The rapid growth of voice assistants powered by large language models (LLM) has highlighted a need for speech instruction data to train these systems. Despite the abundance of speech recognition data, there is a notable scarcity of speech instruction data, which is essential for fine-tuning models to understand and execute spoken commands. Generating high-quality synthetic speech requires a good text-to-speech (TTS) model, which may not be available to low resource languages. Our novel approach addresses this challenge by halting synthesis at the semantic representation level, bypassing the need for TTS. We achieve this by aligning synthetic semantic representations with the pre-trained Whisper encoder, enabling an LLM to be fine-tuned on text instructions while maintaining the ability to understand spoken instructions during inference. This simplified training process is a promising approach to building voice assistant for low-resource languages.

It's not that hard to say my name, Saahil Desai. Saahil: rhymes with sawmill, or at least that gets you 90 percent there. Desai: like decide with the last bit chopped off. More often than not, however, my name gets butchered into a menagerie of gaffes and blunders. The most common one, Sa-heel, is at least an honest attempt--unlike its mutant twin, a monosyllabic mess that comes out sounding like seal.

Despite the fact that AI is already being utilised in a huge variety of ways by a myriad of industry, it's still perceived as being an emergent technology, something of great promise yet to materialise. This couldn't be further from reality. These four examples of AI and machine learning have already been saving you time and effort for years at this point, but it's only really in the last few years that they've become completely viable and effective. I mean, remember Siri when it first came out? AI, along with machine learning, has made a massive number of things possible across industries, proving critical to many different tech platforms.