News of the week: Good books about AI and education

In this week's newsletter, we have given a lot of space to books since a lot of interesting literature on artificial intelligence and education has been published. Therefore, at the end of our newsletter, we list some good books with a short description.

Last week, it emerged that Apple has postponed its Apple Intelligence in Europe due to challenges with DMA. Therefore, Danish Mac users must wait for several functions presented at Apple's developer conference.

Anthropic released Claude 3.5 Sonnet, outperforming competing models such as ChatGPT and Gemini. Claude 3.5 Sonnet works better in some areas than ChatGPT, but the model cannot retrieve updated information online.

Luma Labs released their Dream Machine for anyone to try generating videos in text-to-video and image-to-video. We have tested the Dream Machine and shown a few examples. Sometimes, the videos don't turn out well, and we've also rounded up a few of the weirder ones.

In a new study, the researchers examined how students use ChatGPT and other digital tools for written work. The tools increase efficiency and reduce stress but raise concerns about plagiarism and student learning abilities.

At Viden.AI, we more or less go on summer vacation, but if something wild happens, we'll probably send out a newsletter or two over the summer. Otherwise, we will return at the beginning of August.

Apple Intelligence postponed in the EU

Last week, Apple announced several new AI features for its macOS and iOS. Still, Apple Intelligence and some other features will be unavailable in Europe due to the Digital Markets Act (DMA). In an announcement from Apple, the interoperability requirements of DMA will force them to compromise on user privacy and data security.

Interoperability is the ability of the digital content or digital service to work with hardware or software that differs from that with which digital content or digital services of the same type are normally used.

It is not clear when these features will be available in the EU.

Anthropic releases advanced AI model: Claude 3.5 Sonnet

Anthropic has introduced Claude 3.5 Sonnet, a new and improved language model. According to Anthropic, the new model sets new standards for artificial intelligence and surpasses competing models in some areas, such as OpenAI's GPT-4 and Google's Gemini.

Improved understanding of nuance, humor, and complex instructions.
Advanced coding skills with the ability to write, edit, and run code.
Ability to handle complex tasks such as multi-step workflows.
Ability to analyze images and graphs as well as transcribe text from images.
Claude 3.5 Sonnet can handle large amounts of information, about 300 pages of text (200,000 tokens), whereas ChatGPT-4 is believed to handle about 120 pages of text (32,000 tokens).

One of the more interesting features is that Claude 3.5 Sonnet supports multi-language processing, making it possible to translate between languages in real-time. For example, you can ask the language model to translate from Danish to French and then to German.

Here we test the language function, where Claude 3.5 Sonnet translates a Danish text into English, German, Spanish, French and then back to Danish.

Anthropic was founded in 2021 by former employees of OpenAI and is supported, among others, by Google (approx. DKK 14 billion) and Amazon (approx. DKK 27 billion).

Luma Labs launches Dream Machine

Luma Labs, an AI startup based in San Francisco, recently launched a Dream Machine tool to generate text and image videos. Below, we have played a little with the paid version, which allows you to remove the watermark, use the tool commercially, and get ahead of the queue.

Below, we use the picture-to-video function and ask Dream Machine to make a video from P.S. Krøyer's painting “St. Hansblus on Skagen Beach.” Subsequently, we generated the sound via Elevenlabs.