This Week's News: AI Tools Support Students with Cognitive Challenges
•
12 min read
This article has been machine-translated into English. Therefore, there may be nuances or errors in the content. The Danish version is always up-to-date and accurate. Read the original Danish article here.
Here is an overview of some of the most important news and perspectives on artificial intelligence in teaching.
First and foremost, a new study from Lunds Universitet (Lund University) shows that pupils with cognitive challenges such as difficulties with concentration and impulsivity benefit particularly from AI tools in their schoolwork. These tools help them complete tasks that they would otherwise have struggled with.
At the same time, a survey from Voxmeter shows that 7 out of 10 young Danes consider AI skills important for the future, but many lack knowledge about how best to make use of the technology. In this connection, Google, together with tech expert David Guldager, has visited upper secondary schools (gymnasiums) to teach students to use AI effectively.
We have previously written about DR’s article, which has to a large extent focused attention on the high climate costs of AI technology. However, a number of experts cast doubt on these figures, and the whole question of AI models’ energy consumption is highly complex.
In an opinion piece in Politiken, the editors behind Denmark’s national encyclopaedia, Lex.dk, underline the importance of ensuring access to reliable knowledge in Danish in a digital age marked by misinformation.
Upper secondary school teachers in pædagogikum (the teacher training programme for the gymnasium) now have the opportunity to write about and use generative AI in the theoretical assignment.
A study from Københavns Universitet (University of Copenhagen) shows that ChatGPT performs well in Danish but still has challenges with linguistic nuances.
In this week’s recommendation we link to Ethan Mollick’s new article, where he writes about how AI can already do most homework and tests for students – the so-called “Homework Apocalypse”.
In our article series on faglig sorg I am running a little behind. The fourth article was meant to be published last Thursday, but at the last moment a rather interesting perspective emerged that I had not considered. Therefore, it will be published this coming Thursday.
Enjoy the newsletter!
AI tools are particularly popular among students with cognitive challenges
In a new article published in the journal Frontiers in Artificial Intelligence, researchers from Lunds Universitet investigate the relationship between cognitive challenges and the use of AI tools for schoolwork. They studied 750 students aged 12–19.
The study shows that students with cognitive challenges such as difficulties with concentration, impulsivity and planning experience significant benefits from AI tools like ChatGPT in their schoolwork. These tools particularly help students complete tasks that they would otherwise have found difficult to manage on their own.
However, the researchers behind the study emphasise the need to develop clear guidelines for the use of AI in school, as there is uncertainty about whether these technologies might hinder the development of students’ cognitive abilities. They call for teachers to be better equipped to guide students in a balanced use of AI tools, so that these can support learning without undermining students’ independent cognitive development.
Frontiers | Adolescents’ Use and Perceived Usefulness of Generative AI for Schoolwork: Exploring Their Relationships with Executive Functioning and Academic Achievement
Johan Klarin
Upper secondary students learn to harness artificial intelligence with help from Google
Voxmeter has conducted a survey for Google showing that 7 out of 10 young Danes consider the ability to use artificial intelligence (AI) an important future competence. However, many lack knowledge about how best to make use of the technology. To meet this need, Google’s Danish Head of Communications, Jesper Vangkilde, and tech expert David Guldager are visiting upper secondary schools across the country to teach students to use AI effectively. The students gain hands-on experience in writing effective prompts for AI systems and are introduced to various AI tools.
It is no longer enough for upper secondary students only to be able to ‘google’ – they also need to be able to ‘prompt’, says Jesper Vangkilde, Head of Communications for Google in Denmark, to Sjællandske Nyheder.
The event also includes discussions about the ethical challenges and potential downsides of AI, which supports students’ understanding of the technology’s complexity.
💡
At Viden.AI, we see both opportunities and challenges in these talks. On the one hand, upper secondary students gain insight into the latest knowledge on generative artificial intelligence and are inspired by ways to use the technology innovatively. On the other hand, there may be commercial interests that could potentially influence young people’s views of AI and the tech industry in a way that is advantageous to Google. Although we have not ourselves attended these sessions, we find it important that schools strive for a nuanced and balanced approach when generative AI is discussed in teaching.
Read more about the events below:
- It’s a bit scary, says an upper secondary student about artificial intelligence
Most young people think artificial intelligence is the future, but many do not know how to use it. Google Denmark wants to teach them.
August Tærsker Høgh
Upper secondary students in Kolding get help from Google to use artificial intelligence
A new survey shows that seven out of ten young people consider the ability to master the use of artificial intelligence to be very important. Now some of them are getting help from top-level experts.
Emilie Broe Pedersen empedfyens.dk
The jobs of the future use AI, but young people are not ready. Yet
🚀 Young Danes are facing an AI revolution! But are they ready? 🤔 71% believe that AI skills are the key to the future, but many lack instruction. Read about how we can prepare them for the new technological era! 🌐
Kasper Haagen Skovse
AI technology’s high climate costs worry experts
The use of AI chatbots such as ChatGPT and similar technologies has a significant climate footprint. Data centres that handle these AI tasks require enormous amounts of energy and water for both operation and cooling. In an article on DR, calculations are cited showing that a single query to a chatbot can use as much energy as 40 mobile phone charges.
Although the full climate footprint is still unclear, projections indicate that data centres’ energy consumption could reach the level of Japan’s entire electricity use by 2026. The big tech companies such as Microsoft and Google have already postponed their climate targets because AI development is challenging their sustainability initiatives. There is also growing attention to the enormous water consumption, which creates local challenges, especially in areas with water scarcity.
Several experts cast doubt on these data
However, through our network at Viden.AI we hear considerable doubt about these data, and it is clear that the question of AI models’ energy consumption is highly complex.
For example, as pointed out in Jesper Andersen’s thread on LinkedIn, we simply do not have precise enough data to conclude anything definitive about ChatGPT’s energy consumption, and the calculations that exist may not be directly comparable with real-world scenarios.
At the same time, there is speculation as to whether the figures cited are based on experiments with GPT-J on outdated consumer hardware, which would be a poor proxy for assessing the energy consumption of modern, commercial AI models such as those from OpenAI. Without open and precise data from actors such as OpenAI, we are forced to rely on estimates and proxies that may be misleading.
For example, training Meta’s LLaMA3 emits the equivalent of 7.2 tonnes of CO2 per year, which corresponds to 405 Danes’ annual CO2 consumption (the training emits 1,900 tonnes of CO2, and each Dane uses on average 4.69 tonnes of CO2 per year in 2020).
On the basis of the above, we are very unsure whether DR’s figures are correct. We lack both the basis for their calculations and concrete information from some of the big tech giants before we can carry out these calculations.
We recommend reading the comments thread on Jesper Andersen’s post on LinkedIn, where these figures are hotly debated.
Jesper Andersen on LinkedIn: Chatbots harm the climate: Huge data centres could within a few years swallow Japan’s entire… | 54 comments
As far as I recall, it was Adam Hede who knew of some criticism of that study from Københavns Universitet (University of Copenhagen) last winter, which claimed to have calculated that 10… | 54 comments on LinkedIn
Jesper Andersen
llama3/MODEL_CARD.md at main · meta-llama/llama3
The official Meta Llama 3 GitHub site. Contribute to meta-llama/llama3 development by creating an account on GitHub.
meta-llamaChatbots harm the climate: Huge data centres could, within a few years, swallow Japan’s entire electricity consumption
Companies such as Microsoft and Google have postponed their climate targets because of AI. Danish experts want to make consumers aware of the footprint.
Five years ago I wrote about the internet’s enormous CO2 emissions. Then AI came along and changed everything
Artificial intelligence will make the world’s CO2 emissions rise. But a young Danish student may have found (part of) the solution.
Theis Ehler Molin
Shared knowledge must be safeguarded in a digital age
The leaders behind Lex – Denmark’s national encyclopaedia – stress in an opinion piece in Politiken the importance of ensuring access to reliable, research-based knowledge in Danish at a time when the internet is characterised by misinformation and manipulation.
They emphasise that democracy and public debate are under threat when shared knowledge is replaced by echo chambers and algorithm-driven distortions of fact. They point out that new AI technologies amplify these problems, as they can easily produce content that resembles the truth but often is not. Therefore, they argue that, as a society, we must reinvent the encyclopaedia as a modern source of enlightenment that can underpin a shared democratic conversation.
In Denmark, the Lex.dk initiative has proved a success, with millions of entries and a strong research foundation. To secure a shared reality in the future, they call for collaboration between politicians, researchers and the education sector to prioritise and protect this common knowledge resource.
Leaders of Lex: The internet is critical infrastructure and there is a need for reliable knowledge in Danish
Artificial intelligence and the internet giants are pushing at democracy and eroding our trust in the web. Without a common starting point, representative government dissolves into ignorance and echo chambers. It is crucially necessary that, as a society, we take control and ensure free and equal access to our shared knowledge.
Generative AI will be a theme in theoretical pædagogikum 2024–25
The themes for this year’s pædagogikum candidates have now been published, and one of the new themes is "Learning through generative artificial intelligence". This gives candidates the opportunity to explore and trial generative AI in an educational context. As part of the assignment, candidates can implement AI in their teaching and collect empirical data, which they must subsequently analyse theoretically.
In the outline for Theme 2 there are several references to Viden.AI, which we are very honoured by ☺️
In addition to the AI theme, Theme 1 focuses on disciplinary reading, and, as something new, an elective theme has also been added. At the same time, it is now also possible to use generative AI in the submission itself, as long as good academic practice is followed.
You can read about the three themes here:
The examination
Read more about the Pædagogikum examination on this page.
ChatGPT performs well in Danish, but nuance is lost
A new study by language researchers at Københavns Universitet and Det Danske Sprog- og Litteraturselskab (The Society for Danish Language and Literature) shows that ChatGPT performs well in Danish but still struggles to understand linguistic nuances such as ambiguous words and figurative language. This is partly because ChatGPT is primarily trained on English texts and then translates into Danish, which can result in linguistic errors and biases. Although ChatGPT 4.0 handles many tasks well, challenges remain, especially when it comes to abstract and culturally specific language. The study points to the need for further training of the models on Danish material to ensure they can handle complex linguistic tasks accurately.
ChatGPT performs well in Danish, but linguistic nuance is lost
ChatGPT has impressed with its ability to understand and produce Danish texts. But there is still some way to go before it can handle, for example, ambiguous words and figurative language correctly, conclude language researchers in a new study. One of the problems is that ChatGPT and other chatbots are not trained sufficiently with Danish texts.
View of: Evaluation of language understanding in Danish language models – based on semantic dictionaries
This week’s recommendation
Homework Apocalypse: AI’s Influence on Homework
In a new article, Ethan Mollick describes how the so-called ‘Homework Apocalypse’ has become a reality, with AI able to complete virtually all types of homework and tests for students. The spread of AI has been astonishingly rapid, and studies show that a large proportion of both university and school students already use AI to support their schoolwork. Although many teachers still regard AI use as cheating, there are a number of illusions that hinder an effective response to the problem.
The first illusion, the ‘Detection Illusion’, is teachers’ mistaken belief that they can easily spot AI-generated work, which in reality has become very difficult. The second illusion, ‘Illusory Knowledge’, affects students who fail to realise that AI assistance can actually weaken their learning, even if it earns them higher marks on homework.
Mollick argues that AI should be integrated into teaching in ways that promote thinking rather than merely serving as a shortcut for students. Research shows that AI can be beneficial if used wisely, for example as a teaching tool that pushes students to reflect on their own knowledge. The teacher’s role is therefore central in guiding the use of AI in education, so that it becomes a means to deeper learning rather than simply a substitute for students’ own effort.
Read Ethan Mollick’s full post here:
Post-apocalyptic education
What comes after the Homework Apocalypse
Ethan Mollick
The Homework Apocalypse
Fall is going to be very different this year. Educators need to be ready.
Ethan Mollick
This week’s other news
Norwegian press takes on misinformation and fake news
The two major media organisations NRK and VG are leading the establishment of Center for Kildekritik (Centre for Source Criticism).
Lars Lindevall larslhelsingordagblad.dk
OpenAI Aims to Release New AI Model, ‘Strawberry,’ in Fall
OpenAI aims to release its next-level artificial intelligence (AI) product, called “Strawberry,” in the fall.
PYMNTS
Microsoft inks $10 billion green energy deal as power-hungry AI forces its hand to meet emissions commitments
Driven by AI power demand rising at an “exponential” rate, Microsoft and Brookfield ink largest private energy deal yet.
Researcher: Politicians, industry and the education sector are being dazzled by ‘AI bullshit’ | DataTech (PRO)
AI researcher Denise Utochkin has harsh words for universities and politicians who are rushing to hastily integrate dubious AI chatbots.
Peter Christian Bech-Nielsen
Written by
MSc in IT | Co-founder of Viden.AI | Head of IT and Communications at TietgenSkolen
