New in December 2024

The next generation of AI is on its way, and the new tools I’ve seen in the last two days alone were not even thought possible in the distant future two years ago!

Firstly, there’s ‘Runway’ (https://app.runwayml.com) , which can create absolutely realistic video clips with a simple prompt. So real that you can no longer tell if it’s AI-based. (I made a video from the blog picture, it took 10 seconds, you can see the result on the homepage).

Then there’s Fotorama.ai , which not only lets you turn selfies into passport photos, but also convert them into all kinds of situations. For example, 15 snapshots are enough for a woman to take an absolutely realistic picture of herself with a 9-month baby bump. This is even shown on the home page in the test.

For voice manipulation, there is now  Hume (https://www.hume.ai/), which can shape the voice according to character, tone and mood. So I can have any text set to music as if it were spoken by a human being, according to empathy, expression and voice colour.

Then there is Dream Machine from Lumalabs (https://lumalabs.ai/dream-machine), which can also create visualisations from ideas – either as images or videos. (I found it very convincing for images in the text, but extremely slow for videos).

However, what ‘World Labs’ is doing right now leaves you speechless. This app creates 3D worlds in real time, whether realistic, comic or manga-style. All it takes is a single image and this app creates the entire world around this image in real time, so that you can, for example, walk through the street depicted in Hopper’s famous picture „Nighthaws“ and then just stand in front of the diner or taking a look around the other shops and restaurants. Test videos can be found at https://x.com/theworldlabs . THAT really made a lasting impression on me and I wonder how this development will change our world. Ultimately, you can literally travel to any place. EVERYWHERE.

From a legal point of view, however, it is much more difficult to assume an infringement here, e.g. of an image, which was the basis. Every infringement is based on the unauthorised reproduction of a part of a copyright-protected work, for example. If the AI memorises here – and it has to, because otherwise it could not represent the image 1:1 from which the new world ‘around it’ is created – then one can at least assume copyright infringement for this part. The problem is, however, whether it is not rather a pastiche, which only sees a ‘reference’ to the original work in the form of an homage. Unfortunately, the courts have not yet dealt with these questions. The ECJ is currently still dealing with a crucial question in a Kraftwerk lawsuit: whether the use of a sound sequence in another song can still count as a ‘pastiche’. However, with the speed at which AI and its developers are moving, the courts should be very quick. Otherwise, there will soon be no original works by artists left that have not already been ‘sampled’ in some form.