$ viu feature_img.png

AI a catalyst for innovation

$ whoami

Virginie Marelli

$ date

2022-09-26

AI a catalyst for innovation

With the Energy crisis, it is even more blatant that we need to work on reducing our carbon footprint and better use the planet's resources.

Earlier this year, I wrote an article about the impact that AI has on energy and how much smart implementation of algorithms we need. I discussed the race of big models and the negative impact they can have on energy consumption and thus carbon emissions. I also listed a couple of solutions from the software and hardware perspective that are active area of research.

Is AI an eco disaster? - Virginie Marelli
You hear more and more that technology in general is not so eco friendly. Whatabout AI? Is it also not so eco-friendly? What is the impact of developing AImodels and how good is AI for the planet? With this article, we try to demystify and understand the impact of AI on theplanet and how this c…

With this article, I would like to also emphasis that although AI is energy greedy, AI is also a catalyst for innovation and thus help reducing CO2 emissions and lead to more sustainability.

Innovation with AI

We've seen AI being more and more present in the recent years, conquering many industries and being applied in all types of domains. AI was first applied with 3 main goals: cost reduction, value generation and risk and fraud management. In many cases, this led to positive direct or indirect environmental impact.

AI can be used to better sort and recycle waste, Recupel a Belgian waste recycling company has successfully implemented it into its sorting lines. Microsoft, Meta and Google are all claiming to use AI to optimise energy usage and to cool their data centers. Optimisation of processes is also a very active area for AI, predictive maintenance for example allows for smoother and safer production, and usually also energy savings. Detecting defects in manufactured goods is also a process benefiting from AI. Indeed, in defect detection, the cost of an overkill (false positive) is usually lower than the cost of an under-kill (false negative). You don’t want that a wrong product is leaving the factory and reaching your customers, hence usually, strict controls are put in place, leading to waste of goods. AI, by better understanding the problem, allows to reduce overkill, thus reducing waste.

Not only is AI optimising current processes and production lines, it is also accelerating research. For example, this year, AI made some strikes in nuclear fusion, advancing the field of research and unlocking promising results (a collaboration between DeepMind and EPFL). A new enzime was designed by AI and can devour plastic in a couple of hours, which is very promising to reduce the accumulated waste on the planet.

Finally, AI can help with the environmental crisis and a lot of research is conducted in this area. For example, in Canada, they research how AI can mitigate in effects of climate change, in many areas. In Australia, they monitor real-time the great barrier reef with a model deployed on a Nvidia Jetson (the same that we used in our nowcasting project). They are then able to act proactively and constrict damage of the reef. These are only a couple of examples and there are much more applications out there.

What Dataroots does

At Dataroots, we are also very aware of the impact that our implementations and AI solutions can have on energy.

We are always taking a pragmatic approach to the problems and factor cost and energy consumption in the solutions we propose. We have also worked on some very interesting projects, aiming at reducing energy, mainly in the predictive maintenance and error reduction areas. Defect detections for manufacturing, warehouse stock forecasting and parameters optimisation are a few examples of projects we did implement for our customers which resulted in energy gains and reduced carbon footprint.

We are also actively researching elegant solution designs for energy-constrained computations (like models running on edge devices) and energy efficient communication implementation (for example between devices, like federated learning). We have tried our own quantisation and model distillation as a mechanism to reduce model size when deploying a computer vision model on a sensor, and follow up energy-constrained optimisation research.

Conclusion

AI had to initially prove its value and was applied mainly to reduce cost, generate value and manage risks and fraud. Now, we see AI used more and more in other areas, like art and data for good. AI is a powerful tool and it is undeniably a catalyst for innovation and can definitely contribute to reducing our carbon footprint.

With great power comes great responsibility - Stan Lee

So when you decide to use AI, don't forget to factor energy in your choices :)

You might also like

Weather Nowcasting - Model compression - Margaux Gérard, Omar Safwat
-------------------------------------------------------------------------------- In our previous post, we explained our project of weather nowcasting in ageneral way. Now, we will deep dive into one of the most important steps inmachine learning, which is model optimization. The need for optimiz…

Got an idea?

Let's talk!

sign up to our weekly AI & data digest ❤️