The Beatles' Now and Then, created using AI and an original 1970s lo-fi demo, won Best Rock Performance at the Grammys; McCartney and Starr completed it in 2021
The “Now and Then” track was released in 2023 after machine learning was used to clean up an old John Lennon demo.
Context & Ripple Effects
The track’s release followed reports that a sound editor had isolated John Lennon’s vocal from a lo-fi demo and that McCartney and Starr had turned the recovered material into the group’s first newly released song in decades. The Grammy result closes that arc by recognizing the finished performance rather than treating the restoration process as disqualifying.
It also fits the Recording Academy’s stated position that works containing AI elements can compete when the awarded contribution is human. That distinction matters as music tools move from archival cleanup toward systems that can generate songs and vocals from prompts.
First-order effects
- The Beatles, McCartney, and Starr gain a major institutional endorsement for a recording completed with machine-learning-assisted restoration.
- The result reinforces the Grammys’ existing boundary: AI-assisted production can be eligible when the human artistic contribution remains the basis for the award, as outlined in the Academy’s AI-eligibility guidance.
Second-order effects
- Rights holders and producers restoring archive recordings have a prominent example that technical voice separation can coexist with top-tier awards recognition, provided the human creative role is clear.
- AI-music companies will face sharper comparisons between restoration tools used on authorized source material and models that generate new vocals or compositions, such as the prompt-driven music systems covered in Suno’s generative-music push.
Third-order effects
- If awards bodies continue to distinguish assistance from authorship, music crediting and submission practices may increasingly need to document where AI handled signal recovery versus creative generation.
- The broader market could split between provenance-rich archival and artist-authorized projects, which have identifiable rights and human stewardship, and fully generative releases whose eligibility and attribution remain more contested.
The trend: AI is becoming a production-layer tool in recorded music, with industry legitimacy increasingly tied to provenance, human authorship, and the specific role the system played.