Artificial intelligence, publishing and raspberry jam
Steffen Meier
November 30, 2022 is generally regarded as the "iPhone moment of AI". After ChatGPT made the software version GPT-3 available to the public free of charge on November 30, 2022, one million users worldwide signed up within five days, making it the most downloaded application in the world for some time. After only two months, ChatGPT already had 100 million active users per month. The result was, on the one hand, media hype, a lot of understandable uncertainty, doomsday fantasies, etc., and on the other, an explosion of tools and application scenarios.
What is AI and this ChatGPT anyway?
In the early days, research defined artificial intelligence (AI) as the capacity of a machine to mimic human skills- such as reasoning, learning, planning and creativity. Today, we differentiate between "strong AI", which describes an intelligent computer system that is indistinguishable from the human brain, and "weak AI”, which can perform specific tasks after appropriate training by algorithms. So far, all AI implementations known to us are examples of weak AI.
ChatGPT is a chatbot that aims to simulate conversations that resemble human dialogue. This AI-based tool has been under development since 2020 by OpenAI, a San Francisco-based company co-founded by figures such as Sam Altman and Elon Musk that focuses on artificial intelligence (AI) and machine learning (ML).
Application scenarios of artificial intelligence in the media environment
AI currently offers many tools and applications for all forms of media. In the field of text, the main points are assisting systems that support the writing process - in the form of title suggestions, translations, summaries, and writing suggestions in journalism. In marketing, it includes text suggestions and modules (for social media, for example), advertising copy and claims, as well as the optimization of keywords (in e-commerce) or hashtags.
In the field of audio, AI dutifully performs its services in the area of transcription (speech-to-text) or in the processing of audio files, ranging from the alignment of audio tracks to the elimination of delay sounds (never again “ehs” and “ohs” in a podcast!). In the area of Text2Speech, i.e. the dubbing of text content, the difference between AI and humans becomes apparent: the reading aloud of news texts, etc., still works quite well (and is now also used by many daily newspapers), while in the area of audio books with different vocal suspense arcs, AI still fails. At least Google has already set first scent traces here with " auto-narrated audiobooks".
In the field of videography, we have similar usage scenarios as audio: many assistance systems in editing and transcription. Simultaneous translation in pretty much every language on the planet may not yet work one hundred percent error-free, but it does make access easier for many - AI takes accessibility a big step further. In the area of video, it is now possible to create decent explanatory videos, tutorials, etc., with or without avatars. However, it is still a big step to the first feature film that is completely AI-generated without humans noticing.
Controversy over the legal basis for AI use
Modern AI tools can create impressive content, but they lack actual understanding of what they are creating. They merely identify and replicate patterns - nothing more, nothing less. This leads, for example, to the fact that an incorrect piece of information can become a recurring pattern if it appears in the database frequently enough.
In addition, there are still some unanswered legal questions. The question of authorship alone must be considered from several angles. Under current legislation, AI tools are not recognized as authors, as copyright is based on human authors. There is also the legal question in the context of data mining.
The first lawsuit in this matter has already been filed in the United States. The image agency Getty Images has sued Stability AI, alleging that its AI image generator has committed millions of copyright infringements by using the images for training. This could mark the beginning of a wave of lawsuits, as AI-generated images are continuously being distributed on various social networks, even with the watermark of an image agency.
Will AI eventually replace humans?
Maybe. In times of "weak AI" a simple experiment that the author of this article conducted is enough: ask e.g. ChatGPT for recipe suggestions for casseroles with any kind of vegetables, cheese - and raspberry jam. And AI will dutifully and exactly do what it was told to do. Now there may be people who will pour raspberry jam over an eggplant casserole before putting it in the oven. Generally, however, the reaction will tend to be: Yuck! And as long as AI doesn't also get the idea that raspberry jam is an inappropriate ingredient for vegetable casseroles, we don't have to worry.
PS: The impacts are getting closer. While GPT-3 still bluntly spit out above-mentioned recipes, its successor GPT-4 is smarter: "Compiling recipes for vegetable casseroles with raspberry jam can be a challenge, as raspberry jam is not usually used for savory dishes" ...
What influence does artificial intelligence have on publishing?
- AI-based technologies have the potential to revolutionize the way content is created, curated, and distributed, making the publishing process more efficient and cost-effective.
- AI can help better tailor content to users' needs and provide personalized experiences, which can lead to improved user engagement and loyalty.
- By using AI, publishers can identify relevant trends and patterns in big data to make informed decisions about content strategies, advertising efforts, and audience targeting.
- AI-powered automation of tasks such as content curation, translations, image recognition, and social media management can increase publisher productivity and save time and resources.
- AI can help improve the quality of content by automatically checking for errors, plagiarism, or copyright infringement, reducing the risk of legal issues.
- The use of AI can help publishers monetize content by optimizing relevant ad placements and pricing to maximize revenue.
By the way, the previous points are from ChatGPT