OpenAI has launched an improved version of the paid ChatGPT, which now directly integrates web search, data analysis, and image generation into the chat. That means the user does not have to switch between various tools, as ChatGPT 4 brings all the features together in one platform.
At the same time, they also launched the feature GPTs, which is again for paid customers but where you can make your own trained chatbot. Thus, you can make your version of ChatGPT that is conditioned on your data, and you can decide how it should behave for yourself. However, one must be careful since Leon Furze has succeeded in getting it to hand over the data on which it is trained.
We already see a wide range of personal GPTs, but as I said, they require a subscription to ChatGPT.
The preliminary recommendations of the expert group
The preliminary recommendations from the Expert Group will shortly be published, and in connection with this, IT-vest has made a webinar. The Danish Minister for Children and Education, Mattias Tesfaye, set up the group to investigate how digital aids affect the possibility of cheating on tests at the country's schools and institutions.
It takes place on December 12 from 3-4 pm, and here, Birgitte Vedersø will review the recommendations and, at the same time, answer questions.
Some of the things she will touch on are:
The expert group recommends that work on generative language models be initiated locally in academic environments.
Digital technology literacy should be introduced throughout the education chain, and teachers should be trained as soon as possible.
The target group for the webinar is teachers and leaders in primary, secondary, and adult education.
Registration for the webinar is necessary and can be done via IT-vest's website below:
Below, we select articles that have a slightly more scientific perspective. These are articles that we read ourselves to keep ourselves updated, and at the same time, we also know that they are not so interesting for most people.
Chatbots such as ChatGPT provide better responses when emotions are involved.
We have previously written about what happens when language models become too human and simulate emotions. However, in this study, chatbots can provide better responses when receiving emotional commands. An example would be: "This is very important for my career.". These "EmotionPrompts" should increase performance by up to 11 percent.
It may be time we were emotional with the language models.
Erik David Johnson - Chief AI Officer
Ethan Mollick
Pernille Rohde
103982au.dk
Thomas Telving Filsof, forfatter og foredragsholder
Bill Gates
Morten Svith mosvstiften.dk
Aksel Herlev Rasmussen og Rasmus Nørhaven Johansen, Studenterrådet, Københavns Professionshøjskole
Corinne Becker
Brian Holm Sørensen
Wendy Castillo
Dhanshree Shripad Shenwai
