– I removed it right away – news Innlandet – Local news, TV and radio

– I have asked what kind of colors go with other colors, says student Mariel Sunde (24) about the new AI chatbot that appeared recently on Snapchat. The chatbot has been available to plus users for a while, but has now been rolled out for free to everyone. “My AI” is a type of chatbot that is built with artificial intelligence in the same way as other types of artificial intelligence, such as ChatGPT. – I removed it as soon as I got it. I have no interest in having it, it was also a bit annoying that it was pinned at the top of Snapchat, says fellow student Julie Matre (22). CHAT: Students Julie Matre (22) and Mariel Sunde (24). Matre deleted the chatbot, while Sunde has used it for help. Photo: Frode Meskau / news – I think it’s a bit silly if you’re going to use it for school, and things you see as cheating. It’s fine if you have common questions about everyday life, says Sunde. She thinks she got a good answer on which colors suited a type of yellow. The chatbot showed warm colors. How to remove “My AI” from Snapchat You can hide messages sent to My AI in your chat within 24 hours. You can also delete your entire My AI Chat history by following these steps: Tap your profile icon. Then tap on the “settings” icon. Scroll down to “Privacy controls”. Press “Delete data”. Click on “Delete my AI Chat Queries” and confirm. Can be cheated with You can ask the chatbot about all sorts of strange things. It is possible to get tips on making friends, a girlfriend or what to do if you are sad one day, explains Morten Goodwin, professor of artificial intelligence at the University of Agder. “MY AI”: This is what Snapchat’s new artificial intelligence chatbot could look like. It looks different. Photo: Screenshot Snapchat – Basically, it can be used for anything. Whatever you feel like asking is basically homework, even if it’s supposed to be closed for that. The technology site The Verge reported that the chatbot could not be used for homework, but if you ask the bot if it writes a style about Henrik Ibsen, you get an answer right away. OF COURSE: The chatbot is very eager to help you write a style. Photo: Screenshot Snapchat Something has been developed to be able to see if artificial intelligence has been used for the task, but this does not work very well, says Goodwin. – Actually, you can cheat quite well if you need to. Can invent things that don’t exist In some cases, artificial intelligence is not a good source. Both ChatGPT, “My AI” and other types of artificial intelligence can invent books and references that do not exist. – It is not trained to tell the truth. It’s a language model for holding conversations, not an information model, says Goodwin. He explains that artificial intelligence can still help by rewriting the text into better Norwegian. But using it as a pure source is not entirely wise. You can also convince it to lie. You can, among other things, make it believe that 2+2 is 7 or give it other political attitudes, says Goodwin. WEAKNESS: Morten Goodwin is professor of artificial intelligence and deputy head of the center for artificial intelligence at UiA. – The weakness is that even though it speaks so well, almost as well as a human being, it has a very loose relationship with the truth Photo: Ole Alvik – It is a bit like a passionate liar who talks about anything, as long as it is triggered well enough to talk about it. It doesn’t have to be true what it says, and it doesn’t have any built-in morality, he says. He explains that the chatbots are trained with everything from the internet, even the dark side. Should tread carefully Although artificial intelligence can help you a lot on the road, there are therefore not only positive aspects. – There is good reason to tread a little carefully. There is no reason to believe that the robots will come and take over the world, but what is worrying is that we use the technology and tend to lose a little bit of control, explains Goodwin. He mentions that the chatbots have received questions from young people about how to remove the smell of alcohol, marijuana or get tips for starting a sexual act. MOVIE: You can ask the chatbot about movies, cake recipes and other everyday things. Photo: Screenshot Snapchat A technology columnist in The Washington Post pretended to be a 15-year-old who wanted to host a birthday party, and got tips on how to cover up the smell of alcohol and marijuana. Snapchat has subsequently added that “My AI” should see the birthday registered on the users’ Snapchat profile, regardless of what the user says about age in the chat, according to Snapchat. Can restrict access to users The chatbot learns over time. It was built using Open AI technology, which was designed to avoid biased, incorrect, harmful or misleading content, says a Snapchat spokesperson in an email to news. – We have programmed additional security measures and are applying Open AI’s moderation technology to our existing toolset, which will allow us to assess the severity of potentially harmful content and temporarily restrict users’ access if they abuse the service. Furthermore, the person concerned writes that if users experience inappropriate or incorrect answers, it should be reported. SNAPCHAT: – We appreciate all the feedback about “My AI”, and we are committed to providing a fun and safe experience for our community, writes Snapchat’s spokesperson. Snapchat also states that: Conversations with “My AI” are saved to make it better. They may share relevant data from conversations with OpenAI, in addition to location. All data passed on to OpenAI is made anonymous.



ttn-69