Artificial intelligence that speaks and understands sarcasm: Jack Dorsey's new model

Block, Jack Dorsey's fintech company, has unveiled Dessa – a voice-based AI model trained exclusively on human speech, without using text. The model processes audio in real time and detects tone, pauses, laughter, and sarcasm.

Block, the fintech company founded by Twitter co-founder Jack Dorsey, has unveiled Dessa – a voice-based AI model trained exclusively on human speech, without using text. The company claims it is the world's first AI capable of conducting live, realistic conversations trained in this manner. Dessa can listen and respond in real time without converting speech to text, using a speech-to-speech architecture that allows it to listen and speak simultaneously. The model processes audio directly and can respond to cues such as tone of voice, pauses, laughter, and sarcasm. Block stated that the model learned language from a much smaller number of words compared to large language models (LLMs). In February, the company announced layoffs of over 4,000 employees, about half its workforce, following the AI revolution.