Whisper

More than a dozen software engineers, developers, and academic researchers say OpenAI's voice transcription tool Whisper has a major flaw: it sometimes generates large paragraphs or even entire sentences of false content out of thin air, according to a report by the Associated Press on 27 May. The experts note that the generated text, commonly referred to in the industry as "AI hallucinations," can involve racial remarks, violent language, and even fabricated medical advice. This is particularly worrisome, according to the experts, because Whisper is already being used in a wide range of industries around the world, including for translating and transcribing interviews...

At its DevDay event on October 1, OpenAI announced the release of the Whisper large-v3-turbo speech transcription model, which has 809 million parameters and is 8x faster than large-v3 with virtually no loss in quality. The Whisper large-v3-turbo speech transcription model is an optimized version of large-v3 and has only 4 Decoder Layers as compared to large...

NVIDIA NeMo, the leading open source conversational AI toolkit, announced the release of the Parakeet ASR model family, a series of state-of-the-art automatic speech recognition (ASR) models that transcribe spoken English with excellent accuracy. Developed in collaboration with Suno.ai, the Parakeet ASR models are a breakthrough in speech recognition, paving the way for more natural and efficient human-computer interaction. According to the developers, the models are robust to non-speech segments such as music and silence, and outperform OpenAI in benchmarks…