If artificial intelligence was the big protagonist in 2022, announcements like MusicLM’s confirm what we could already expect, on the other hand, that 2023 promises to be quite similar in this regard.. In addition, the impact of new artificial intelligence launches may not be as famous as last year, because it is becoming something normal for the general public, but with total security we will see more new such developments in 2023 than in 2022.

A lot has to do with the success of ChatGPT (if you haven’t tried it yet, we’ll tell you how), although the picture would be incomplete if it didn’t also include all the AIs that generate images from text in the equation, like DALL-E 2, with whom we became acquainted during the past year. And of course, these are just the latest developments, but they are based on a substrate that has been evolving for decades and now seems to have reached a good degree of maturity.

Every technological revolution is usually associated withappearance of new companies they devote themselves to these innovations, while the oldest ones, established in other markets, have to try to compete with “the new ones”. This brings us back to ChatGPT, or more precisely OpenAI, the company behind this AI, and how its products have forced other tech companies to step on the accelerator. MusicML is a clear example of this and how Google is in a difficult but interesting situation.

Yann LeCun mentioned in his statements about ChatGPT Google as a company closely associated with innovation in the field of artificial intelligenceand he was absolutely right, the problem is that the search engine company has not made their advancements and technology very visible in this regard. However, it seems that this will soon change, and although the most relevant announcements in this regard are expected for Google I/O 2023, we already have interesting progress with MusicLM, an artificial intelligence capable of creating a musical composition from a textual description.

The company published a paper, which you can download for free here and has also published an extensive list of examples that allow us to verify the effectiveness of MusicML. We’ve seen other AI-powered music generation services before, but from what we can see in the examples, it looks like we have the most advanced development of this type yet.

Google decided, yes, not to launch it as a service, at least for now. Trained with 280,000 hours of recordings, those responsible fear the same thing we’ve seen with other types of AI, the risk that AI might incorporate copyrighted material into its compositions, an interesting sign that the debate that has arisen around AI the intelligence of image creation exceeds many other creative fields.