# Speed-of-Light Computing Could Get Us to AGI Very Fast Source: https://www.youtube.com/watch?v=AmEWj0PB12s Recap page: https://rapidrecap.app/video/AmEWj0PB12s Generated: 2025-12-13T16:38:27.991+00:00 --- ## Quick Overview Photonic computing, which uses light particles (photons) instead of electricity for processing, promises substantial speed, efficiency, and multichannel advantages over current electronic chips, but it currently faces significant drawbacks in storage and universality, meaning that while Chinese scientists achieved a 12.5 GHz optical AI chip, the technology still requires electronic components for memory storage and needs more research to overcome limitations like error rates before it can fully replace silicon designs or achieve AGI. **Key Points:** - Photonic computing utilizes photons instead of electrons, offering speed advantages where calculations occur at the speed of light (12.5 GHz demonstrated by a Tsinghua chip vs. 100-500 ns for electronic chips). - Key advantages of photonic computing are speed, efficiency (requiring only 10% of the power of electronic equivalents), and multichannel capability via different frequency bands. - Major drawbacks include the inability to store information using photons, requiring supplementary electronic components for storage, and a lack of universality for general logical operations. - Chinese scientists at Tsinghua University created the world's fastest optical AI chip running at 12.5 GHz, serving as a strong benchmark, although the presenter rates the hype surrounding this as a 5/10. - US startups like Lightelligence and Qant are also developing photonic processors, claiming up to 50x performance improvements and significant energy reduction (up to 90%). - The technology is already being used in aerospace, biomedicine, and finance, but achieving the precision and reliability of established electronic methods remains a challenge. - The presenter concludes that while photonic computing is promising, more research is needed to address error rates and universality issues before it can completely supplant electronics. ![Screenshot at 0:05: The introduction of photonic computing highlights its core promise: computing at the speed of light, contrasted visually with a burning electronic chip to emphasize the potential breakthrough.](https://ss.rapidrecap.app/screens/AmEWj0PB12s/00-00-05.png) **Context:** This video provides a scientific news update, hosted by Sabine Hossenfelder, discussing the emerging field of photonic computing—the use of light (photons) rather than electrical current to perform computations. The discussion centers on recent progress, particularly developments in China and the US, weighing the touted benefits like extreme speed and low energy consumption against inherent technological limitations like data storage and operational universality. ## Detailed Analysis Sabine Hossenfelder reports on photonic computing, which uses light particles (photons) instead of electricity, suggesting it could pave the way for superintelligent AI. She notes recent hype, particularly surrounding a 12.5 GHz optical AI chip developed by Tsinghua scientists in China, which serves as a fast benchmark. The advantages of this technology are speed (photons move at light speed, resulting in 250 picoseconds per operation versus 100-500 nanoseconds for electronic chips), high efficiency (requiring only 10% of the power), and multichannel capability using different light frequencies. However, critical limitations exist: photons cannot easily store information, necessitating hybrid electronic components for memory, and they lack the universality of electronic transistors for general logical operations. The video references news about US companies like Lumai (promising 50x performance improvement and 90% energy reduction) and Qant (developing a native processing server). Hossenfelder rates the current hype level as a 5 out of 10, cautioning that while progress is real, the technology still faces hurdles in precision, reliability, and error rates compared to mature silicon technology, meaning it is not yet ready to completely replace classical computing for all tasks. ### Photonic Computing Overview - Speed advantage (12.5 GHz chip) - Efficiency (low energy loss) - Multichannel data transfer - Drawbacks: Poor storage and universality (cannot perform all logical operations without electronics) ### Industry Progress - Tsinghua scientists create world's fastest optical AI chip at 12.5 GHz - US companies Lumai and Qant are developing processors aiming for 50x faster computation and 90% energy efficiency ### Current Limitations - Photons cannot store data, requiring electronic components for memory - Photons lack the non-linear interaction needed for universal logic gates, unlike transistors - Error rates and reliability need significant improvement before widespread adoption ![Screenshot at 0:00: Host Sabine Hossenfelder begins the science news segment against a futuristic blue/purple digital background.](https://ss.rapidrecap.app/screens/AmEWj0PB12s/00-00-00.png) ![Screenshot at 0:07: An animation shows a computer chip engulfed in flames, overlayed with the text "Photonic Computing," illustrating the heat/energy inefficiency of current technology.](https://ss.rapidrecap.app/screens/AmEWj0PB12s/00-00-07.png) ![Screenshot at 0:21: A bar graph showing an upward trend appears next to the text "Light-Speed Computing," visually representing the promised performance gains.](https://ss.rapidrecap.app/screens/AmEWj0PB12s/00-00-21.png) ![Screenshot at 1:51: An abstract graphic illustrates photons \(colored beams\) moving through a circuit board, representing the core mechanism of photonic computation.](https://ss.rapidrecap.app/screens/AmEWj0PB12s/00-01-51.png) ![Screenshot at 2:43: The key drawback of photonic computing is highlighted with the word "Storage" and a thumbs-down emoji, indicating poor memory capability for photons.](https://ss.rapidrecap.app/screens/AmEWj0PB12s/00-02-43.png)