AI Economics 101
The goal of this article is to discuss the amazing value creation potential of artificial intelligence, using healthcare as an example.
What is money?
Money is defined as something that can be exchanged and which represents value. The more value you create the more money you accumulate. The rate of growth of your money creation is dependent on your productivity.
Why were coins or currency created? It represented means to exchange for an asset (e.g. some item) or, for intellectual or manual work done by another person. For more work, more coins are created to represent the value exchanged for the deeds.
One thing is for sure: Our planet’s assets are fixed. Therefore, the global wealth of our earth’s resources is fixed. However, there is a way to create more value through manual or intellectual work.
The boom of economy through industrial revolution, was an explosion in wealth due to increase in manual work capacity from machines that were fueled with steam and eventually electricity. A century later came the invention of internet, the neural network of global information. Internet caused the increase in information flow, which led to high capacity to perform intellectual work thus more value creation.
Increasing humanity’s capacity to perform manual or intellectual work leads to wealth creation, increasing our GDP.
We live in a very unique time in history (now is the greatest time to be alive) that we can engineer intelligence to augment our potential to create intellectual output. This has led to Andrew Ng naming Artificial Intelligence (AI) as the new electricity.
AI economics in Healthcare:
A vital duty of our providers is to triage. Triaging is when a trained medical personnel examines a patient and assigns a risk to this patient for hospitalisation so that high-risk patients can be prioritized.
Few years ago, Alex developed an AI application that analyzed various parameters of a patient’s case and automatically triage patients. After a year’s worth of effort with a team of 3 and numerous trials and errors, the team finally managed to create an AI driven application that cleared the “it makes sense” test from a specialist doctor.
To prove AI’s value, a mini “Medical Turing Test” was conducted. Accuracy and performance was measured in a live test in which, the AI competed against 6 specialist doctors. A third party doctor, created 30 cases and then all the 6 doctors vs. AI application triaged those cases. The expectation was that if AI was close to the performance of specialists within 10% margin of error (you need 15 years of training to become specialist doctor) it would suffice.
The “Medical Turing Test” was successful and outperformed 5 out of the 6 specialists in this niche task. That was totally unexpected!
The best part of this AI application was it’s favourable economics.
The 6 doctors that triaged the 30 cases needed 1 hour each, to triage them and were paid $250/hour.
Whereas, the cost of the AI triaging the 30 cases is less than $0.00001 (it was living in a t2.large instance of Amazon web services) and it took a millisecond.
AI triaging 30 cases cost less than $0.00001 and it took a millisecond. This was shocking!
The only reason that we do not triage the whole US population every day is that we would need 10,630,000 of human doctor hours, which equals to, for a normal work day of 12 hours, 885,833 doctors triaging at the same time.
The total cost of a “Population Triaging Manhattan Project” by doctors would be $2,65B/day.
How much would it cost for an AI application to do the same work? It would triage 30,000 medical cases per second with a cost of $0.01. At this rate, it could triage the whole US population in 88.5h. How about firing four instances of it, and triage the whole US population in 24h?
The complete US population could be potentially AI triaged with the total cost of just $9.20/day.
How about spending $1000 / day to triage the whole population every 14 minutes? We would spend $1000/day and create the today’s equivalent of $265B. Imagine the potential of such a task!
If we were to triage the whole US population in real time, we would catch many complications faster and be able to address them before it is too late. Such a scenario is beyond imagination for our current state of healthcare.
I know that all assumptions made above, are over simplified but one of the major limitations of our current value creation is the potential to use highly trained specialists in niche tasks. This limits our capacity to deal with important problems. If we tackle this bottleneck with AI, a new world of capabilities opens in front of us!
Intelligence — artificial or human augmented — is the ultimate wealth creation machine that can outperform any other investment in value creation and it scales both in size and time!
What do you think? We strongly favor the abundance scenario for our AI future!