the same time, we can learn from history that business leaders today should be even more careful to understand AI’s potential to change things.”
What is the future of AI Adoption?
Even though incumbents and startups are still in a tug-of-war, it’s clear that AI hasn’t reached its full potential yet. And in the future, AI is likely to be used in new ways and with new systems that are very different from what we’ve seen so far.
Agrawal, Gans, and Goldfarb said, “We think there are still a lot of opportunities to be had by using AI as point solutions or applications that don’t disrupt businesses too much.”
“The real change won’t happen until AI technology has improved so much that it makes sense to think about building new systems around it. We hope that time will come, but until then, there is a lot of value to be found in the “smaller” parts of the technology.
To read more Tech Related Articles Click here.