HOW IT ALL COMES TOGETHER
Automation and Artificial Intelligence
That look like an android… The most important AI contributions do not have a tangible physical presence: the automatic pilots of commercial airplanes move millions of people around the world every day … These AI systems have been learning and maturing for years. Which is why we trust them to transfer that idea to manufacturing (autonomous vehicles that handle goods valued in millions of dollars) and other means of transport (autonomous trains). Also, AI systems collaborate in life-or-death surgical interventions that multiply the chances of survival.
In a public lighting system that optimizes energy consumption while illuminating those areas that need it; in an algorithm that processes millions of stock market transactions per minute; and in an expert system that prescribes personalized medical treatments by crossing the patient’s historical data with current symptoms. There is also AI in our word processor, which learns how we write and offers alternative words (or directly writes them) when we make a typographical error.
Why not both? The objective is to improve human intelligence with AI, not to have human intelligence replaced by AI. Industry invests a lot in AI to replace human beings with machines. What for? To automate repetitive tasks, minimize errors, improve accuracy, increase the speed of execution of processes, etc. But all that does not mean “making human intervention unnecessary“. The primary objective when automating a task or process is not to eliminate human intervention, but to eliminate the inconveniences that human intervention entails.
It has become a fundamental tool in the context of IoT (or we should say IoE) and Industry 4.0: we have a population of robots working autonomously and generating data on a large scale. And we want them to incorporate AI, so they can make the right decisions for themselves: where to move, at what speed, with what force to tighten a screw or move a package…
What if we use the same data to make the same decisions, but without needing to physically transport and store that data anywhere else? After all, many of these data are no longer necessary as soon as the decision is made (move an arm, weld, push a package, move on…), so the robot can remove those data right after the decision has been made.
The data that a robot needs to make its decisions is processed at the robot itself, without moving it elsewhere. This quickens the response time, simplifies communications and turns each robot into an entity with its own autonomous intelligence. There is still massive data transmission in and out of a central entity to coordinate different robots, orchestrate an entire assembly line or an industrial logistics warehouse.
2019 will be “the year of Edge Computing“: Ericsson, Akamai, Limelight and Fastly are introducing it, and discovering that it is cheaper and easier to manage than centralized cloud computing. So not everything depends on a large data lake and a central computing center.
- A lot of medical tests and diagnoses are based on images:
- X-rays, mammograms, Doppler studies, etc…
- So, it makes sense to use thousands of these images to train algorithms so they can identify tumors or malfunctions.
- Also, algorithms can compare symptoms and evolution of different patients, as well as analyze in great detail the same patient over time.
- In addition to diagnosis, Deep Learning may be used for prevention:
- Example: Alphabet (an American Medical Research company) uses predictive software to analyzee images of the retina and detect risk factors for future diseases.
- They have a key advantage for medicine:
- Machine Learning algorithms based on them are not invasive…
- Doctors or algorithms can try and learn (and make mistakes) as many times as necessary with the Digital Twins, without any danger for patients.
- The body of the patient will not be invaded for a doctor to “see” the state of tumors on a screen.
- It will be possible (in the long term) to obtain the exact Digital Twin of each person:
- The advantages this will bring are not yet clear:
- The hyper-personalization of medical treatments
- The possibility of robots performing pre-programmed surgical interventions tailored to each patient.
- This is not science fiction, as it is already done:
- Printing in 3D prothesis and valves for hyper-personalized blood circulation.
- Designing exoskeletons for paraplegic people.
- The advantages this will bring are not yet clear:
- Many companies are investing in AI for the health industry.
- To understand why, we must analyze it coldly:
- Medical treatments constitute an industry that moves huge amounts of money every year, both in public and private health services.
- For private health insurance companies, the more profitable investments in the long term are those aimed at preventing their clients from falling ill, avoiding expensive treatments.
- There are synergies with areas not previously related, such as private vehicle insurance companies:
- Treatments and health care caused by accidents involve astronomical costs for both insurers and medical companies (among others).
- So, they work together in road safety programs, driving support systems, smart roads…
- Thus, more and more synergies appear with other types of industries.
- Example: IBM and Pfizer (an American biomedic research company) use sensors to monitor patients of Parkinson’s disease:
- Making real-time Analytics of patient functions (to gauge their risk when they are on their own).
- It is also applied to patients with cancer or with neurological diseases.