According to French futurologist Laurent Alexandre “Today, Europe does not control any of the components of AI, whose industrialisation is based on the marriage between the power of computers, the mountains of big data and the neural networks for deep learning”. It is true. Remember that without large amounts of data, the machine learning algorithms, even the most efficient, will be of no use.
Across seven of its products, Google has more than 1 billion unique users: Gmail, Chrome, Maps, Search, Youtube, Google Play and Android (more than 2 billion platforms equipped). On Facebook, there are 3 products that exceed this billion users: Facebook, Whatsapp and Messenger. They have widely enough data to provide the best targeted advertisements.
The battle on personal data is lost but not on physical and industrial data.
More than 70 million km of power lines installed in the world, More than 2 million wind turbines installed in 2050, more than 50 Ha of photovoltaic panels installed every day. The examples of substantial volume of industrial assets are not lacking. Industrialists have taken the step of digitisation but the transformations are in their infancy.
Guess that if only 5% of the french electrical distribution grid was captured, more than 4.5 million pictures could be generated
Reaching 100% data coverage of the grid would be useful to better monitor it. There are 2 main reasons why it is not done that way today:
- Picture analysis are mostly done manually. Who would like to check more than 4.5M pictures?
- The combination of drone, AI and sensors are just now reaching critical maturity levels
If drone inspection market is valued at more than $ 127 billion by PWC it is mainly because physical world will definitely get closer from the virtual world. In a few years there will be an other google map much more accurate in time and space with many different information : water temperature, pollution in the air, corrosion levels on electrical transmission towers, health levels of plants, and so on.
The GAFA or BATX do not have a clear competitive advantage to take a leadership role in physical world data. Here are a few reasons:
1) More than 200 applications can be affected by drones. This implies inspection markets that are very heterogeneous in terms of size and applications. The task of taking into account the industrial environment and the operational values brought is to be carried out here. Facebook and Google have only one market today, that of advertising.
2) Data on these markets do not necessarily exist today and manufacturers are still working to establish acquisition standards (equipment and methods) with experts in data capture automation like Sterblue.
3) The expertise of the industrialists is necessary in order to build different weak AI’s specific to each industrial application. As an example on distribution grids, more than 130 types of faults are to be appraised with the necessary development of several detection algorithms.
4) The data is considered to be more sensitive, hence additional precautions to be taken with respect to existing platforms.
So far Sterblue has been working on these different application with many leading utility companies.