Throughout the last few years, I had the opportunity to be at the forefront of witnessing how transformational the application of AI is to different businesses. The ability to automatically extract insights from data to make better, smarter and more informed decisions is no longer a “nice to have,” it is one of the most essential ways to successfully compete in today’s marketplace.
Already, multiple business functions have realized the benefits of using AI:
- Increased employee proficiency in the digital workspaces
- Enhanced self service capabilities on communities through machine based recommendations
- And more recently with the e-commerce rush, AI has helped deliver higher conversion rates through personalized online consumer experiences.
We have numerous examples illustrating the ROI of applied AI, and there is ample literature about this topic. But in my opinion, the broader economic impact of AI adoption merits a deeper evaluation.
Simple question: What does AI reduce the cost of? AI profoundly reduces the cost of Prediction
I recently came across the book Prediction Machines: The Simple Economics of Artificial Intelligence co-authored by professors Joshua Gans, Avi Goldfarb, Ajay Agrawal. They argue that AI serves a transformative, economic purpose: it significantly lowers the cost of prediction.
Because of this, I believe AI is fundamentally more important relative to other technology innovations.
A parallel technological innovation discussed in the book that helps equate the possible impact on the cost reduction of a useful outcome is semiconductors. The book illustrates how semiconductors democratized arithmetic. Semiconductors drastically reduced the cost of calculating now elementary statistics such as averages and medians.
And then 3 things happened:
- The first obvious consequence: cheaper cost increased the availability and adoption of the technology. Current industries that were already using arithmetics expanded their usage into other business functions such as inventory management, accounting, and cost forecasting. It was easier and cheaper to run a software program, and it greatly reduced human errors.
- The second consequence is a little less intuitive, but has a profound impact on innovation and ultimately on society. When the cost of arithmetic went down, it became compelling for other non traditional users to contemplate using it. As a result, arithmetic started being used to solve problems that hadn’t traditionally been framed as arithmetic problems. The book discussed how we used calculations to solve for the creation of photographic images by employing chemistry (film-based photography). Then, as arithmetic became cheaper, we began using arithmetic-based solutions in the design of cameras and image reproduction (digital cameras).
- The third thing that happened was that it changed the value of other things—the value of arithmetic’s complements went up and the value of its substitutes went down. Going back to the photography, the complements then were the software and hardware used in digital cameras. The value of these increased because we used more of them, while the value of substitutes, the components of film-based cameras, went down because we started using less and less of them.
The democratization of prediction through cheaper AI is likely to have similar impacts.
My experience working with diverse companies and partners has demonstrated already profound impacts both in the corporate world but also in the society at large.
- The first consequence: Increased adoption of AI across a myriad of use cases: I firmly believe that it is now an “AI take all economy”. Even the most traditional industries are forced to reinvent themselves as they are facing new competition (i.e. Marriott vs. AirBnB, Hertz vs. Uber). Companies cannot rely on a simple website or an app to check the digital box. Digital is a given. Companies now need to deliver the most relevant experience to each unique buyer – and the only way to do this at scale is through AI.
Traditional retailers have been hit really hard in 2020. Local pet food retailers are now competing against e-commerce giants such as Cheewie. Cheewie delivers faster, cheaper and targeted product recommendations through AI.
Every partner I am working with is expanding their capabilities to cater to this new environment; even within the most conservative industries such as banking, government and insurance.
- The second consequence to the democratization of AI – similarly to what we experienced with arithmetic – is how we are using predictions to solve problems that have not historically been thought of as prediction problems.
The medical field, more specifically radiologists, is a great first example. Radiologists traditionally relied on their experience and judgement to interpret scans. Technology then focused on improving the quality of the image for better decisions. Reducing the risk of mistakes and facilitating decision making are prediction problems. Using AI methods, more specifically extracting insights from data through machine learning, doctors can now automatically recognize complex patterns in imaging data and provide quantitative, rather than qualitative, assessments of radiographic.
Another example is around programming. Traditionally, programming was built using logic in a controlled environment: “If this then that”. AI has allowed programs to use prediction to respond more effectively to uncontrolled environments. Solving problems such as autonomous driving is well documented. Traditional programs would have tried to code all options manually: If a human crosses the street then stop. If the light is yellow then slowdown. But no matter how many lines of code you build you can never fully capture the whole picture.
We never thought of autonomous driving as a prediction problem but it is the only way to solve it. The AI needs to predict the answer to the question: What would a good human driver do? Then ingest the data and learn from it.
In every scenario, the AI makes a lot of mistakes at first. But it learns from its mistakes and updates its model every time. Its predictions start getting better and better until it becomes so good at predicting what a human would do that we don’t need the human to do it anymore. The AI can perform the action itself.
We could not do this until the cost of predictions was significantly low
- Finally, as the cost of prediction will continue to drop, just like demonstrated with arithmetic, there will be complementary products that will see their value go up and substitute products that will go down. The main complement to machine based prediction is human based one.
A few months ago, GPT-3 an autoregressive language model that uses deep learning to produce human-like text, the third-generation language prediction model in the GPT-n series created by OpenAI has already demonstrated that GPT-3 can take normal speech, and automatically turn it into ‘lawyer speak’. It can also take written text, and turn it into fully functioning code, in real-time. It can even write poetry! GPT-3 gives but a glimpse into the power we’re soon to unleash.
There are a lot of ethical aspects around this element that fuels a lot of passionate debates.
- Is my job safe or will a robot replace it?
- Innovation is happening so fast, what will we do as a society to support the ones that have obsolete skills and can’t reinvent themselves?
- Is our education system built in such a way that the next generation will be properly trained to strive in this new environment?
All very valid questions. Thankfully, organisations like @OpenAI (who built the GPT-3 model) are pathing the way to safe & sustainable AI. Employing the world’s greatest minds to ensure that AI remains a force for good.
Finally, by understanding the impacts of an increased adoption of AI, I am hoping it will make it easier for business leaders and public servants to figure out the most valuable ways to apply AI for the benefit of all stakeholders.
Got a comment, a question – I’d love to hear from you!
Marie
Marie-Michele Caron 