This Week's News: Pensum_GPT Challenges the Literary Canon
•
14 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.
Viden.AI has now passed 2,000 members who receive our newsletter about artificial intelligence in education. Thank you all 🎉
This week’s newsletter was slightly delayed, but is now finally ready. This week we focus on the launch by Dansklærerforeningen (the Danish Association of Teachers of Danish) of a spiteful and misogynistic language model, trained on the literary canon. There are also developments regarding the funds for technology comprehension in the Danish primary and lower secondary school (folkeskole), which were previously allocated in the government’s digitalisation strategy – suddenly they have been reallocated and now form part of the school’s time bank. At the same time, Digitaliseringsstyrelsen (the Danish Agency for Digital Government) has published a guide to the safe and secure use of artificial intelligence tools.
Among new AI tools, Adobe Premiere Pro will soon gain artificial intelligence features, while Meta has released its Llama 3 language model, which reportedly performs better in many areas than, for example, ChatGPT. Microsoft has also showcased its project: VASA-1, which is an AI-generated talking face created from an audio clip and an image.
In addition to the above, it is interesting how language teaching in the USA and Australia appears to be strongly influenced by AI tools. At the same time, people are becoming increasingly dependent on artificial intelligence to generate and access information, something Andrew J. Peterson describes as a knowledge collapse. Read his very interesting article at the end of the newsletter.
Happy reading
Pensum_GPT challenges the literary canon
The Danish advertising agency Marketsquare, launched a chatbot last week that they called Pensum_GPT, designed to challenge and reflect on the literary canon in Danish A in upper secondary school (gymnasium). Here, they trained a language model on works primarily by male authors, which resulted in a model that came across as spiteful and misogynistic.
The project has been donated to Dansklærerforeningen (the Danish Association of Teachers of Danish) with the aim of influencing a renewal of the literary canon so that it includes more female authors and thus a broader perspective. Marie Bjerre (V) has subsequently commented on X, where she calls for more female authors.
We have compiled a number of links below, and you can try Pensum_GPT on Instagram yourself (this requires an account).
New chatbot gives wild answers when you ask it about women – to highlight a problem
The current literary canon is 20 years old, and the Minister for Children and Education, Mattias Tesfaye, agrees that the curriculum needs updating.
Julie Tantholdt
Chatbot based on the Danish curriculum is both ‘spiteful’ and ‘misogynistic’ – Dansk Markedsføring
AI chatbot shocks with sexist statements
Read more here
More women on the canon list
Over the past year there has been a heated debate about the canon list for Danish in the gymnasium, as only one out of fourteen authors on the list is a woman. Here is a suggestion as to which female authors should be included. When Karen Blixen is the only woman who appears on the gymnasium’s
Lise Vandborg_0.png?itok=ZwK2bUan)
Login • Instagram
Welcome back to Instagram. Sign in to check out what your friends, family & interests have been capturing & sharing around the world.
Reallocation of DKK 160 million to the Danish primary and lower secondary school (folkeskole)
On 8 February 2024, the government wrote in its digitalisation strategy: "A total of DKK 160 million is allocated to the initiative in the period 2024-2026 for technological understanding in folkeskole teaching".
On 19 March 2024, a new agreement on the folkeskole was reached, and the folkeskole agreement parties (S, V, M, LA, KF, RV and DF) have reallocated the amount so that it has now become an annual appropriation of DKK 160 million earmarked for a quality boost to supportive teaching. The amount will in future be included in the school’s time bank, where it is decided locally whether the time bank should be used for initiatives and activities or to strengthen quality, for example through two-adult arrangements, co-teaching, etc.
Thus, all constraints that required the amount to be used for technological understanding have been removed, and it is up to individual schools locally to decide whether to spend money on it.
New agreement on funding for technological understanding in schools does not make the subject compulsory for all
DKK 160 million has been earmarked for technological understanding in the folkeskole in the government’s digitalisation plan. Exactly how the money will be used depends on the ongoing negotiations on the government’s folkeskole proposal.
Karen Ravn
New agreement on the folkeskole
Guide: Safe and secure use of artificial intelligence tools
Digitaliseringsstyrelsen (the Danish Agency for Digital Government) has just published a guide with three good tips for when you use tools with built-in artificial intelligence.
- Be aware of whether you receive incorrect or made-up responses
- Be aware of the risk of biased responses
- Do not share personal information
Guide: Safe and secure use of artificial intelligence tools
This guide is for those who want to know more about how to use tools that employ generative artificial intelligence safely and securely.
Artificial intelligence is being implemented in Adobe Premiere Pro
Adobe is actively enriching its applications with artificial intelligence, and the video-editing application Premiere Pro now gains access to Adobe Firefly. This means users can remove or add AI-generated objects to their scenes. One rather striking feature is the ability to let generative artificial intelligence extend video clips when a few extra frames are needed.
No exact release date has been given for the new features, but they are expected to launch later this year.
These capabilities also mean that, for example, media studies students with access to the Adobe suite can hardly avoid using artificial intelligence. Enforcing a ban will in any case be difficult, as it is very hard to tell when AI has been used. Conversely, we may also see a counter-movement to all these possibilities, such as a set of dogma rules where everything must again be authentic without the use of the technology.
See a demonstration of the possibilities below.
Watch this: Adobe shows how OpenAI’s Sora will change Premiere Pro and video editing forever
Premiere Pro’s new AI tools are a glimpse of the future
Mark Wilson
Meta releases Llama 3
Last week, Meta released an early version of its Llama 3 model as open source, which is on a par with ChatGPT 4, Anthropic's Claude 3, Google Gemma and Google Gemini. On some measures, the new version is also reported to perform better in benchmark tests.
As Llama 3 is free and open source, this now affects the language model market. Google’s and OpenAI’s models cost around DKK 160 per month, and Llama 3 may prove a serious competitor to them.
In addition to releasing Llama 3, Meta is developing a successor with 400 billion parameters that will be capable of delivering multimodal responses.
It is possible to download and run Llama 3 on your own computer via LM Studio.
Meta says Llama 3 beats most other models, including Gemini
The models have some pretty good general knowledge.
/cdn.vox-cdn.com/uploads/chorus_asset/file/23951355/STK043_VRG_Illo_N_Barclay_1_Meta.jpg)
Meta Is Already Training a More Powerful Successor to Llama 3
The open source Llama 3 AI model released by Meta today is just the start, according to the company’s chief AI scientist, Yann LeCun. He said a new, much larger version is in the works.
VASA-1 - AI-generated talking face
Some time ago we shared a research article in which researchers, from a single image and an audio clip, could animate it and create a short video. Microsoft has a similar technology, which they call VASA-1, and it appears highly convincing and much better than what we have previously seen.
0:00
/0:57
0:00
/0:11
The technology can be used in teaching materials to animate content from audio clips and images, but it can also be used to create misinformation. Therefore, Microsoft does not intend to release their product until they have ensured that the technology is used responsibly.
Read more below
VASA-1 - Microsoft Research
Opens in a new tab
This week's news
Science Academy: We are failing the disadvantaged if technology comprehension is not for everyone
If we want children and young people to master digital technologies, the subject of technology comprehension must be compulsory for everyone in the Danish primary and lower secondary school (folkeskole), write Sine Zambach, Jan Damsgaard and Michael E. Caspersen.
Upper secondary school (gymnasium) teachers: There is no need for a new subject on digital technology comprehension
We should drop the idea of a new subject on digital technology comprehension and instead focus on weaving digital understanding into the existing range of subjects in the upper secondary school, writes Tomas Kepler.
New professor wants more focus on the human aspect of software development
Emilia Mendes has always been interested in technology’s digital quantum leaps. She is the new professor of software technology at Aarhus Universitet (Aarhus University)’s Institut for Elektro- og Computerteknologi (Department of Electrical and Computer Engineering).
443142au.dk
English teachers’ 5 creative tips for tackling ChatGPT and AI in teaching
For three English teachers, it makes no sense to ban the use of artificial intelligence in teaching. So they have come up with different ways to handle it, both to make it easier to detect cheating and to teach students to use the tools appropriately and correctly. Read their advice here.
Emilie Palm Olesen
Upper secondary school teachers: There is no need for a new subject on digital technology comprehension
We should drop the idea of a new subject on digital technology comprehension and instead focus on weaving digital understanding into the upper secondary school’s existing range of subjects, writes Tomas Kepler.
“I think I’m brave every day”
Text_ Benjamin Mikkelsen Photo_ Private
English teacher: “It doesn’t make sense to spend time marking essays that are written by a machine”
Artificial intelligence is here to stay, and particularly for language teachers in the final years of the Danish primary and lower secondary school (folkeskole), tools such as ChatGPT bring both opportunities and challenges, especially in the run-up to the written exam. Read about the ups and downs three English teachers have had with AI so far.
Emilie Palm Olesen
Debatt: – Chat GPT – vår tids «Titanic»?
(Åpen for alle) Vi må sikre at vi utdanner studenter som kan tenke selv, skriver Margrete Dyvik Cardona (NHH).
Foto: AP / NTB
Data consortium aims for colleges to “own” their AI future
American Council on Education will lead global effort to pool data on tens of millions of students to improve learner success and collaborate on AI tools.
Forslag til deklarasjon av KI i oppgaver - KURT – Kompetansesenter for undervisning i realfag og teknologi
Dersom du har mulighet til, og ønsker, å bruke generativ kunstig intelligens (KI) i din oppgave, så skal bruken deklareres. Under følger et forslag for hvordan dette kan gjøres.
Registration of music using AI
Building LLM Applications: Serving LLMs (Part 9)
Learn Large Language Models ( LLM ) through the lens of a Retrieval Augmented Generation ( RAG ) Application.
Vipra Singh
Teaching elementary schoolchildren the rights and wrongs of AI is just as important as sex and drug education, Northeastern expert says
Ethics professor Hossein Dabbagh says it should be mandatory for children to learn about the dangers of becoming over-reliant on AI.
Patrick Daly
AI Can Transform the Classroom Just Like the Calculator
AI can better education, not threaten it, if we learn some lessons from the adoption of the calculator into the classroom
Don’t Make This Mistake When It Comes to Teaching AI Literacy
Teachers can provide the lessons without AI-powered tools.
Alyson Klein
The Opportunities and Drawbacks of AI-Powered Reading Coaches, Assistants and Tutors - EdSurge News
The edtech market is saturated with various tools designed to improve children’s literacy from e-readers to apps to digital libraries. Over the past…
Natalia Kucirkova
The future of language teaching may be at risk due to AI technology
According to an article published in The Atlantic by Louise Matsakis, we may be seeing a possible decline in the need to learn foreign languages due to the rise of artificial intelligence. In countries such as the United States and Australia, there has just been a significant decline in the number of students studying foreign languages, and the article questions the value of traditional language teaching.
It is, among other things, technologies such as deepfake videos and advanced translation programmes that are helping to change our approach to language learning. For example, Louise Matsakis notes that software such as HeyGen can create videos in which the author speaks fluent Chinese, even though they have limited experience with the language. However, it is cautioned that, although the technology can offer new conveniences, it cannot replace the deeper human experience and cultural understanding that is also part of learning another language. The article also emphasises that educational institutions should focus more on intercultural competences rather than just grammar and vocabulary.
Read the full article below:
How AI Is Reshaping Foreign-Language Education
A new generation of translation tools threatens to upend how people understand different cultures.
Damon Beres
AI and the Problem of Knowledge Collapse
Andrew J. Peterson from the University of Poitiers has examined the consequences of artificial intelligence for society’s understanding of knowledge and culture. Here he introduces the concept of knowledge collapse, which describes a situation in which reliance on artificial intelligence to generate and access information gradually diminishes human knowledge and cultural richness over time.
Read the full article below:
AI and the Problem of Knowledge Collapse
While artificial intelligence has the potential to process vast amounts of data, generate new insights, and unlock greater productivity, its widespread adoption may entail unforeseen consequences. We identify conditions under which AI, by reducing the cost of access to certain modes of knowledge, can paradoxically harm public understanding. While large language models are trained on vast amounts of diverse data, they naturally generate output towards the ‘center’ of the distribution. This is generally useful, but widespread reliance on recursive AI systems could lead to a process we define as “knowledge collapse”, and argue this could harm innovation and the richness of human understanding and culture. However, unlike AI models that cannot choose what data they are trained on, humans may strategically seek out diverse forms of knowledge if they perceive them to be worthwhile. To investigate this, we provide a simple model in which a community of learners or innovators choose to use traditional methods or to rely on a discounted AI-assisted process and identify conditions under which knowledge collapse occurs. In our default model, a 20% discount on AI-generated content generates public beliefs 2.3 times further from the truth than when there is no discount. Finally, based on the results, we consider further research directions to counteract such outcomes.
Andrew J. Peterson
To avoid AI-driven “knowledge collapse”, humans must actively preserve specialized expertise
Large language models can make information more accessible - but AI researcher Andrew J. Peterson warns of a negative long-term effect: a “knowledge collapse”.
Maximilian Schreiner
Written by
MSc in IT | Co-founder of Viden.AI | Head of IT and Communications at TietgenSkolen
