One of the key challenges is that cloud compute resources aren't necessarily geo-located (although external factors, such as data sovereignty requirements in healthcare that forbid patient information being accessed/ stored outside the country where it was collected, can mitigate this in practice) so national policies alone may not be enough to ensure incentivise (e.g. green tax credits) or enforce (legislating for the mandatory procurement of sustainable options) . Therefore policymakers are needed to advocate for alignment in approach between different countries.

The philosophy for everything from the ground up needs to be to design in the use of less power from the start - if a sensor reads temperature an fires the result over the network, keep the traffic down - will a reading every second do, oe once a minute or ten per hour.

We do not care now but if the 'IoT' grows ar the current rate for too long, we sure will.

Keep the payload small - do the processing at the origin and only send what data is needed to solve the problem, not several kb of padding. Run lines idle when there is nothing to do. Power down totally when we can be sure of not being needed.

As an example of how not to do it, consider the whole battery backed up fibre phone fails in a long power cut fiasco.

All that could have been avoided if the initial standard had had a 'power cut mode' from the beginning, where the whole thing shuts off until there is activity on the right laser wavelength for incoming traffic, and/or there is outbound data on the phone line. An emergancy handset phone pick up can be detected with very little idle current indeed.

Or Why is your WiFi on when everyone is in bed ? Does it need to be?

In a country with tens of millions of households, every negawatt* counts.

Data centres are big, but there are not that many of them. Households look small, but there are a great many of them - much as there are more insects by weight than mammals. 

Mike.

* negawatt - thing not using power at the moment because it is turned off ..

Agreed Mike, we should all connect a time clock on our routers and this would stop pinging noises in the middle of the night as well.  Win, win situation.

Other problem is that all photos sent to me clog up my computer memory and certainly should not be allowed to be stored in my cloud memory.  How can we delete received photos automatically after say a month?    Any idea??

mapj1 said:

Or Why is your WiFi on when everyone is in bed ? Does it need to be?

I suppose it does if you have security cameras that are connected via your WiFi or if there's anything else that relies on the connection such as the monitoring gizmo connected to a friend's pacemaker that is also used to remotely adjust it as they sleep .

My phone is set to go into DND mode at 9pm every night and only my parents and my partner can bypass that i.e. messages or phone calls from them will get through but no one else. This minimises the pings and other noises that can suddenly wake me up in the middle of the night! 

Re photos and the amount of storage space they take up, strangely enough, one of my old blog posts  Getting Your Digital Self In Order... has seen a sudden surge in viewership over the last few months which makes me think that it's now becoming quite a hot topic!  

maybe - though I'd be worried by something like a pacemaker that needs a network connection- if the adjustment is a matter of life and death then really the decision needs to be made and acted upon local to the person  not reliant on any external equipment - if  not how restrictive is that in terms of where she can go without losing contact, and what happens in a power cut ?

Similarly well designed security cameras can do the 'looking at' the images, and only send the ones where something moves or whatever over the network so a slower less power hungry protocol is needed - perhaps the network sleeps for 99 % of the time, and just checks once a second if there is any traffic and it needs to wake up. Such things need to be designed into the protocol from the outset but by and large, have not been, except perhaps with things like Zigbee to a degree.

And again, what happens in a power cut - the length of  time a battery can keep your vital kit going is inversely proportional to the current consumption.

Mike

That’s a really good point, Peter. There are also some great examples of how AI can be used to improve sustainability itself but I wonder if there are any examples of emerging technologies that can mitigate the environmental impact of the AI in the first place, in addition to the policy levers you mentioned.

mapj1 said:

maybe - though I'd be worried by something like a pacemaker that needs a network connection- if the adjustment is a matter of life and death then really the decision needs to be made and acted upon local to the person  not reliant on any external equipment - if  not how restrictive is that in terms of where she can go without losing contact, and what happens in a power cut ?

I would hope that such a device would handle an intermittent network connection.  Without a network, it would go into "business as usual" mode, carrying on doing whatever it was previously.  Settings updates should only be installed once they have downloaded fully and been verified as correct.

I do know someone with a built-in defibrillator, which has a link via a mobile phone to the hospital.  Mostly to alert the hospital if there's a problem.

To strike a balance between innovation and environmental responsibility, the tech industry can focus on several key areas:

1. Energy Efficiency: Invest in research and development to improve the energy efficiency of data centers and AI algorithms. This could involve optimizing hardware design, implementing advanced cooling technologies, and utilizing renewable energy sources.

2. Circular Economy Practices: Embrace circular economy principles by designing products for longevity, repairability, and recyclability. This includes exploring options for refurbishing and repurposing hardware components to extend their lifespan and reduce waste.

3. Data Optimization: Implement data optimization techniques to reduce storage requirements and energy consumption in data centers. This could involve data compression, deduplication, and efficient data processing algorithms.

4. Collaboration and Knowledge Sharing: Foster collaboration among tech companies, research institutions, and policymakers to share best practices and develop innovative solutions for sustainable technology. This could include forming industry consortia, participating in standards development, and sharing data on environmental performance.

5. Transparency and Accountability: Prioritize transparency and accountability in sustainability efforts by publicly disclosing environmental impact assessments, setting measurable sustainability goals, and regularly reporting progress towards those goals.

As for policymakers, they can play a crucial role in promoting sustainability within the tech industry through:

1. Regulations and Incentives: Implement regulations and incentives to encourage companies to adopt sustainable practices. This could include tax incentives for renewable energy investments, carbon pricing mechanisms, and energy efficiency standards for data centers.

2. Research and Development Funding: Allocate funding for research and development initiatives focused on sustainable technology innovation. This could involve government grants, public-private partnerships, and funding for academic research.

3. Education and Awareness: Raise awareness about the environmental impact of data and AI technologies among policymakers, industry stakeholders, and the general public. This could involve educational campaigns, workshops, and conferences focused on sustainability in the tech industry.

By working together, the tech industry and policymakers can promote innovation while minimizing the environmental impact of data and AI technologies.