AI will save the world

Daring? Absolutely, but behind this futuristic vision lies a tangible reality. While the title might appear provocative, by exploring the potential of Green AI, we can uncover how artificial intelligence is indeed playing a significant role in the pursuit of sustainable solutions to address global environmental challenges.
However, caution is warranted; it's essential to view this reality as a two-faced Janus entity. In fact, AI can - at the same time - both save and endanger the environment we inhabit.

From a positive point of view, AI can play (and is already doing so!) a vital role in searching for solutions to support sustainability and the environment. For example, by leveraging data collected from machinery and IoT devices, it is possible to reduce energy consumption or material waste. In the context of smart cities, historical mobility and behavior prediction data provide a fundamental contribution to making our cities more livable.

The use of AI, however, can pose **risks** to the environment. For instance, the implementation of AI infrastructure requires a significant amount of energy and resources, especially for data processing in large data centers. Additionally, widespread use of AI could lead to an increase in the production and disposal of electronic devices, contributing to the rise in electronic waste.

- **High Energy Consumption** in Data Centers: training and running artificial intelligence models require a significant amount of energy, especially in data centers. This can have a substantial impact on the environment, contributing to carbon emissions and the use of non-renewable energy resources. It is important to explore solutions to reduce energy consumption, such as model optimization, the adoption of energy-efficient hardware, and the use of renewable energy sources to power data centers.
- Training machine learning models requires a **vast amount of data**, and maintaining a historical record of this data over time can consume a significant amount of resources, including storage space, energy, and processing capacity. This can have an environmental impact, as data centers hosting these resources consume energy and generate waste. In the development of new models and algorithms, it is important to question which and how much information is actually valuable: when it comes to data, it's easy to accumulate too much of low-added-value data.
- **Electronic Waste Generation**: artificial intelligence relies on sophisticated hardware, such as servers, storage devices, and sensors, which can contribute to the growing production of electronic waste. It is essential to explore strategies to address this issue, such as responsible recycling of electronic components, extending the useful life of devices, and adopting modular design to facilitate the replacement and upgrading of obsolete components.
- Use of **Unsustainable Materials**: Some components of the hardware used in artificial intelligence may require the use of expensive, rare, or unsustainable materials. This raises concerns about resource availability and the environmental impact associated with the extraction and production of such materials.

Are there **solutions** to these problems?

Green AI focuses on creating energy-efficient machine learning models, optimizing data collection processes, and utilizing adaptation and model evolution techniques that are environmentally responsible in terms of resource management.

The traditional lifecycle of a machine learning algorithm involves starting from scratch whenever you want to update the model or train it for a new task. Typically, in this process, you either expand the training dataset with new data or modify the model's structure, allowing it to evolve over time. This process presents two major challenges. On one hand, it's necessary to maintain a memory of all the data over time, leading to resource usage and privacy risks. On the other hand, starting the training process from scratch consumes a significant amount of energy, especially for more advanced models.

Large technology companies like Google and Microsoft are working to optimize the energy efficiency of their data centers, which house the servers used for training and executing artificial intelligence models. This includes using more efficient cooling systems, optimizing infrastructure, and adopting low-energy consumption technologies.

Some companies are developing specialized hardware for artificial intelligence, such as high-performance graphics processing units (GPUs) or artificial intelligence accelerator processing units. These devices are designed to perform AI operations more energy-efficiently compared to general-purpose computing solutions.

The most important aspect, and potentially of greater impact, is to directly intervene in the development of innovative algorithms and training paradigms that require fewer computational resources. In this regard, the most promising methodologies belong to the so-called Continual Learning. These techniques allow models to evolve over time without starting from scratch each time, but by retaining memory of what has already been learned, thus avoiding the need for lengthy retraining or maintaining overly deep data history.

In summary, awareness of the environmental issues caused by AI urges us to act responsibly, promoting the research and development of Green AI. As Luciano Floridi states, "Green and blue are two colors that will save the world." This phrase emphasizes the importance of combining the environment (represented by green) and digital technologies (represented by blue) to address the challenges of the 21st century. Green AI, which merges artificial intelligence and environmental sustainability, will be crucial for humanity. By harnessing the potential of digital technologies in a responsible and sustainable manner, we can work towards a future where AI is an ally in solving environmental problems and improving people's lives.

Only if it becomes green

Pitch Your Innovation Industry 4.0 Edition

We were in Bologna for the event "Pitch Your Innovation Industry 4.0 Edition", organized by the Startup Community of CRIF and G-Factor, the incubator-accelerator of Fondazione Golinelli.

We had the pleasure of introducing Deix S.r.l. to the attending companies and getting to know other startups that are about to take off.

Deix becomes a member of Assintel

This month Deix has become a member of Assintel, the national association for ICT and Digital companies of Confcommercio.

The association will allow us to expand our network, connect with new realities, and have our say on the role of digital companies in the Italian landscape.

Demo Day in TERNA

Today our own Jacopo Corno is at TERNAIDEASpark by Terna SpA in Turin for the Demo Day of the Data Science for Resilience call!

We have been selected as one of the top 10 finalist startups for our proposal on asset management and optimization of maintenance interventions on the Terna grid. 🏅

Our ability to combine data analysis and prediction algorithms with optimization models has been particularly appreciated, enabling us to provide prescriptive suggestions to define the best strategies. 🚀

Deix among the 10 finalist startups

AI Week 2023

On April 20th and 21st, we were at Rimini for the fourth edition of AI Week.

Two days of presentations, workshops, and discussions on the hottest topics revolving around the world of Artificial Intelligence. We also had the opportunity to meet interesting companies with whom we can establish collaborations.

On our YouTube channel, you can find the video of our Digital intervention in collaboration with Prysmian Group.

Privacy is dead. Or is it?

No, privacy is not dead, it has simply changed. And as a result, our approach to its protection must also change.

At the beginning of 2021, Amazon installed AI-powered cameras in the delivery vans of one of its warehouses in Los Angeles. Derek, a delivery driver at the facility, shared that the camera in his van started incorrectly penalizing him whenever other cars cut him off, which is a daily occurrence in Los Angeles traffic. "Maintain a safe distance", said the camera installed above his seat when a car cut him off. The data was then sent to Amazon and used to evaluate his performance for that week and determine whether he would receive a bonus.
The incident initially caused panic for Derek and his colleagues, and subsequently for the entire sector of employed workers who feared for their privacy.

What Derek is unaware of is that he himself is also a surveillant in turn. Whether we like it or not, all of us play a role in "surveillance." Sociologist David Lyon, in his latest book "The Culture of Surveillance," no longer approaches the subject with the notions of "state" or "society," but rather speaks of culture. These outdated notions described a type of top-to-bottom surveillance, the same one depicted by George Orwell in his 1984.

The culture of surveillance in which our society finds itself goes beyond that. While acknowledging what happens in the surveillance carried out by organizations, it highlights the different roles that each of us plays in relation to it. For example, the data we create using Facebook (or any other social media) is not only exploited by organizations to sell us to the highest bidder but is also the tool we ourselves use to surveil, de facto, what our friends say and do. We self-surveil.

And it is in this last analysis that Lyon finds one of the core issues: performance. In a world where we are constantly surveilled, everything becomes a performance. And this inevitably has side effects. The awareness of potentially being observed alters our behavior: it makes us harmless, conforming to the behavior we know to be the most reassuring, it denaturalizes us, leading to what he defines as the "collapse of contexts," where, returning to the example of social media, too many different contexts of our lives merge and collapse into a single amalgamation.

Professor Antonio Casilli explains that what privacy has undergone is not an end but a transformation. It has shifted from an individual right to a collective negotiation. In his book, this is presented as a transition from "privacy as penetration" to "privacy as negotiation," which is a notion inherited from Anglo-Saxon jurisprudence of the 19th century that portrays each individual as an isolated entity at the center of a private sphere. American jurist Louis Brandeis stated in 1890 that "privacy is the right to be left alone," but online, no one wants to be left alone. Moreover, what we face today is a system where we consciously sell "ourselves," as seen in the startup Weople: an app founded by the company Hoda that promises its subscribers compensation in exchange for the surrender of their personal data.

With this premise, Italian writer Giovanni Cattabriga, better known by the pseudonym Wu Ming 2, reminded us back in 2008 of the essential formula to use in this fight against the culprit: "the usual old speech. Since the enemy uses a tool, the tool becomes the enemy." Hence, the exhausting realization that regardless of the means, the owner of a company will always have an advantage in the Hegelian master-slave relationship. But should this be enough to limit technological development?

In industrial workplaces, a thorough risk assessment is the first step towards increased safety, using tools that allow the identification of risks before they can lead to potentially dangerous situations. When incidents occur, they should be analyzed and used together with all other appropriate preventive measures to avoid future occurrences.

Collecting real-time data, in fact, enables us to broaden our understanding of the daily issues that workers have to deal with and, consequently, make changes to support the creation of a safer work environment.

Furthermore, "the surveillance culture," like any other culture, develops in different ways and often transforms unpredictably, especially in contexts of increasing social fluidity. It sprouts and blossoms differently depending on historical and political circumstances.

Therefore, before succumbing to psychological terrorism dictated by the fear of being devoid of protection, passive victims of a technocratic system, it is necessary to analyze the issue based on the involved countries and their contribution to citizen security.

In this regard, the European Union has always been at the forefront. Since 2016, through the General Data Protection Regulation (GDPR), it has promoted the protection of personal data and sought to restore control to citizens over their own privacy.

In conclusion, although privacy has undergone a transformation, and the way we conceive it has changed, we cannot claim that it has died completely. It is true that we live in an era where surveillance is widespread, and our digital and physical lives are constantly monitored. However, we must not succumb to fatalism. Technology offers new scenarios that can support and improve our lives. It depends on how we use it. The responsibility to use technology for the benefit of others and not to their detriment remains in our hands.

Or is it?

Privacy is dead

Deix at Ideas to Impact

On March 10th, Deix participated in the demo day of Ideas to Impact, an initiative by Open Italy.

We were selected among the top 10 most interesting startups in the Safety Management and Digital Health sector. 💪

We presented Varda-AI, our platform to enhance workplace safety through computer vision.

Selected among the top 10 most interesting startups in the safety sector

Interview on Story Time

We were on Story Time on the Radio Italia channel to talk about Deix, what we do, and our perspectives for the future.

On [our YouTube channel](https://www.youtube.com/watch?v=0-juElc8G4o&t=15s), you can find the video for more in-depth information!


Articles and news