According to news from IT House on July 27, Microsoft recently launched a speech model called NaturalSpeech2, which uses a "potential diffusion" design and has an outstanding effect at the level of zero-sample speech synthesis. "Class" speech/singing solution, which can give users high-quality and diverse speech synthesis experience. Unlike traditional speech-to-text (TTS) systems, Microsoft's NaturalSpeech2 uses "continuous vectors" instead of "discrete tokens" to represent speech, resulting in more complete speech fragments that do not produce "unsentimental" "stick readings" (a speech in one word)" phenomenon. The experimental results show that the speech generated by NaturalSpeech2 under the zero-sample condition is nearly consistent with the prosody of the speech prompt and the real speech, and the naturalness (measured by CMOS) on the LibriTTS and VCTK test sets is indistinguishable from the real speech.

Jinse Finance reported that researchers at Dresden University of Technology in Germany recently published a breakthrough study demonstrating a new material design for neuromorphic computing, a technology that may have a revolutionary impact on blockchain and artificial intelligence. Using a technique called "reservoir computing," the team developed a pattern recognition method that uses magnon eddies to perform algorithmic functions almost instantaneously. The researchers not only developed and tested new storage materials, but also demonstrated the potential of neuromorphic computing working on standard CMOS chips, which could disrupt blockchain and artificial intelligence. The main benefit of neuromorphic computing is its extremely low power consumption relative to classical and quantum computing. This means that neuromorphic computers can significantly reduce time and energy costs when operating blockchains and mining new blocks on existing blockchains.