This Week's News: Denmark in a Technological Slumber
•
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 another look at the latest in artificial intelligence and the education sector. Many are currently on the autumn break, but there are still a couple of news items we think are interesting, and the tech world is not standing still – on the contrary, our new Minister for Digitalisation tells Børsen that we need to accelerate the technological express train so that Denmark and Europe do not fall behind.
Meanwhile, our ways of writing and communicating are changing as generative AI intervenes in our text culture. That is why a new basic research centre at Aarhus Universitet (Aarhus University) will investigate this impact – both the opportunities and the challenges.
Another news item that has caught our interest is that the videoconferencing system Zoom is working on having our digital avatar attend online meetings or reply to an email. That saves time that can be used for other things.
In our newsletters we have a strong focus on the environmental impact of artificial intelligence, especially because it is a very energy‑intensive technology. But perhaps it does not have to be that way, and a new scientific article explains how, in language models, floating‑point multiplications can be replaced with simple additions. Their conclusion is that energy use can be reduced by up to 95%. In the meantime, the tech giants focus on nuclear power to supply electricity to run their data centres.
Additionally, you can read about:
- The podcast “Læreværelset”, which discusses AI in education
- The Nobel Prize in Physics, awarded to AI pioneers Geoffrey Hinton and John Hopfield
- Jeffrey R. Young discussing how smart glasses could change teaching
Happy reading and enjoy the autumn break – it’s beautiful out in the real world, so enjoy autumn’s vibrant colours.
Denmark must strengthen AI development to remain competitive
Denmark’s new Minister for Digitalisation, Caroline Stage (M), underlines the importance of boarding the technological “express train” within artificial intelligence and digitalisation. She believes that Denmark and the rest of Europe need to wake up from their “Sleeping Beauty slumber” and take AI seriously in order to compete globally. To advance this, the minister will launch the new Dansk Data Space Forum, which will promote data sharing between businesses, researchers and public authorities. In addition, Denmark will join the European Gaia-X-organisation to increase data sharing.
The minister sees AI as a decisive factor for Denmark’s future competitiveness and points out that both the private and public sectors must participate in development if Denmark is to keep pace with global players. This may also affect how technology is integrated into the education system.
New minister: Europe needs to wake up from its Sleeping Beauty slumber
Minister for Digitalisation Caroline Stage (M) signs Denmark up for European data sharing and establishes a new Danish data forum “It is absolutely essential that we get a move on, both when it comes to A
Ulrik Horn, Nicolas Stig Nielsen
New basic research centre to study artificial intelligence and text culture
Aarhus Universitet is on its way to establishing TEXT: Center for Contemporary Cultures of Text, a new basic research centre supported by Danmarks Grundforskningsfond (the Danish National Research Foundation). The centre, which is expected to receive 38 million kroner in funding over the next six years, will examine how artificial intelligence, especially generative AI, affects our text culture. Professor Mads Rosendahl Thomsen will lead the centre, which will bring together researchers from disciplines such as computational linguistics, literary studies and education research.
The research will focus on the opportunities and challenges created by generative AI, including efficient information search and the risk of misinformation. In addition, the centre will work to develop tools to minimise these risks.
TEXT will promote interdisciplinary collaboration between researchers both in Denmark and internationally to understand the relationship between AI and text, including how language models may affect creativity and professional writing. Mads Rosendahl Thomsen emphasises that AI has created a marked before‑and‑after in our text culture, which the centre will explore in depth.
New basic research centre to study artificial intelligence and text culture | Aarhus Universitet: Arts
Artificial intelligence is fundamentally changing how we communicate. Therefore, a potential new basic research centre at Aarhus Universitet named TEXT: Center For Contemporary Cultures of Text will investigate how artificial intelligence affects our text culture, and what we can do about it.
Zoom introduces AI avatars that can speak for you
The videoconferencing platform Zoom is taking another step towards a future where you can send an AI avatar to meetings instead of attending yourself. As part of Zoom’s expansion with new AI features, users will soon be able to create a digital avatar that both looks and sounds like them. To create an avatar, the user records a video that Zoom’s AI uses to generate the avatar. You then write the message you want the avatar to deliver and let the AI speak on your behalf.
This feature will work via Zoom’s Clips feature, where users can send short video updates to their colleagues. Zoom is aware of potential risks such as deepfakes and has implemented advanced authentication and watermarking technology to ensure responsible use.
The AI avatars will launch at the start of next year as part of Zoom’s AI Companion add‑on, which will cost 12 dollars per month. If you use Zoom without this add‑on, you can still use preset AI avatars and voices, but you cannot create your own.
Zoom’s CEO, Eric Yuan, has previously talked about a future where AI can take meetings and answer emails on users’ behalf, freeing up time for other tasks.
Zoom will let AI avatars talk to your team for you
When will AI avatars get to attend meetings for you?
New algorithm dramatically reduces energy use in large language models
In the article “Addition Is All You Need for Energy‑Efficient Language Models”, Hongyin Luo and Wei Sun present a new algorithm called L‑Mul that significantly reduces the energy use of large language models. The algorithm can save up to 95% of energy during both training and inference by replacing floating‑point multiplications with simple additions.
The researchers behind the method point out that it can be applied widely in AI models and could potentially improve energy efficiency in data centres, where the energy demands for AI processing are high. The next steps will be to develop dedicated hardware that can further optimise L‑Mul and to train AI models specifically designed to use this new algorithm.
Addition is All You Need for Energy-efficient Language Models
Large neural networks spend most computation on floating point tensor multiplications. In this work, we find that a floating point multiplier can be approximated by one integer adder with high precision. We propose the linear-complexity multiplication L-Mul algorithm that approximates floating point number multiplication with integer addition operations. The new algorithm costs significantly less computation resource than 8-bit floating point multiplication but achieves higher precision. Compared to 8-bit floating point multiplications, the proposed method achieves higher precision but consumes significantly less bit-level computation. Since multiplying floating point numbers requires substantially higher energy compared to integer addition operations, applying the L-Mul operation in tensor processing hardware can potentially reduce 95% energy cost by element-wise floating point tensor multiplications and 80% energy cost of dot products. We calculated the theoretical error expectation of L-Mul, and evaluated the algorithm on a wide range of textual, visual, and symbolic tasks, including natural language understanding, structural reasoning, mathematics, and commonsense question answering. Our numerical analysis experiments agree with the theoretical error estimation, which indicates that L-Mul with 4-bit mantissa achieves comparable precision as float8_e4m3 multiplications, and L-Mul with 3-bit mantissa outperforms float8_e5m2. Evaluation results on popular benchmarks show that directly applying L-Mul to the attention mechanism is almost lossless. We further show that replacing all floating point multiplications with 3-bit mantissa L-Mul in a transformer model achieves equivalent precision as using float8_e4m3 as accumulation precision in both fine-tuning and inference.
Hongyin Luo
Tech giants bet on nuclear power to reduce their carbon footprint
Large tech companies such as Amazon, Microsoft and Google are increasingly turning to nuclear power to supply their energy-hungry data centres. Microsoft and Amazon have already struck deals with US nuclear power plants, while Google is considering following suit. Nuclear power, which provides stable, round-the-clock energy, is a good fit for data centres that have a constant demand for electricity.
The tech companies see nuclear power as a way to reduce their CO2 emissions, as their rising electricity consumption, driven especially by artificial intelligence, makes it harder for them to meet their climate targets.
Big Tech has cozied up to nuclear energy
Major tech companies want to run data centers on nuclear energy.
The podcast "Læreværelset" discusses AI in teaching: Opportunities and pitfalls
In the latest episode of Læreværelset, Morten Gundel talks with upper secondary school (gymnasium) teachers Laura, Sune and Simon about their experiences of using artificial intelligence in teaching. The starting point is an article from Magasinet Gymnasieforskning, which covers experiences from six East Jutland upper secondary schools that have explored both the opportunities and pitfalls of AI. The teachers share their own experiences and reflections, inspired by the article.
The mathematics teacher, Sune, highlights how high-attaining students can use AI to deepen their learning, while lower-attaining students often lack the necessary skills to get the full benefit from AI. Laura, who teaches Danish and English, observes that students sometimes use AI as a shortcut instead of developing their own language skills, which can hinder their learning. She has therefore introduced draft notebooks to gain better insight into students’ own abilities without help from AI.
The three upper secondary school teachers say that AI can be a useful tool in teaching, but that it also creates challenges, especially when it comes to maintaining students’ critical thinking and engagement. The teachers also discuss how AI can affect the assessment of students’ performance, and whether teaching should in future focus more on the process than on the final product.
Listen to Lærerværelset here:
LÆRERVÆRELSET READS: How can we use artificial intelligence in teaching without falling into the traps? - Lærerværelset
Together with three upper secondary school (gymnasium) teachers from Københavns åbne Gymnasium (Copenhagen Open Gymnasium), host Morten Gundel explores how we can use artificial intelligence in teaching in a way that reaps the rewards and avoids the pitfalls. We take our starting point in a current article f…
Nobel Prize awarded to AI pioneers
Geoffrey Hinton, often called “The Godfather of AI”, and John Hopfield, both distinguished researchers, have been awarded the Nobel Prize in Physics for their groundbreaking work on artificial neural networks, which laid the foundations for today’s AI technologies. Since the 1980s, their discoveries have enabled the development of machine learning in which neural networks mimic the brain’s way of forming connections. This has revolutionised areas such as image recognition and language models.
Hinton, who developed the Boltzmann machine, and Hopfield, behind the Hopfield network, have had an enormous influence on AI’s advance. Hinton has, however, expressed concern about AI’s potential dangers and left his position at Google in 2023 in order to warn about the very technology he helped to develop.
Scientists who built ‘foundation’ for AI awarded Nobel Prize
One of the winners reportedly “regrets” his life’s work, citing risks posed by AI.
At the same time, Demis Hassabis and his colleague John Jumper from Google DeepMind received the chemistry prize for their breakthrough with AlphaFold, a program that predicts protein structure, one of science’s great challenges.
Google DeepMind scientists and biochemist win Nobel chemistry prize
Demis Hassabis and John Jumper of DeepMind and computational biologist David Baker share prize for protein structure breakthroughs
Hannah Devlin
Smart glasses and AI: Jeffrey R. Young’s thoughts on the classrooms of the future
In an article by Jeffrey R. Young from EdSurge, he discusses what smart glasses could mean for education. Meta recently unveiled its prototype Orion glasses, which combine the physical world with digital holograms and, according to Young, could change how we use technology in classrooms in the future.
Young points out that there is already a trend towards banning mobile phones in schools because they often disrupt students’ concentration. The introduction of smart glasses, which may be even harder to regulate, raises entirely new challenges. There are already reports of students using smart glasses to cheat, with AI and built-in cameras helping them to find answers to exam questions.
Although the technology is not yet ready for the public, Young also sees potential for positive uses. He highlights that smart glasses combined with AI could create new learning opportunities. The glasses could provide real-time feedback by using AI to analyse and respond to tasks such as laboratory exercises. Augmented reality could also create more interactive and collaborative learning environments, which could improve students’ engagement and learning outcomes.
Young emphasises, however, that educational institutions will need to find ways to navigate between technological advances and the ethical challenges that come with them, especially regarding academic integrity and potential cheating.
If Smart Glasses Are Coming, What Will That Mean for Classrooms? - EdSurge News
SMART GLASSES IN CLASS?: Meta recently announced its bet that smart glasses are the next big thing that will replace smartphones in a few years. With a…
Jeffrey R. Young
Man suspected of using smart glasses to cheat on univ. entrance exam
Police plan to refer to prosecutors an 18-year-old man for allegedly using camera-equipped smart glasses to cheat on an entrance exam at Tokyo’s prestigious Waseda University, according to investigative sources.
KYODO NEWS
Other news this week
GPTs and Hallucination: Why do large language models hallucinate?: Queue: Vol 22, No 4
The findings in this experiment support the hypothesis that GPTs based on LLMs perform
well on prompts that are more popular and have reached a general consensus yet struggle
on controversial topics or topics with limited data. The variability in the…
Authors Info & Claims
In a digital world, we need teachers more than ever before
In an increasingly digital society, teachers are needed more than ever to strengthen pupils’ creativity and critical thinking, write DLF (the Danish Union of Teachers), GL (the upper secondary teachers’ union), FSL (the Independent Schools Teachers’ Union) and Oxfam Danmark (Oxfam Denmark) in this joint opinion piece to mark Teachers’ Day.
The pupils can already do it: now the teachers need to get on board too
Pupils were asked how they use AI. The answers took some of their teachers by surprise.
Ida Smith
ChatGPT has become the ‘best teammate’ to these Sydney university students – but is there a limit?
Universities are pivoting to incorporate generative AI into academic life – without the plagiarism – in the belief it will become an intrinsic part of life and work
Caitlin Cassidy
Written by
MSc in IT | Co-founder of Viden.AI | Head of IT and Communications at TietgenSkolen
