Home / Blog / Megatrends for this decade – VII

Megatrends for this decade – VII

The chip developed by Imec and Global Foundries using analog techniques for a low power machine learning engine for edge AI. Image credit: Imec

6. Everything is smart, embedded intelligence

This electric grill for steaks caught my attention the other day. The packaging and the shop label claimed this was an “intelligent” steak grill. You just place the steak and it would decide how to best cook it! Photos taken by me at Kasanova shop in Turin

This Megatrend may appear “outdated”. If you look around with a lay-person’s eye you’ll see plenty of ads claiming “intelligence” of products, most of the time in a ridiculous way. The first (that I know of) “intelligent” toothbrush debuted at CES 2017, now you can find quite a few intelligent toothbrush on Amaszon starting at 24.99$.

On the left a photo I took a few days ago in a shop in Turin, selling an “intelligent” steak grill. The box pointed out that the grill was “intelligente”, you just place the steak on it and it would decide the cooking temperature and the time needed. I guess it had a sensor to detect the temperature inside the steak and based on your preference (raw, medium or well done) it knew when to stop the grilling. A chip provided the algorithm that I suppose would start with a high temp to broil the surface and them a lower temperature for cooling until the inside of the steak reached a certain, predetermined temperature.

Now this is fine, and probably result in a grill that can cook a steak better than what I would be able to do, but calling it intelligent is quite a stretch.

It is just an example out of many you can see for yourself just looking around. Sometimes an object is defined “smart”. In a way it is an intelligence restricted to a very specific task, and it is usually much more appropriate than claiming for intelligence. For sure, we have seen many more products being defined as smart and the words “artificial intelligence” has grown to the point of becoming a ubiquitous presence to the point that we are now seeing variations on the theme.

The Hype Cycle for emerging technologies is out and it is interesting to look at what changed since last year… Image credit: Gartner

Gartner in its 2020 emerging technologies hype-cycle identifies :

  • Generative Adversarial Networks (a tech for self-learning by a machine)
  • Self-supervised learning
  • Adaptive Machine Learning
  • Composite Artificia lntelligence
  • Generative Artificial Intelligence
  • Responsible Artificial Intelligence
  • AI augmented development
  • Embedded Artificial Intelligence
  • Explainable Artificial Intelligence
  • AI assisted design

This list of 10 emerging topics related to AI is out of a total of 30, one third of the predicted emerging technologies are in the AI field.

So, on the one hand we have a pervasive perception of AI presence in our world today, on the other hand we see more and more research on different aspects of AI (as in the Gartner list) that will result in an expansion of capability and of application areas in this decade. There is more.

On November 18th Halide, a company developing software for computational photography, published a review on the iPhone Pro Max 12 photo capabilities.

In that review they discovered an amazing, although hidden feature of this phone. As you may know, the iPhone Pro Max 12 has three lenses, i.e three cameras/three sensors. The wide angle lens/camera has the biggest photo sites  (the buckets gathering the incoming photons) 1.7µm versus the 1.4µm of the other two cameras (yes I know 1.7µm may not seem something to call “big” but it is 20% bigger than the others, meaning it can harvest 20% more incoming light and that makes a difference when you are dealing with very low light, as in night time photography.

Here comes the amazing part. When you take a photo in low light condition and select the telephoto camera you’ll see, as you would expect, in the iPhone screen the “zoomed in” image. However, the phone will not use the telephoto camera since the software has detected a low light situation and switched to the wide angle camera. Then, in real time it crops the image frame to match the frame size you will get if it was using the telephoto lens. Well, you might say: “this is a trick that really does not require any intelligence!” and you would be right expect that you are not!

When you take a wide angle photo and then a telephoto you get a zoom in effect that can be obtained by cropping the photo taken with the wide angle lens but that cropped image will not be like the one taken by a telephoto lens: the perspective changes (using a tele depth distances are squeezed) and the bokeh is dramatically different so that our eyes (brain) would spot the difference. Not so with this iPhone. The software (computational photography) will look at the scene, “understand” the image and reshape it to match the result that would have been produced by a telephoto lens. Now, this is what I call “intelligence”! And yet, it is not advertised as such (I guess for marketing reason: Apple might consider better to hide this little cheating, letting you feel that you have control on the selection of the camera…).

Hence we are seeing:

  • hype (often just pure hype and no substance),
  • ongoing research effort in bettering AI and extending its application, and
  • AI becoming a tool in a growing number of industries, both as part of a product and as part of industrial processes. although we may not be perceiving it.

These latter two are obviously the most important and are the ones that are sustaining this Megatrend “everything is smart, embedded intelligence”.

A significant evolution that is happening in these last years, and that will be in full swing in this decade, is the capability to embed intelligence (at least a little bit) inside more and more objects leveraging on the evolution of electronics “designed” to support ML. An example is given in the opening figure: a chip developed by Imec and Global Foundries using analog techniques for a low power machine learning engine for edge AI. Here the crucial point (taking for given the lower and lower cost) is the low power demand, allowing its embedding in a variety of products. Notice that the chip on its own would not be enough: it needs to be fed by data and these are provided by embedded sensors -IoT-, an integral part of the Industry 4.0 evolution AND it needs to become part of a network of local intelligence leading to an emerging “ambient intelligence”.

This latter is what is probably going to characterise the next decade (2030-2039) with 6G playing a major role.

Digital Twins may also play a significant role in this distributed smartness and embedded intelligence, creating a bridge between an object and the cyberspace and (stage IV) contributing to the object functionality (and behaviour). A Digital Twin can embed a local (object) intelligence and can be part of a shared/emerging intelligence.

About Roberto Saracco

Roberto Saracco fell in love with technology and its implications long time ago. His background is in math and computer science. Until April 2017 he led the EIT Digital Italian Node and then was head of the Industrial Doctoral School of EIT Digital up to September 2018. Previously, up to December 2011 he was the Director of the Telecom Italia Future Centre in Venice, looking at the interplay of technology evolution, economics and society. At the turn of the century he led a World Bank-Infodev project to stimulate entrepreneurship in Latin America. He is a senior member of IEEE where he leads the Industry Advisory Board within the Future Directions Committee and co-chairs the Digital Reality Initiative. He teaches a Master course on Technology Forecasting and Market impact at the University of Trento. He has published over 100 papers in journals and magazines and 14 books.