Olivier Ezratty shot again! After a guide to startups and a report on the CES in Las Vegas, he publishes a very interesting document around the use cases for artificial intelligence!
This document addresses 8 major sections:
- Main streams of AI
- Building blocks of AI
- AI and IT infrastructure
- Generic applications
- Business Applications
- Actors of AI
- AI in the society
- The AI and the company
It is a vast subject on which Olivier provides an extensive effort of more than 360 pages.
For once, and I prefer to say it right now, I did not read everything and I will not make a full synthesis here!
However, I invite you to read the section on the fundamental building blocks of artificial intelligence. It is sometimes a bit austere, or complex for newcomers. It is nonetheless very well documented and repositioned a number of things in their place.
He talks about:
- brute force and decision trees,
- Statistical methods,
- Expert Systems,
- Machine learning,
- Neural networks,
- Deep learning,
- Artificial General Intelligence.
Explanations are useful, especially on supervised and unsupervised learning!
By the way, it will surely make me review and specify my categorization of technologies for this blog and for the panorama Insurtechs .
Key takeaways from the use cases for artificial intelligence
In the use cases for artificial intelligence, where we can be demanding towards Olivier Ezratty, it is not only to redo a history and a description of the technology, but to dwell on the use cases ! On this topic, the subject is treated and well treated, by typology of actions first, then by business line, with a section for insurance.
Here is what I retain.
Typology of uses
The technologies in presence (cf face.com in Israel used by Facebook) are able to identify with 97,25% of success the person presents on a photo. For the record, the human reaches 97.5%.
Such power is now used for unlocking iPhones. It can also be used for the categorization of snapshots for example, or for medical imaging interpretation and the identification of diseases.
The use cases also apply to driving obviously, with the ability to recognize signs or the presence of elements on the road (other vehicle) or around (pedestrians).
Finally, text recognition (OCR) or the description of images in text also open up new perspectives.
In recognition of the voice and more generally the language, one allows any type of action in” speech-to-text”, but also why not in” speech-to-speech”, with possibly a stage of translation in the middle …
Concrete applications have recently arrived in our phones and homes (Siri, Alexa, etc.), with sometimes surprising results, even in ambient noise.
The Human Machine Interface (HMI) is therefore reduced to its minimum portion. The next step will be a direct connection to the brain.
Robotics, HR and Cybersecurity
In the wake of chatbots, a chapter is dedicated to robots, but that will interest us less in the case of insurance. However, this is to be read for the interested parties, especially on the aspects of improving the physical reception agency.
The HR aspect is seen solely in terms of recruitment, and not the necessary evolution of trades.
Finally, cybersecurity comes back to the use of automatic detection of spam, phishing or other attacks. However, another dimension is open, with the new risks opened up by artificial intelligence. For example, when you know how they work, it is possible to mislead artificial intelligence systems and incorporating errors or bypasses into their training.
The overview is very wide.
I only dwell here on the section on insurance, but I invite you to take a look at
- transport: section that deals in detail with the question of the autonomous vehicle, and its techniques. This can be used by any insurer to help him better anticipate his risk for tomorrow.
- health: the issues of connected health, e-health, medical research, etc. are huge. More specifically, we will make the link with the insurance sector with connected health objects (wearables in particular) that provide extremely rich information to feed real-time risk analyzes for insurers.
The 2 pages dedicated to this sector (217 and 218) are full of examples of the technologies mentioned above that it would be useless to summarize, because it is already very synthetic and gives a very real vision this time of the solutions existing, from fraud detection to automatic claims settlement, with of course a description of IBM Watson.
Conclusion of use cases for artificial intelligence
To conclude, I really invite you to review this document, not to read it in detail, but rather to have it as a reference under the elbow in case of question or doubt on the subject. It’s probably not perfect, but it’s complete enough to give examples of all-round use of artificial intelligence.
A nuance to bring however. This type of technology evolves very quickly and I fear that, if there is no regular update, this type of document is quickly outdated.