Can the planet afford AI?
AI and climate change, how teens are using AI, Apple Intelligence, pick of the podcasts and much more ...
What’s been happening
Amid all the hype on the potential for generative artificial intelligence, we all need to note the environmental cost of the AI frenzy – and how this might affect how teachers and schools use and contextualise it.
Microsoft’s emissions have risen almost a third since 2020, the company reported in its annual sustainability report last month. The tech giant has expanded its data centre capacity to meet the booming demand for computing resources to train and run AI models. The power needed to run and cool the servers is astronomical. It is so large that Goldman Sachs estimated that in 2023, AI data centres used as much energy as the total capacity of all US utility-scale solar farms deployed last year and could consume as much energy as Japan by 2026. It is also estimated that about 700,000 litres of water have been used to cool the machines that trained ChatGPT-3 at Microsoft’s data facilities.
This is all happening as the world steadily heats up. A study released last week by 50 leading scientists found that ongoing high emissions of warming gases mean the world is moving closer to breaching the symbolic 1.5C warming mark on a longer-term basis.
Some economists, such as UCL’s Mariana Mazzucato, argue that greater regulation is needed to tackle the climate costs of AI, alongside a more comprehensive, systematic approach to meeting the challenge of cultivating less extractive methods.
Alessio Terzi at the University of Cambridge acknowledges the failures of poor regulation and the policy mismatches that aggravate the climate cost of AI – it can take as little as one year to build a data centre, and up to five years to build renewable energy facilities. However, he also points to the potential for AI to accelerate technological progress, including in the realm of environmental sustainability:
“From a climate perspective, AI should neither be glorified nor vilified. Alone it cannot install the necessary solar and wind capacities, promote a circular economy, or restore degraded ecosystems and biodiversity. These challenges require action by consumers, businesses and governments navigating trade-offs between living standards, energy security and sustainability. But AI can augment human ingenuity in finding a solution to these wicked problems.”
Greenwashing? Perhaps. But is there anything individuals can do in the meantime? A quick checklist includes:
Think before you click – avoid firing up generative AI for simple answers when it uses an estimated four to five times the energy of a conventional web search
Clear the virtual clutter
Minimise cloud storage
Keep devices for as long as possible (mending a smartphone just once can save more than 77kg of carbon emissions)
These actions are all things that teachers and schools can consider, as well as including the climate implications of AI in AI literacy lessons. When we previously covered this topic here, we mentioned theJisc report by Scott Stonham, Exploring digital carbon footprints, on the hidden environmental cost of the digital revolution and the steps universities and colleges can take to address it. While the reports’s focus is on digital carbon impacts in general rather than AI specific, and on the tertiary sector, there’s a lot in there for schools and individuals. For example, there is advice on choosing video call providers (Teams, Zoom, Google Meet etc) based on their carbon emissions and notes that one of the quickest and possibly easiest areas for improvement is email signatures and minimising unnecessary attachments, including images in signatures.
AI roundup
Apple Intelligence
Apple’s new personalised AI system – Apple Intelligence – is a tie-up with OpenAI that promises to deliver personalised AI services while keeping sensitive data secure and protected through its Private Cloud Computer system. Reaction has generally been muted (see, for example, ChatGPT on your iPhone? The four reasons why this is happening far too early) but it’s worth reading Ethan Mollick’s context-setting exploration of how the release highlights something important happening in AI right now: experimentation with four kinds of models – AI models, models of use, business models, and mental models of the future.
Teen GenAI use
A new study from Common Sense Media and Harvard examines how young people perceive and interact with generative AI technologies, with special attention to race and ethnicity, age, gender and LGBTQ+ identity. Key findings include:
Half of the survey respondents (ages 14-22) have used generative AI at some point in their lives; however, only 4% reported being daily users.
The most commonly reported uses of generative AI were for getting information (53%) and brainstorming (51%).
Among those who have ever used generative AI, Black young people are significantly more likely than their white peers to turn to it for a variety of reasons, including to get information (72% vs 41%), brainstorm ideas (68% vs 42%), and help with schoolwork (62% vs 40%)
AI in Australian schools
A teenage boy was arrested after allegedly creating and circulating fake explicit images of around 50 female students at a private school in Victoria, Australia. The images appeared to be created using AI by combining photos of the girls' faces from social media with digitally manipulated nude bodies. However, more positive use of AI in an Australian school can be seen in this experiment teacher Lee Barett did with year 11 students using and not using AI to support essay writing.
Just ‘trained on the internet’?
That LLMs are simply ‘internet simulators’ is a common criticism of GenAI – but how true is that? “This used to be true! And is still mostly true. But it’s increasingly becoming less true,” claims Allen Pike.
Quick links
Be Internet Legends Year 6 assemblies by Google are interactive virtual events to teach children to be safer, more confident online explorers as they move from primary to secondary school. There are two new dates on offer to schools: 9 and 11 July
How effective is digital and media literacy education? Can it help children navigate a world of disinformation and deep fakes? Common Sense Media has released an independent evaluation of the Common Sense Digital Citizenship curriculum in UK primary and secondary schools, conducted by the LSE Department of Media and Communications
Kingsley St John Primary headteacher Rachel Jones claims that using technology- based interventions such as lumii has helped support pupils’ wellbeing. We’d be keen to see an independent evaluation of such interventions
Digital resilience is a bit like learning to ride a bike, involving a ‘ying and yang’ of risk and resilience, argues Simon Hammond in this overview of his research synthesising existing evidence on digital resilience
“You may not notice in everyday life when a building you walk into is constructed of words and numbers in addition to bricks and mortar.”
We’re fans of Cliff Manning’s More Than Robots newsletter and the latest edition is a corker, with a wealth of fascinating links to delve into, from a week in the life of a young person’s smartphone use and the effects of digital media on early childhood to more quirky offerings, such as this intriguing exploration of ‘addressable space’
We’re reading, listening to…
The third in a mini series from Parent Zone about the suggestion to ban smart phones for under 16s. What could a dumb phone look like? What do we do about children who are only able to do and upload their homework via a phone?
On the Ezra Klein Show, D Graham Burnett calls the attention economy an example of “human fracking”: with our attention in shorter and shorter supply, companies are going to even greater lengths to extract this precious resource from us
Pekka Mertala (University of Oulu) talks to Neil Selwyn about the history of the ‘digital natives’ concept in academic literature, why the persistence of the idea is damaging, and how we need to actively campaign against its future use.
Give it a try
Animated drawings
This tool from Meta AI Research is cute, free and very easy to use: simply upload a child’s picture and bring it to life with animation. There’s no requirement to contribute the image to a public dataset after creation and no personally identifiable information associated with the image.
Connected Learning is by Sarah Horrocks and Michelle Pauli