It is already foreseeable that in 2025 the focus will be on more practical applications for AI, but the technology behind it will also continue to develop. Although the ways in which AI can be further developed are far too diverse for an exact prediction, we still dare to take a look into the crystal ball for a few selected examples.
1. AI for better gameplay
An AI that plays Dungeons and Dragons with a user has long been a reality, but so far there have been comparatively few applications for (modern) AI in video games. However, the potential behind it is enormous, as the community itself demonstrates:
There is already about oneMod for Skyrim, through which you can converse normally with NPCs. They even notice what is happening around them. Thinking a little further, there is a dynamic narrator à laclose, or perhaps a game director who oversees the gameplay as indynamically controlled. AI can offer even more possibilities here.
Indie developers in particular could incorporate such relatively new technologies into projects faster and better than large studios, which often work on a project for several years. That's why we don't yet see a AAA game with AI as a core element in 2025.
At the start of development of games that will be released in 2025, the big boom triggered by ChatGPT two years ago was simply not yet foreseeable. However, we are very optimistic about exciting indie projects.
About the author: Tim is currently writing his doctoral thesis on generative artificial intelligence (KI, draw me a picture of a cat riding a unicycle!
), after his passion for computer games lured him into computer science.
He is enthusiastic about explaining artificial intelligence, much to the chagrin of those around him.
2. Triumph of the video generators
In 2024, the first video generators were introduced that actually delivered usable results: high resolution and quality, and increasingly plausible videos. What has worked well for a long time in pictures has taken significantly longer with video generators because not only the appearance of our world but also the behavior has to be learned. Like how a ball bounces or a piece of clothing wrinkles while running.
While creating videos is more of a cool gimmick, editing existing videos has dozens of uses, for example in the advertising industry. The problem is that there are simply no large data sets with examples (before / instruction / after) on which an AI can be trained.
The first steps to overcome this are already about meta with hercurrent approachWith a few tricks, such training data can be artificially generated.
After the first steps, in 2025 we expect a big explosion of more and more high-quality videos and also more and more useful applications such as text-based editing. This could create completely new opportunities this year, especially for smaller filmmakers.
3. Smart (and useful) AI assistants with good connections
Previous AI assistants are often an extra application: We need something, type it into an input field, and get a solution that we can then send out.
However, smartphone manufacturers like Google and Apple in particular want to move away from this path and towards onepersonal assistant
, who can actively do all sorts of things. For example, book a hairdressing appointment at our regular shop so that it fits into our calendar.
For the next year, now that the technology behind many text-based assistants is already usable for many practical purposes and (we deliberately put this in quotation marks)mature
a much stronger and seamless integration is obvious.Okay Google, order me something to eat
could become a reality this year.
By the way: By the way, you can see our expert Tim Elsner in the following Tech Talk as he talks to Jan and Maxe about AI myths and thembusts
:
4. The hunt for the holy grail
The ultimate goal of development, a strong AI that can understand and do all the tasks that humans can (also often called AGI), will probably not be around in 2025. However, we are currently seeing what some first steps in this direction could look like.
As a small example: Current AI models such as ChatGPT have long relied on factsto explain in more detail. Similar to a person who solves a math problem piece by piece (5 + 3+8 = 8 + 8 = 16
), an AI assistant like ChatGPT makes fewer mistakes when solving a task in more detail.
This is often done by explaining the task in more detail, followed by a detailed solution in many steps. ChatGPT version o1, for example, does exactly that. However, formulating it in human language is of course not ideal - and this is exactly where innovations come into play that give the AI model a freely usable spacethink
give, thenot in scraps of wordsis bound.
Here too, progress can be expected, but no real miracles. Ultimately, the basic principle remains to predict word-scrap-by-word-scrap, which many researchers consider to be a dead end (see X-Post below). For a real revolution, completely new approaches would probably be needed, after all, almost all current AI assistants are essentially based on oneWork from 2017.
Recommended editorial content
At this point you will find external content from Twitter that complements the article.
You can display it and hide it again with one click.
I agree to content from Twitter being shown to me.
Personal data may be transmitted to third-party platforms. More about this in our.
Link toTwitter content
What AI itself says about it
Of course, we also asked an AI chatbot how AI itself sees developments for 2025. There were a number of points that the AI (we think is plausible) sees as important. But the focus is a little more practical, almost more boring, than in our prediction. The statements can be roughly summarized as follows:
- Further rise of generative AI: This includes both AI chatbots and models that create images or videos. Practical applications such as product previews are mentioned, as is generated music according to user requirements.
- More focus on ethics and laws: Stronger regulation and more transparency required on the part of providers as well as a labeling requirement for AI-generated content.
- Greater automation in the workplace
- More focus on sustainability: The huge energy requirement to train and operate an AI model is already the focus of hardware manufacturers and researchers alike.
Regardless of whether in small or large steps, we expect some exciting developments to happen in 2025 that we cannot even expect right now. Above all, we expect AI to become more and more present in our daily lives, for example through better integration into smartphones.
Are you curious about how else AI in the gaming industry can change our favorite hobby? Then we can only recommend our GameStar Talk on the topic:
GameStar Talk: Is AI destroying creativity in gaming?
How have you perceived the development of artificial intelligence over the last year? Did you find the developments rather manageable, or are you (like the author) completely fascinated by the speed at which new developments emerge? What is most interesting about the developments? Feel free to tell us your opinion in the comments!
