CPO Review in Nature Electronics Draws Criticism Over Errors and Suspected AI Generation

SK-affiliated researchers published a review, "Co-packaged optics for high-performance computing and artificial intelligence," in Nature Electronics, offering a systematic survey of CPO for HPC and AI. Blogger @jwt0625 found it worth reading despite minor errors, while also mocking the continued use of PDFs for technical knowledge in 2026 and a interconnect startup's use of a suspected AI-hallucinated ELSFP image in a serious presentation.

Confirmed

• The paper was published in Nature Electronics as a review of co-packaged optics for HPC and AI.
• @jwt0625 identified multiple technical flaws: the tier diagram conflates optical-engine and fiber materials, inter-rack fiber links are mislabeled as copper, and the key technology NPO is never mentioned.
• @jwt0625 criticized the unnecessary complexity of the drawn QPSK modulator and questioned the implication that MRM people would use supercontinuum sources.
• @jwt0625 argued CPO is currently bottlenecked by laser sources, yet the paper barely discusses external lasers; the integration technologies covered are not production-ready, and mature packaging-level laser integration (e.g., Intel's high-volume shipments) is omitted.

Unconfirmed

• Some netizens suspect the paper was largely AI-generated; @jwt0625 relayed these doubts, but this remains speculation with no response from the authors.

Why it matters

• The review is a top-journal systematic treatment of CPO, valuable to readers tracking AI compute interconnects.
• The controversy over figure errors and AI-generation suspicions reflects renewed debate on academic rigor in AI-assisted writing.
• The critique that external lasers and production-grade packaged laser integration are CPO's key bottlenecks flags potential blind spots for the industry.

2026-08-22 ~ 2026-08-22 · 5 related posts

• Reviewing CPO Paper and Roasting Hallucinated AI Images in Tech Decks — jwt0625 · 2026-08-22
• [source] Nature Electronics review maps co-packaged optics for AI compute interconnects — jwt0625 · 2026-08-22
• [source] HPC optics paper accused of being AI-generated due to technical errors — jwt0625 · 2026-08-22
• Critique of CPO paper: Overlooking laser source complexity and ignoring bottlenecks — jwt0625 · 2026-08-22
• [source] CPO现状吐槽：被激光源卡脖子却仍无视成熟封装方案 — jwt0625 · 2026-08-22