What is generative AI?
Generative AI is a technology in which computer systems are trained to generate "new" content synthetically. This can be, for example, texts, images, video or audio content. To do this, the computer draws on existing training data and uses the approach of machine learning (or deep learning). AI generators create new interfaces for users and thus new possibilities, but also a fundamentally higher awareness of the topic of "artificial intelligence".
The chatbot ChatGPT is one of the best-known examples of text-generating AIs to date. The bot conducts human-like conversations and answers questions in natural language. To do this, it draws on an extensive database of texts from the Internet and other publicly accessible sources. In the teaching material "How reliable is ChatGPT" we offer more information on this topic.
The chatbot My AI on Snapchat is based on ChatGPT and is used for chatting and simple research. More information about My AI and what children and young people should be aware of when using it can be found in our news article.
Google's chatbot Bard can also create texts and conduct dialogs in human-like language. As with ChatGPT, parts of the answers can also be untrue.
Text-to-image generators can be used to create or edit images, graphics or illustrations from text descriptions. Midjourney is one of the best-known tools in this area. Other examples of image-generating offerings would include Stable Diffusion, DALL-E-2 from OpenAI, Canva, or Adobe Firefly. However, there are other types of image generators that do not rely on text input.
The use of such synthetic image content is diverse. Image generators are used, for example, in online gaming to create avatars, the game environment, or even objects. AI is also getting better at generating real-looking images of non-existent people . Against this background, the phenomenon of "virtual influencers" has been observed sporadically in social networks for several years.
Well-known examples are Lil Miquela, an influencer with over 2.8 million followers on Instagram, or Rozy Oh, with about 154 thousand followers. Both characters are AI-driven, generate high reach and have been part of advertising campaigns for internationally known brands. There are even real-life influencers like Caryn Marjorie, who are taking advantage of the AI trend and transforming themselves into a virtual version in order to implement new business models.
AI technologies, including those that generate new content, are neither "good" nor "evil" in themselves. They are tools that people make, deploy, use, and shape through user interaction. Just as with other technical achievements, what you do with them matters. The responsibility therefore lies with people - especially the tech companies that develop such systems and make them available to the public.
It is not unusual for technical progress to be polarizing. The public debate around the topic of "AI" also often moves in a field of tension between two extreme camps. On the one hand, the use of AI is feared to result in a loss of control, with machines making us humans superfluous or even "wiping us out." On the other, driven primarily by tech companies, AI is touted as a solution to every problem. Also known as "solutionism."
But both extremes should be viewed with distance. Because exaggerating AI technologies as "super intelligence" creates a false impression of what AI systems can actually do. The belief that we are dealing with a "higher power" that is beyond our human understanding can lead to passivity in some people. If the worries and fears of the skeptics are too great, this can result in neither addressing the potential nor the real dangers posed by AI. On the other hand, an exaggerated fascination with AI on the part of proponents can lead to the technology being uncritically understood as a "higher truth." In this context, it would be important to ask which capabilities of an AI would be desirable for us as individuals and society and which would not.
Mystification of AI technology makes it difficult for society at large to address the complex of issues and the "real" dangers posed by AI.