View source for Building cross-stakeholder awareness and understanding of the direct and indirect environmental impacts of digital/Internet technologies and how to mitigate them – WS 04 2023

20 June 2023 | 16:45 - 17:45 EEST | Auditorium A1 | [[image:Icons_live_20px.png | Video recording | link=https://youtu.be/FMX6pYK3eWg]] | [[image:Icon_transcript_20px.png | Transcript | link=Building cross-stakeholder awareness and understanding of the direct and indirect environmental impacts of digital/Internet technologies and how to mitigate them – WS 04 2023#Transcript]]<br />
[[Consolidated_programme_2023#ws04_23|'''Consolidated programme 2023 overview / Workshop 4''']]<br /><br />
Proposals: [[List of proposals for EuroDIG 2023#prop_56|#56]] [[List of proposals for EuroDIG 2023#prop_57|#57]] [[List of proposals for EuroDIG 2023#prop_65|#65]]<br /><br />
== <span class="dateline">Get involved!</span> == 
You are invited to become a member of the Session Org Team by simply subscribing to the [https://list.eurodig.org/mailman/listinfo/ws4_2023 '''mailing list''']. By doing so, you agree that your name and affiliation will be published at the relevant session wiki page. Please reply to the email send to you to confirm your subscription.

Kindly note that it may take a while until the Org Team is formed and starts working.

To follow the current discussion on this topic, see the [[{{TALKPAGENAME}} | discussion]] tab on the upper left side of this page.

== Session teaser ==
Building cross-stakeholder awareness and understanding of the direct and indirect environmental impacts of digital/Internet technologies and how to mitigate them

== Session description == 
Internet-connected technologies are an integral part of our lives, but their deployment and operation also have a significant environmental impact. As governments and other stakeholders have grown more aware of this impact, greater priority has been placed on the need to address it, via regulation or other means. But effective mitigation requires an informed understanding of the technologies and interactions involved. 

This workshop will bring together key participants involved in a number of different technologies - data streaming, artificial intelligence, and quantum computing - that are at different levels of maturity, commercialisation, and adoption. The session will discuss the challenge of governance as it relates to the environmental impact (particularly energy consumption, but also considering other indirect drivers like mining for rare earths) of these technologies, and what different stakeholders can do (and are doing) to develop strong awareness, understanding, and cooperation towards practical and effective mitigation strategies. 

== Format == 
Brief introduction from the three key participants, followed by an open discussion. 

== Further reading == 
* [https://www.oecd-ilibrary.org/science-and-technology/measuring-the-environmental-impacts-of-artificial-intelligence-compute-and-applications_7babf571-en/ The AI Footprint: Measuring the environmental impacts of artificial intelligence compute and applications, OECD]
* [https://github.com/Lizaterdag/QIH-2022/raw/main/Qubits_for_the_Kids.pdf/ Qubits For The Kids: Realizing a sustainable quantum internet for the smallest/future researchers]
* [https://www.greeningofstreaming.org/less/ Greening of Streaming “Low Energy Sustainable Streaming” (LESS) Accord]

== People == 

'''SME''' 
*Minda Moreira
The Subject Matter Experts (SME) support the programme planning process throughout the year and work closely with the Secretariat. They give advice on the topics that correspond to their expertise, cluster the proposals and assist session organisers in their work. They also ensure that session principles are followed and monitor the complete programme to avoid repetition.

'''Focal Point''' 
*Chris Buckridge
*Vadim Pak
Focal Points take over the responsibility and lead of the session organisation. They work in close cooperation with the respective Subject Matter Expert (SME) and the EuroDIG Secretariat and are kindly requested to follow [https://www.eurodig.org/get-involved/planning-process/#tab-organising-a-session EuroDIG’s session principles] 

'''Organising Team (Org Team)''' ''List Org Team members here as they sign up.''
*Emilia Zalewska
*Rainer M Krug
*Amali De Silva-Mitchell
*Katrin Ohlmer
*Michael J. Oghia
*Gianluca Diana
*Paolo Gemma
*Nadia Tjahja
*Steven Setiawan
The Org Team is a group of people shaping the session. Org Teams are open and every interested individual can become a member by subscribing to the mailing list.

'''Key Participants'''

* Celine Caira and Johannes Kirnberger, OECD
* Vesna Manojlovic, RIPE NCC
* Eero Lindqvist, Finnish Data Center Association
* Steven Setiawan, EMJM CIRCLE 

'''Moderator'''

* Patrick Penninckx, Council of Europe

'''Remote Moderator'''

Trained remote moderators will be assigned on the spot by the EuroDIG secretariat to each session.

== Current discussion, conference calls, schedules and minutes ==
See the [[{{TALKPAGENAME}} | discussion]] tab on the upper left side of this page. Please use this page to publish:
*dates for virtual meetings or coordination calls
*short summary of calls or email exchange
Please be as open and transparent as possible in order to allow others to get involved and contact you. Use the wiki not only as the place to publish results but also to summarize the discussion process.

== Messages ==  
''Rapporteur: Francesco Vecchi, United Nations University – CRIS''
 
# '''Nexus between digital transition and environmental impact:'''<br />The Council of Europe has recognised the nexus between digital transition and environmental impact, and its connections with human rights, child abuse, and exploitation. The most effective critical paradigm to unpack this nexus is composed of direct (e.g. energy consumption, mining of rare minerals and raw materials, etc.) and indirect environmental effects (e.g. results of the implementation of digital innovation in industries, etc.). However, a standard measure to analyse these outcomes still needs universal acceptance. 
# '''Environmental impact of hardware infrastructures:'''<br />Though many think AI is software and ephemeral, it is actually rooted in concrete infrastructures, as well as cloud services that are operated through huge factories and data centres filled with computers and storage devices. Moreover, quantum Internet is far from being sustainable. So, to decrease the environmental impact of the Internet, it is first crucial to determine the green metrics for measuring it.
# '''Decision-making process:'''<br />The current decision-making process lacks knowledge regarding the environmental cost of each decision and of new digital technologies, and struggles to concretely implement sustainable technology by design. Therefore, regulation should take a consultative and iterative approach, starting from improving measurement, standards and collaboration on data collection, then looking at the complete life cycle impact.

== Video record ==
https://youtu.be/FMX6pYK3eWg

== Transcript ==
Provided by: Caption First, Inc., P.O. Box 3066, Monument, CO 80132, Phone: +001-719-481-9835, www.captionfirst.com


This text, document, or file is based on live transcription. Communication Access Realtime Translation (CART), captioning, and/or live transcription are provided in order to facilitate communication accessibility and may not be a totally verbatim record of the proceedings. This text, document, or file is not to be distributed or used in any way that may violate copyright law. 


>> PATRICK PENNINCKX: We won’t be interrupted as long as we have this representation, then we are all set. 

So we’re online, are we? 

Okay.  Well, hello, everyone.  Thank you for coming to this afternoon session.  I suppose that for some, this is still the morning session, and for others, they’ve had a late evening session.  We know that our interconnected technologies are an integral part of our lives.  I don’t think I need to repeat that, but the development and the operation also have a significant environmental impact.  And as governments, and other stakeholders have grown more aware of this impact, it has been a greater priority that has been given to it.  So give you an example, two weeks ago, we had a Council of Europe summit a few weeks ago, which identified two major issues, one digital development; and second, environmental impact.

So I think it is important that we look at those developments jointly and make sure that these are the challenges for our societies, not only for all the citizens but also for the international and regional organizations.  That will be a little bit of the debate that we will be having.  And therefore, we are having key participants which are involved in a number of different technologies, data streaming, data intelligence, quantum computing that are at different levels of maturity, commercialization and adoption.

We will be challenging governance as it relates to the environmental impact, particularly the energy consumption, but also considering other indirect drivers like mining of rare earths and rare materials.

And so the Council of Europe, has been working on digital technologies for some time now.  We have been, as I said yesterday, in the opening panel, we have been involved with the EuroDIG and with the IGF since its creation, and it’s a very high interest on the topic of ecological development, environmental development.

Human rights, we stand for human rights.  It obviously cannot take place if we don’t have a healthy biosphere and that healthy biosphere, we are all in charge of.  Without a healthy functioning of our ecosystems, there are no human rights.  So we definitely need to look at what are those possible impacts and how can we mitigate those impacts?  But before – but also our European court of human rights, when confronted with serious ecological issues has looked at it through the lens of human rights and we are thinking of new instruments in order to respond to all of that.

But before we go into the matter of how are we going to respond to it, I think it is also important that we look at what is actually that impact?  What is that environmental impact?  And what are the ecological consequences of use of this, but also of the main data gatherings and online communications and the use of digital technologies in general, including artificial intelligence? 

So I asked Steven Setiawan to give us a little bit of an oversight and then I will give the floor to the other panelists, Vesna Manojlovic, and Celine Caira.  Steven, it’s your job to enlighten us a little bit on what the sequences are.

>> STEVEN SETIAWAN:  If we talk about the environmental impact and I will use the life cycle assessment for 2001 as like the main document.  Because, like, we are talking about government and we need to see on the international standard.  So we have four stages in assessing impact, basically environmental impact, and I want to make it simple.

Basically, if we talk about environmental impact, we are talking about things that are – become the output to the environment.  For example, if we produce this water bottle, what comes out of it?  We have energy and mass and there will be something that will be produced out of it.  And like there are several categories.

For example, acid that will be released to the environment or, like, fresh water or like, ozone layer depletion, like that, but the consequence will be divided into three major categories.  So basically, these links that are emitted into the environment, they reduce the quality of life and it becomes the consequence that we, like we experience in the front ways.  Yeah.  Of.

>> PATRICK PENNINCKX: Thank you.  Thank you so much, Steven.  Of course, well, I forgot to say that we have three online speakers here in the panel.  Johannes Kirnberger and Celine Caira.  And they will look at the direct and indirect impacts of the online impact.  Can you, I don’t know who will be speaking first, but you have the floor, as the two of you are based in Paris, and speaking from Paris. 

Celine?

>> CELINE CAIRA: Thank you.  Apologies we can’t be with you physically today.

So, indeed, you know when it comes to AI, you do have – we find it useful to use this direct and indirect framework as well.  It’s interesting because when we think about AI, you know, often we think of something that is perhaps software or something, you know, ephemeral, but it is really rooted in concrete infrastructure, like data centers, like the computing stack that train these large models that process data at a large scale.

And of note, you know, we’ve really seen over the last decade or so, around 2010 with the advent of deep learning, we’ve seen these models really balloon in terms of their capacity, their size, their parameter count.  Notably, you know, since late last year, again, with generative AI, we have now really seen the rise of big large language models and some experts are coming out and saying, you know, this has environmental impacts that could grow by several magnitudes as a result of the models getting bigger, processing even more data.  And so that’s for the training.

But also the inference, you know, being so widely used, what we call, you know, inferencing, every time you look up a recommendation, it sends, you know, an inference is done to get you that information, ChatGPT, for instance.

We are seeing the scales grow and we are seating the models getting bigger.  Why is this problematic?  As I alluded to, simply put, it’s used to train these models and also to use them, have their own environmental impacts.  I like the idea of using the life cycle impact as well, because at each stage of the life cycle of this resource, there are emissions.  There’s energy use and there’s also water.  Water is something that’s often not talked about to cool data centers, for instance.

So at the OECD, we look at this, through the OECD AI expert group on AI computing climate.  We released a report last year at COP27 that looked into kind of a broad – it sketched a broad framework for how one can think of it, direct and indirect, environmental impact.  So I will briefly pass it to Johannes to give a very high level overview of what that framework is and then we can jump into the discussion.

Thank you, Patrick.  Over to you, Johannes.

>> PATRICK PENNINCKX: Thank you so much, Johannes, would you like to add something to that?

>> JOHANNES KIRNBERGER:  To use the framework, something that was already in the ICT industry has been used and the differentiation between direct impacts which means the direct effects of the infrastructure that is behind the technologies such as AI and then the indirect impacts which concern the application of AI and in our case, or of digital technologies, both in a positive and negative way.  So just as scan.  When you use a networking server that is part of the compute stack, part of the data center, when it comes to the direct effects it needs to be – you need to kind of mine minerals and it needs to be manufactured and shipped to the data center.  And then the majority is it consumes energy and water, which both have an associated carbon footprint.  And then it’s being discarded at the end of its life cycle.  That’s the direct impact.

Now the indirect impact would be the application of a given, for instance, AI model which could be used for a lot of positive things, for smart energy systems in networks, for transport, for climate prediction, but it can be used in an active way, for instance, in oil and gas exploration. So what we really try to have the framework as wide as possible and look at all the impacts that exist, and as Steven said, the life cycle approach is really useful there. 

>> PATRICK PENNINCKX: Thank you. Thank you so much. I was almost thinking we should have a carbon footprint and H2O footprint on each time we are using our computers here in this room or anywhere else so there’s like a barometer that comes up – pops up on our computer and gives us an indication of what the impact is.

I don’t know, Guillermo, if that’s what the Finnish data center was looking into. You said you may give us a little bit more insight on the EU energy efficiency directive, which, of course, also includes data centers, including the network data centers. So can you give us a little bit more insight? Am I bit too far reaching by putting a barometer on each of our devices?

>> No, I think that’s a good goal because everybody is talking that we are using cloud services. So what are cloud services? They are these huge factories, data centers, filled with computers and storage devices. And they consume really a lot of electricity and also water when we are in the warmer climate zones. And if you take a look at the environmental reports from, for example, Meta, Microsoft, Google, and their greenhouse gas emissions, what would you think is the biggest contribution there? 

The consumption of electricity, their own operations. No. Over 90% of their emissions come from the supply chains. So the devices that we put into the data center, that are manufactured, that are delivered, and all of these has a carbon footprint and it’s a footprint. And I think that this is not transparent. And we have to make this more visible to the users.

For example, training of the GPT 3 model, it consumed 1300 megawatt hours, and produced 550 tops of CO2 just to train.

>> PATRICK PENNINCKX: Thank you. It’s an excellent illustration.

Well, I’m looking at Chris Buckridge here. I’m not going to ask you to intervene, Chris, but you were so kind to invite me to Iceland and Iceland is one of the countries that promotes the installation of data centers very much. So I was going to ask Vesna Manojlovic from RIPE NCC if she has anything to add. She’s also the, let’s say, fourth or fifth person in the room that joins us remotely. So Vesna, would you like to add something to the discussion that we have had so far? 

>> VESNA MANOJLOVIC: Yes, thank you. I’m pleased to be invited and to be in your company. And in addition to what the previous speaker said, I would like to bring the voices of the academics and researchers that are gathered together in several events that I was part of. I’m kind of just speaking on behalf of them, because I’m neither the academic, nor the researcher. I’m the community builder. So I cross connect these various communities, mostly technical communities.

And the main paper that was suggested here for this session was the one that we worked together about the quantum Internet, which is – which has been working on the connection between the sustainability and the quantum Internet and they are really the opposite from each other.

But, on the other hand, these young researchers went really deep into trying to figure out how can this very new technology develop in a sustainable way in the future, because right now, that is still kind of a science fiction or like on a very early stages of the development. So now is the time for them to think about their responsibilities towards decreasing the environmental impact on their side.

And then the second community that is busy with, this the standardization body for the protocols that are underlying all the Internet networking and they have organized a workshop about the impact which has been published as a report. And their pain work is on determining the green – the green metrics for actually measuring impact so it can be decreased.

And finally – sorry, one more last week there was another academic conference called Computing within Limits and they research even more alternative and kind of fringe topics about how can the computing itself, as a computing science take into consideration the environmental impact of the Internet? 

>> PATRICK PENNINCKX: Thank you so much. I would ask here in the room, because I’m also following online, if there’s any interaction here, or intervention. My colleague, Vadim Pak, who is also with Chris, was preparing this session. Chris, you have a question. In this room, you have to speak out loud, because there are only the microphones here, which capture the voices in the room. So Chris Buckridge, maybe introduce yourself.

>> CHRIS BUCKRIDGE: Yes, Chris Buckridge, I’m here as a technical community representative of GFMAG. And I actually have to jump to the other session. And so I’m very sorry to miss the rest of it. I wanted to note something, why we put this workshop together and thought we had. What I found really interesting in listening to each of these speakers so far and some of the issues that they are talking about is that each of them seems to have some surprising facts.

Like, you know, the energy comes from here. Or actually the impact is in this place. And the fact that we are still getting those surprises, the fact that we don’t fully yet understand impact all of these technologies have is something where a multistakeholder governance forum can bring together people and start to dig into those unknowns. Instead of saying, okay, we have done some research and here’s where the impact is and maybe others will say, okay, that makes a lot of sense. That’s not what I thought initially. But then how do we work with that?

So it’s really moving that discussion forward, hopefully in productive ways about how you actually mitigate the impact that these technologies have. So I’m very glad, but also very sorry.

>> PATRICK PENNINCKX: Thank you so much, Chris, for also giving us a little bit of frame of why we are organizing this and maybe a request to you, because you are dashing out.

But the key thing, even though this is not first time that ecology and environment is being discussed in an Internet Governance Forum or in EuroDIG, but I think it is important that this topic, the topic remains on the agenda and also for the IGF in Kyoto, I’m quite clear.

I have Vadim Pak online who is asking a question. So Vadim, please go ahead.

>> VADIM PAK: Yes are, thank you very much, Patrick. So my name is Vadim Pak. I was one of the focal points for the session. And just bouncing off of what Chris mentioned, one of the things that really puzzled us and we really wanted to have the view of the participants on this, human decision making as it stands now, in respect of all of these things which create environmental impacts, is what it is. The decisions are taken with not necessarily no – when decisions are taken, what would be the environmental cost of these decisions, and my question to the panels would be, in your view can we organize decision making on these questions better? 

So are decisions taken the right level hierarchically, do people who take these decisions have access to the right type of information or maybe the whole game should be rearranged.

That would be my question. Thank you very much.

>> PATRICK PENNINCKX: Thank you. Thank you very much. That’s a million dollars question, I would say. Who of the panelists would like to respond? Can I ask Celine to start to kick start us? 

>> CELINE CAIRA: Thank you, sure. Yes. I think that gets at the heart of international policy making as well, which is even more complicated because you have the regional governments, the national governments, the international level and much of the work that we do at the OECD is to try to build those bridges when it comes to decision making.

And so you know, what we in the policy space love to call complicated word called “Interoperability” and so it’s the basic concept that, you know, your rules and your laws, your regulations work together in the same direction.

If you have decisions that have been taken in different jurisdictions, and this applies to many policy areas that are not interoperable, it could incentivize actors to engage in forum shopping and find loopholes to ultimately evade reporting, you know, in the context of AI we have heard some ways of reporting, you know, to hide emissions in the cloud, for instance.

So, you know, regulation needs to be made in a way that’s interoperable at a global scale, and also to have basic definitions to ensure that we are talking about the same thing. It sounds like a simple point, but it’s actually quite complicated.

One point that I will end on, the definition of an AI system is, at the EU level, around the world, we have been discussing this for some time, with the OECD, and the OECD AI principles we have a working definition of what an AI system is. Of course, generative AI and new applications will, you know, shed light on, perhaps how that definition needs to be flexible and adapted to the times to be time proof. I guess the main take away there is to ensure that regulations and decisions, you know, lead to something that’s interoperable, and that’s rowing in the sim same direction and are we discussing the same topic when we talk about these policy discussions.

And as we have been talking Vesna has been putting quite a number of documents in the chat. If you are connected at the same time, please do have a look at the different documents that have been posted there on the indirect and the direct impacts of digital Internet technologies and how to mitigate them. So I think that’s a crucial issue. 

Now, from your input, Celine, I have the there is a question that pops to mind, and I would like the audience to react. Are you saying that there is a need to regulate the environmental impact, regarding energy consumption and also considering other drivers of Internet connected technologies? I won’t give the floor back to you. I would like Johannes or maybe my colleagues here in the room. Johannes, would you like to answer that?

>> JOHANNES KIRNBERGER: The short answer is yes. The real question is how and what to regulate. Because –

>> PATRICK PENNINCKX: Johannes, these are probably going to be the next questions. So don’t anticipate.

(Laughter).

We will keep it with the short yes for the time being then. And maybe I will ask my colleagues here in the room to also respond to that. Maybe Guillermo, is that a response that is appropriate?

>> I think that it’s a step to the right direction. So on the EU level, it has been seen that the energy consumption of data centers is growing and growing. And since EU is trying to reduce the overall energy consumption, they have to address this is in some way.

And the steps that have been laid out in the cost the energy directive is that first, the data centers will be obliged to report certain key performance indicators, mainly the energy usage and also let’s say water usage, their let’s say ICT capacities and some level of those.

And first, this information is gathered, May next year is the first reporting round. And then the EU commission shall analyze that data. And the next year they will propose their report to the EU parliament and to council and also give a message for parameters. I think we are taking steps in that direction that there will be more regulation coming to this area. 

>> PATRICK PENNINCKX: Thank you. Thank you. 

I have a couple of people in wanting to intervene right now. And follow up question will definitely be to which extent will we then come to this more comprehensive picture of the impact that will – that will need to be seen and that’s a story to be continued, I believe, because as Chris was saying, for the time being, err we are still discovering new areas. And I think it’s important.

Vesna, you represent, and I will afterwards give the floor to another Chris, Chris Appleyard. So get ready, Chris, in order to prepare your question. But Vesna, you said you represent a technical community. What does that regulation mean for you? And is that needed? Not only regarding the energy consumption, I would say, which is, of course, extremely important for the network that you represent, but also I see these rare materials being used at all times to – for our digital technologies. So maybe Vesna, can you maybe say something more about that question of regulation?

>> VESNA MANOJLOVIC: Yes, through this answer, I also want to address two things that were mentioned by previous speakers from Celine and the Finnish gentleman in the room which is – and then your question, Patrick.

So although I personally believe that the regulation is needed. I agree with the short answer, yes. The technical community and the network operators have been traditionally strongly for industry self regulation, and also traditionally the Internet has been kind of very hard to regulate. And so I think that’s going to be the case or it has been the case also with regards to the – to the regulations of energy consumption, water consumption and specifically because what Celine is said that resonated and made me laugh. Then we have to agree on what is the definition of the Internet. And then everybody starts pointing fingers to the other, other part, no, no, it’s the manufacturing. No, no, it’s the provisioning. It’s not us. We don’t have to do anything.

Now stepping out from this kind of representation and becoming, like prescriptive of what I believe has to be done, is that general decrease in everything. This is what we on the highest level this was an agreement, that would be the Paris Agreement that we initially have to decrease by a certain percentage, it used to be 7% and now it’s 10% per year everybody. So in the network engineers and the network operators believe in the self regulation then they should decrease all of their usage, emissions, water, energy, equipment that they are buying, by 10% per year and then start reporting that so then the regulation cycle can actually check has that been going on or not.

But looking into the measurements that have been done already, that has not been the case. So the most the network operators can report is that their energy usage has not been increasing but that’s for specific operator in general the whole sector has been increasing every year and that is unacceptable according to the Paris Agreement. 

>> PATRICK PENNINCKX: Thank you, thank you, Vesna. Obviously, we always put the responsibility elsewhere, as you said and I think it’s also in reaching those 7.5 or 10%, next year 12.5%, it’s always the others that need to do it. 

Chris, Chris Appleyard from Nominet, the floor is yours. Do you agree with this analysis?

>> AUDIENCE MEMBER: I had there. I hope you can hear me well.

>> PATRICK PENNINCKX: Yes, we can hear you very well.

>> AUDIENCE MEMBER: Thank you. So I like that were focusing our time on reducing and to usage, and improving efficiency, but shouldn’t we also focus our time on what to do with electronic waste, to maybe ensure that we send out electronic waste back to the manufacturers so that we can create a more circular economy. 

>> PATRICK PENNINCKX: Thank you. Excellent question. I think it’s definitely part of the paradigm. So can we maybe, Steven, from your background, as industrial ecologist, can you maybe say something more about this proposal and also about the regulation issue? I haven’t given you floor on that yet.

>> STEVEN SETIAWAN: So if we are talking about regulation, I will say that physical infrastructure, and since there is physical impact on the environment and also energy impact. So it should be regulated with the same manner as other aspect of, like, for example, like with industry and also regulated. But the question is, like, who will become the actor.

We know that the Internet is a network of networks. And the agency, are and like the ownership of the one product or service will be failure. So the goal is to make it more regulatable.

And regarding the circular economy, I completely agree that we know that European Union has proposed – EU critical materials act that one of the key points, we need to increase the recycling, and we have to talk about another proposal to go to deep sea mining that also still in infancy. Some companies want to start this this July, although the environmental impact is not known. So the priority is to reduce the primary consumption of the raw materials, and it’s true the recycling of those ewaste. I agree.

>> PATRICK PENNINCKX: Thank you. Thank you very much. We also continue to say that reality is often the opposite. Are there any reactions here from the room? Yes, please? Present yourself, YOUthDIG no doubt. 

>> AUDIENCE MEMBER: Thank you so much for this discussion. I think this it’s highly essential for the time being to speak about sustainability and technology, but I would like to just stress two points that we didn’t speak about, and it’s a related topic.

From the user point of view, I think that what has to be addressed is the greenwashing that we are victim to. Even if we want to be sustainability, we first think that we are not, but this is something I would like highlight.

And the second point is something I have already spoken about, but I would like to stress it more. It’s regarding the – if we speak about mining, if we speak about the tasks of technology, and the human rights violations that are taking place in order to do that.

So I would really like a more environmental sustainable and protection of human rights go hand in hand.

>> PATRICK PENNINCKX: Thank you. Thank you so much for these very important points that indirectly come from the YOUthDIG. I think it’s extremely important and our rapporteur here, Francesco is eagerly typing the greenwashing and the human rights violations down. I think Francesco, if I may use my capacity as moderator, I would add child exploitation to that, which is very closely linked also to the human rights violations that we see. 

Are there any other questions from the room? I see that you are eagerly waiting. Yeah, yeah, yeah. Also from YOUthDIG.

(Laughter).

>> AUDIENCE MEMBER: Well, since last year –

>> PATRICK PENNINCKX: No longer. Okay. You passed the barrier. 

>> AUDIENCE MEMBER: Actually, I remember a couple of years ago, an application for streaming, which was set to drive the data consumption and thus energy consumption. It seems like the attention has shifted to AI. I find it was so interesting it was mentioned how much energy consumption –

>> PATRICK PENNINCKX: The training.

>> AUDIENCE MEMBER: Associated with the training and the AI model and the use of the analysis. And so my question would be, what do you think of this AI consumption, and how do you think it is from streaming and how do you think it will evolve in the future? Is AI the future driver of energy consumption?

>> PATRICK PENNINCKX: Thank you. Thank you. I hope that our colleagues online understood the issue. It’s basically comparing between different forms of the use of digital technology, what is most energy consuming. It’s almost like a ranking.

Celene, has the OECD or Johannes, has the OECD done some work on that?

>> CELINE CAIRA: Sure. Maybe I will start and I will pass it to Johannes. I think just to make take one comment around the training of ChatGPT that was just mentioned, as Chris before, you know, we have a bunch of surprises, things that are counterintuitive. I would like to highlight two of those quickly as well. One that we talked about is actually the embodied emissions of hardware is something that’s not reported. And so that’s one potential surprise.

The second one is what we’re actually hearing from experts and it would be great to get some concrete evidence and time based series data on this and that’s what we are working to do. We are actually hearing from experts that as I mentioned before, the training although huge is not the most compute intensive and energy intensive part of the AI system. It’s actually the inference. When you are talking about a system used over time, it’s the inference of the model, the use of the model that’s more compute intensive. I wanted to highlight that, because it’s counterintuitive and everyone talks about the training all the time, but we might not actually be highlighting the most compute intensive portion over the lifetime of the model.

And then I think the question related to – how does AI compare to ICT measures overall, if I kind of understood?

>> PATRICK PENNINCKX: Yes, and specifically, we specifically mentioned streaming in this context. So basically not the people in this room here are in charge, but the ones actually following online, even though they saved some CO emissions – CO2 emissions by not coming here, but it’s actually the streaming that would be problematic. Do you agree, Celine?

>> CELINE CAIRA: It’s interesting, and the IEA, the International Energy Agency has done some analysis comparing Internet traffic to data center energy use, and while Internet traffic has gone up quite significantly, the data center energy use has remained flat. That’s interesting. That seems to point efficiency gains of the hardware even though the domain is increasing. We would love to have comparable numbers add another line on that graph when it comes to AI. Unfortunately, we are kind of early in the data collection space when it comes to A I. and we’re working to standardize these measures.

Some measurements from Google says they use 15% data center usage for AI. Maybe Johannes has more to add or Pat trick, you want to respond as well.

>> PATRICK PENNINCKX: No, it Johannes would like to say something more about, it please do. And then we have a very interesting comment job line, which I will read out afterwards. Johannes?

>> JOHANNES KIRNBERGER: So it’s very hard to put numbers behind these numbers. It’s really hard to kind of disentangle the AI from overall data center use and ICTs. That’s what we have been trying to do over the past year, but it tends to be hard. As AI proliferates, I think that’s true for the entire ICT, to point – what is sure is that AI use will be increasing and ICT use will be increasing and since we have enormous opportunities for the carbon free transport, it is really vital and the very short answer is, yes, we need regulation. It’s vital that the infrastructure that’s being used for these applications is as environmentally friendly as possible. And hopefully we can decrease the 7 to 10% as we need to, as with use this for the overall sustainability goals that we have.

>> PATRICK PENNINCKX: Thank, thank you, Johannes. Before I give the floor to Vesna, I think Jorge’s question, actually the question and answer at the same time. In the – is the idea of a sustainability by design in many use, in security by center sign and privacy by design are already well known but they are not fully utilized and required. The main reason cloud services have security incidents is that the configurations are not readily secure and private. By extension, configurations are not by default the most energy efficient. Can this be required in norms? Ask this have a potential as a solution, or is this misunderstanding of the technologies themselves? 

Vesna, you wanted to respond to this? And I would ask my colleagues panelists in the room also to react if – just indicate, because now with these energy systems people can indicate that they want to speak but we also have hands. Vesna, are you are –

>> VESNA MANOJLOVIC: Yes, and I find myself kind of lagging behind all the topics that are being discussed because there was a question from the room about the rankings, like, could we have –

>> PATRICK PENNINCKX: That’s right.

>> VESNA MANOJLOVIC: Like, could we have the ranking and from me, coming from the communications side, I would say that maybe there are numbers about that, but for me, it’s kind of also even intuitive. Like, if we are trying to communicate with, like the Internet of humans when we want to talk to each other, I would say that the ranking on what requires the most resources goes from the VR, like virtual reality, to the video call, to the audio call to the text messages like chatting, and then to the intermittent text messages like emails or even less demanding texts which is not going there, like between users realtime.

And this is for me, let’s say the recommendation for the future, but for, let’s say dystopian future where all of these other, more demanding ways of communication will not be available and they are already not available to a lot of people in the different parts the world. So there is a huge inequality in terms of what type of speed and bandwidth are available to different people. I think if we want to decrease the environmental impact, we could, as individuals and as consumers decrease this following this ranking to some kind of level that we are comfortable with.

But this is a very minor impact that we have as individuals. So, again, we can go back to the regulation or other methods that we still have to talk about, like – and then coming back to you, like, if this would be sustainable technology by design, how would we make sure that these norms are actually followed and implemented? And that requires the changes in the culture, let’s say, and, of course, in economic and political systems, because it’s part of this cycle of demand and supply and people wanting to have everything faster and then the providers actually making sure that that is available without thinking of what the actual impact is.

So it’s a conundrum and it is a difficult question.

>> PATRICK PENNINCKX: Thank you. I think it’s with also with the extrapolation and explosion of ICT instruments, maybe an advice to the YOUthDIGers that maybe they should start with a ranking of the worst ecological imprint of the ICT tools and this – because in many of the big Internet companies, the zero ecological imprint has already been put forward as one of the positive arguments for the buying of their products.

So maybe that’s an idea for the future. 

I also already asked the question, are the current regulations, the proposals, the regulatory discussions, are they informed by a good understanding of those technologies themselves? Their markets, their interconnectivity, their interconnections involved, as well as their direct and indirect impacts on the environment? 

So are these regulations already appropriate? Celine, sorry to turn back to you. 

Because you are in Paris.

>> CELINE CAIRA: Thanks, Patrick. I think that’s a good question. And, indeed, you know, one of the very difficult things around, you know, is AI – how do we regulate AI and environmental impacts is, you know, a lot of these large language models and foundation models, they are black boxes in a sense. Even the brightest minds in AI can’t quite understand how they work. And it’s really hard to regulate something and to create rules for something that you don’t completely understand. So, you know, there are applications to societies and economies are also very new. When you think of generative AI and how it’s diffusing greatly through the economy. So there’s a lot of unknowns, but this is why, I think, regulators need to take a consultive approach, need to take an iterative approach. And while we have seen a lot of activity when it comes to AI, in the AI safety and ethics space, the sustainability dimensions have lagged behind, until recently, I would say.

And so we were – it was great to see that there’s actually some language now and the most recent version of the EU/AI act that is, you know, making its way through the appropriate kind of validation processes, but there’s language around life cycle impacts when it comes to AI and the mitigation of environmental impacts and the reporting of environmental impacts when it comes to high risk systems. This is very interesting and so it remains to be seen how Europe will eventually implement these regulations, but it was a positive signal that at least is included there.

And I’m sure, Patrick, in the work that you are doing at the Council of Europe, you are looking at both people and planet. I know UNESCO has done some work in their recommendation on AI ethics and the OECD AI principles, and so at the international level, we try to capture it. So that is an interesting development. 

>> PATRICK PENNINCKX: Yes, definitely. Thank you so much, Celine and since I’m also together with Vadim, representing the Council of Europe here, the link between digital development and especially the development of artificial intelligence and the ecological impact have been made and that’s certainly something that we will be looking into in the future as well. That’s why we co organized this very specific session, but also proposing to continue this work in the meantime, but also leading up to the IGF and we hope to be able to, together with our Japanese partners there as well to look more directly into the ecological questions related to this.

Are there any questions from the room here? Yes, please. 

Speak up loud because the capture here is not so good.

>> AUDIENCE MEMBER: (Inaudible).

>> PATRICK PENNINCKX: Thank you. Thank you for your opinion on that no regulation on AI would really be needed in the current times.

May I ask my colleague Vadim maybe to give the perspective of the Council of Europe.? I’m sorry so draw you into the panel like this Vadim, but I think this is your area of work. So please.

>> VADIM PAK: Thank you very much, Patrick, for giving me the floor.

The Council of Europe works on certain aspects of, you know, sustainability, protection – and protection of environment. 

But focus of human rights. We are looking at human rights effect of environmental changes more than anything else. So if we speak about the current work in the community of artificial intelligence, the angle that we’re taking is human rights angle. We are also developing a –

>> PATRICK PENNINCKX: If I may just interrupt, the angle that you are taking is also on regulating AI and I think the question was definitely primary on that.

>> VADIM PAK: Yes, in regards to regulation on AI, this is a very complex question, because – because of the diffused nature of artificial intelligence, and the way they permeate societies, it’s very likely that the approach that national regulators will take will be extremely context specific.

And while it’s clear, that regulators may choose to regulate technology as such, this is something that is up to regulators to decide, some countries may also go for, you know, much more cautious approach which is I didn’t the work that we do in the committee on artificial intelligence would have to be able to accommodate various possible regulatory approaches when it comes to AI. That essentially means that countries that would like to go after technology as such would be able to join the convention that we’re working on and then countries that would like to, you know, have more limited or context specific measures would have to be able to join this instrument as well. This is the answer.

>> PATRICK PENNINCKX: Okay. Thank you. Thank you, Vadim. So that the opinion in the audience, is looking at what is happening in the environment.

We are already coming to the end of our session here. And if you could, inning so, what are the solutions that we could steer towards a more effective approach of containing, let’s say, the ecological footprint with the development of digital tools? Celine, you are always my spearhead. So I’m asking you again to kick off there.

>> CELINE CAIRA: Sure. Thank you.

>> PATRICK PENNINCKX: And others please be prepared.

>> CELINE CAIRA: That’s great. I will answer that question with a bit of my wish list as well and obviously a very typical OECD response. So I think as they say, we can only change what we measure, right? And so this would just be, again, a call for improved measurement, improved standards, collaboration on data collection, and really that – that data collection accompanied the regulation, but we also build capacity so that people are able to collect data and it’s usable.

In terms of solutions, I think one thing that we have been reflecting on and I will just leave the group with is this concept that has emerged, AI model cards. AI models need to come with a nutritional label on them or information that reports the number of parameters, data used but also how much water and how much energy and what were the carbon emissions so that those fine tuning these models know what they are working with. And then we have something standardized. They have been called various other things. Maybe it’s something else but I just put that out this as food for thought. Thank you, Patrick.

>> PATRICK PENNINCKX: Thank you very much, thank you for participating. Johannes.

>> JOHANNES KIRNBERGER: Yes, so to kind of add to the wish list is what we have been talking quite a lot to look at the entire life cycle impact and going out to the direct/indirect impact. And the indirect impacts meaning the application of digital technologies to leverage them for the key sectors so they can have the most impact and not use them where they have adverse impact on the climate. Energy waste is the largest of them all and only a fraction is recycled. We heard about the critical minerals that are needed to add the infrastructure. So really having the life cycle approach because energy and carbon get kind of all the attention and in a lot of reports. And so widen the angle a bit and look more comprehensively at life cycle impacts would be very well valuable. 

>> PATRICK PENNINCKX: Thank you so much. Here Steven. What is your wish list?

>> STEVEN SETIAWAN: Okay. So I think from this dialogue, both from panelists and from others, we all agree that AI is like intertwined and something in the future. So my wish list is what is the main thing that the ICT needs to see in the environment, it’s not the AI that is in had the future, but it’s the current situation. What is the biggest environmental impact that ICT give? And it’s mainly physical thing. And if we talk about energy contribution. It’s not the root problem. The root problem where the energy comes. So we need to so many of these basic things and give this to researchers so after we solve these basic things in ICT, we can focus on the later development.

>> PATRICK PENNINCKX: Thank you. Thank you. And that would lead us to another one hour discussion, but since we are being called by the organizers, Jiro, what is on your wish list?

>> We need to measure more clearly what is the impact of the ICT to the environment and also forming a holistic bit of that and make it transparent.

>> PATRICK PENNINCKX: And with the economic barometer on each of our devices.

I still have Vesna. No, I haven’t forgotten about you, Vesna.

Vesna, your wish list.

>> VESNA MANOJLOVIC: So my wish list, degrowth and urgency. 

>> PATRICK PENNINCKX: In that order? 

(Laughter).

>> VESNA MANOJLOVIC: Yes.

>> PATRICK PENNINCKX: Okay. We heard some wish lists here as well. The greenwashing issues, the human rights violations and the use of models and which are so intensive, computer intensive, the proliferation of ICT the explosion of ICT maybe the ranking of the worst ICT tool with the worst ecological impact. All of them are issues that we discussed.

But I will give the floor to Francesco to give us a two minute roundup before we are kicked out of the room and the red lights start buzzing.

>> It’s almost impossible. Okay.

>> PATRICK PENNINCKX: Francesco. 

>> We have started focusing on the digital change, which is important especially for quantum computers and AI with direct and indirect impacts, but actually, this is really important to focus on because environmental and human rights are deeply connected and so it also connects this to digital transition, as well as child abuse and exploitation as the moderator asked me to add.

Human rights cannot be placed in the biosphere and so they are doubling for data gathering. I already mentioned direct/indirect, and I wouldn’t just focus that much open those because there’s been an much better explained by other people during the panel, but, of course, the problem is that to analyze these outcomes, it’s really important to define a standard measure which was also in all the wish list, of the panel points.

Furthermore, many people think that AI is software and ephemeral, but it’s problematic when it comes to the environmental side, the environmental problems and LLM’s impacts are actually related to processes that actually take place under the concrete and the hardware level. Of course, there’s indirect effects of the application of AI, for example, I don’t know the negative outcomes if AI is applied to sectors that work against climate change and anything else. But what is also important is that I already mentioned quantum Internet. It’s important to think about cloud services because it’s huge systems with storage and computer devices. And here we have the hardware issues for the environmental issues and of course, great somebodies like meta and Microsoft and Google will have the biggest impact contribution, 95%, am I right? That comes from the supply chain and not just for the direct emissions. It must be more than just now. We spoke about ranks which was the example given by Vesna.

And we saw that when we speak – when we focus on the rankings we see that there are huge inequalities and so this is human rights, equality.

I’m almost finished, I promise.

Of course, during the discussion, we highlighted the importance of reflecting on the decision making process, both from the measurement point of view and so the need to support a measure or a number of measures. And maybe the reflection of if sustainable technology by design is possible and beneficial. But we also tried to reflect on the – if regulations can be performed and we actually had a debate on this. Especially when it comes to AI, and also what kind of approach regulators should take.

And finally, of course, we made your personal wish lists when it comes to several solutions. And you basically agreed on the need to find the common standards and have a greater collaboration, data collection, improve measurement and also focusing on where the energy comes from and not just develop the process. 

>> PATRICK PENNINCKX: Thank you. Thank you so much, Francesco. This is the 15 minute delay, which is not my usual thing to do. We will conclude this session. Thank you for this very good online participation as well. The discussions that were held there, please check what you missed.

Also, Min who supported a number of the panelists in their interactions. So very much welcome. Thank you for the numerous participation here in the room, and active participation from the YOUthDIG and the other dig, the EuroDIG. Thank you so much and have a great evening. Thank you so much.

(Applause)

[[Category:2023]][[Category:Sessions 2023]][[Category:Sessions]][[Category:Workshop 2023]]