Are there any proposals or research on making bitcoin's PoW useful for real-world computation?

Bitcoin’s proof of work (PoW) mechanism secures the network by requiring miners to perform massive amounts of computations to find a valid nonce. However, these computations—though essential for Bitcoin’s security—are purely wasteful in terms of real-world utility, as they serve no purpose beyond reaching the target hash.

Are there any proposals, research papers, or experimental ideas exploring ways to make PoW contribute to useful computational tasks while still maintaining Bitcoin’s security?

For example, could mining be designed to also perform:

• Scientific simulations
• Protein folding for medical research
• AI model training
• Other forms of distributed computing

If such ideas exist, what are the technical and economic challenges preventing their implementation? Would altering PoW in this way compromise Bitcoin’s decentralization or security?

Curious to hear thoughts and any research on this topic!