Will an open-source text-to-music LLM model be able to create an entire album as convincing as a human by end of 2025?
12
61
270
2026
28%
chance

Criteria is a work in progress but in essence, if someone can simply prompt “Make the next Drake album” and it makes an album as immersive and good as one of Drakes real albums then obviously it’s YES. Since the words immersive and good are subjective, all arguments for it should be considered equally and I will treat each with equal care. Look for more LLM prediction timelines from me. I’m thinking of betting on this. I’m convinced by 2035 text-to-anything will be possible.

Get Ṁ600 play money
Sort by:

Does this have to be as good as humans across the board, all genres, or any genre? I expect some genres will be way easier to generate than others.

@agentydragon My hypothesis is that if it can do any genre, it can do all, so based on that assumption if you show me any genre can be a YES or any genre can be a NO, that's good enough but naturally the conversation would continue showing multiple LLMs showing NO for YES in certain areas and vice versa. From general consensus this will all be measured on a final overrall YES/NO determination across the board.

More related questions