Text-to-speech technology struggles to convey natural emotions and humor, while voice-to-voice technology excels in replicating inflections and cadences.
TikTok could potentially replace human influencers with synthetic influencers that work for free and are fully owned by the platform.
Fully generative platforms like TikTok pose risks of dark mimicry, manipulation, extreme division, and disinformation, highlighting the importance of regulation and detection tools.
As synthetic voices become more realistic and scalable, efforts must be made to balance the potential benefits with the risks of voice-based scams and implement safeguards.
Deep dives
Text-to-Speech and Voice-to-Voice
Resemble.ai offers both text-to-speech and voice-to-voice technologies. Text-to-speech involves converting written text into synthesized audio, while voice-to-voice directly converts one person's voice into another person's voice. Voice-to-voice enables the replication of a person's inflections, emotions, and cadences, resulting in a more natural and nuanced audio output.
Challenges in Text-to-Speech
Text-to-speech technology still faces challenges in conveying natural emotions, emphasis, timing, and humor, as text-based language is inherently ambiguous. While advancements in large language models have improved the inflection and emphasis in text-to-speech, it may struggle to achieve the same level of authenticity as voice-to-voice technology.
Current Use Cases of Voice-to-Voice
Resemble.ai's voice-to-voice technology has been utilized in various industries, including Hollywood and post-production environments. It has been employed in documentary films, movies, AAA games, and even in the docu-series 'The Andy Warhol Diaries' on Netflix.
TikTok as a Generative AI Platform and Future Possibilities
TikTok's algorithm-driven video selection creates a powerful engine for content creation. A future fully generative TikTok could direct, shoot, and optimize its own videos, ushering in a golden age for influencers. However, TikTok may no longer need human influencers if it can produce content at digital speeds and optimize engagement through generative tools. Synthetic influencers would work for free, be fully owned by the platform, and could replace human influencers. TikTok's community features and ability to create viral hits would remain, and the influence of synthetic influencers may rise.
The Dangers of Fully Generative Platforms and Dark Mimicry
Fully generative platforms like TikTok pose risks of dark mimicry and manipulation. A future fully generative TikTok could produce optimized media for addiction, persuasion, or sales, and may not need human influencers. Synthetic influencers would have the ability to alter appearance, produce massive output, and optimize engagement far beyond what human influencers could achieve. However, a fully generative TikTok could also lead to extreme division and disinformation if not regulated properly. Detection tools and watermarks are crucial in combating these risks, and the importance of staying up-to-date with generative AI advancements and addressing the issue of spam and scams is highlighted.
The Power and Pitfalls of Synthetic Voices
The emergence of synthetic voices in AI raises concerns and opportunities. Platforms like TikTok and generative AI technology enable the creation and optimization of synthetic voices. As the technology evolves, synthetic voices could become increasingly realistic, versatile, and scalable. However, this development can also lead to malicious use, such as voice-based scams. Efforts are being made to combat these risks, including the development of deep fake detection tools and watermarks to protect against misuse of audio data. It is important to balance the potential benefits and risks and be proactive in implementing safeguards.
The Threat of Labor-Intensive Scams in a Generative AI Future
Generative AI technology may render labor-intensive scams cost-effective and ubiquitous. Currently, scams like pig butchering require significant manual effort, limiting their scalability. However, as AI chatbots and generative models become more advanced, the cost and labor involved in executing scams will dramatically decrease. Synthetic voices and video manipulation will enhance the quality and realism of the scams, making them more difficult to detect. Steps must be taken to develop detection algorithms and safeguards to protect vulnerable individuals from falling victim to these scams in a future with widespread generative AI technology.
Will the glittering dawn of the genai era be accompanied by a dark tsunami of pixel-perfect deep fakes? I discuss this prospect with Sam Harris, as well as the CEO of synthetic audio pioneer Resemble.ai.
Remember Everything You Learn from Podcasts
Save insights instantly, chat with episodes, and build lasting knowledge - all powered by AI.