Dessa

Dessa, Block's new voice AI, represents a significant leap in artificial intelligence's ability to mimic human speech. Unlike many existing systems that rely on text-to-speech conversion, Dessa processes audio directly, allowing it to capture and replicate the subtle emotional cues, pauses, and natural cadences that define human conversation. This direct audio processing is key to its remarkably lifelike output, as demonstrated in company recordings where the AI exhibits laughter and reacts with apparent emotion. The model's ability to run locally on high-end consumer hardware, such as a MacBook Pro with an M4 Max chip, suggests a potential for widespread adoption and integration into various applications.

The training methodology, exclusively on human speech and conducted on a single powerful computing system, highlights an efficient approach to developing sophisticated AI. Block's decision to acquire the original Dessa AI startup in 2020 underscores a long-term commitment to advancing this technology. The company's current strategy of releasing Dessa as a research preview to AI safety researchers indicates a cautious approach, aiming to identify and mitigate potential risks before a broader public release. This measured rollout is crucial given the dual nature of the technology's reception.

Jack Dorsey

Jack Dorsey, co-founder and CEO of Block, has been a vocal proponent of Dessa, sharing his amazement at its human-like qualities. His endorsement, particularly on platforms like X, has amplified the technology's visibility and sparked widespread discussion. Dorsey's direct involvement and the AI research lab's reporting structure to him suggest a strategic priority for this technology within Block's broader ecosystem, which includes Cash App and Bitkey. The emphasis on AI safety, with early access granted to researchers, aligns with a growing industry trend towards responsible AI development, though the inherent risks of such powerful voice synthesis remain a significant concern.

Dorsey's vision for Dessa appears to extend beyond mere technological advancement, hinting at its potential to revolutionize human-computer interaction. The ability to converse naturally, with emotional nuance, could unlock new possibilities in areas like education, customer support, and accessibility. However, his public statements also acknowledge the dual-use nature of the technology, implicitly recognizing the need for robust safeguards against malicious applications. The company's acquisition of the original Dessa AI startup in 2020 further solidifies its dedication to this field, positioning Block at the forefront of voice AI innovation.

Scam Fears

The uncanny human resemblance of Dessa's voice has ignited significant apprehension regarding its potential for misuse, particularly in perpetuating scams and deception. Critics and observers have voiced concerns that such advanced voice AI could be weaponized to create highly convincing fraudulent communications, making it increasingly difficult for individuals to distinguish between legitimate and malicious interactions. This fear is amplified by the technology's capacity to replicate emotional tones and natural speech patterns, which are often key indicators used to detect fraudulent calls.

Concerns about the proliferation of sophisticated scam calls, potentially reaching "pandemic level," have been raised by journalists and social media users. The ability of Dessa to handle interruptions and maintain a natural conversational flow means that AI-generated voices could bypass current detection methods that rely on robotic speech or unnatural pauses. This poses a substantial threat to personal security and financial well-being, especially for vulnerable populations who may be more susceptible to emotionally manipulative tactics employed by scammers. The lack of disclosed benchmarks or a release date, coupled with the current limited access to AI safety researchers, highlights the nascent stage of risk assessment for this powerful technology.