Addressing claims of Azure AI Speech being labeled a ‘deepfakes creator,’ Microsoft asserted that it has incorporated protective measures in response to the criticism
At its developer conference this week, Microsoft unveiled its recent advancement in artificial intelligence: software capable of producing new avatars and voices or mimicking a user’s existing appearance and speech. This development has sparked concerns about potentially amplifying the generation of deepfakes—AI-generated videos depicting fictitious events.
Revealed during Microsoft Ignite 2023, Azure AI Speech is trained on human images, enabling users to input a script that can be spoken by a photorealistic avatar generated through artificial intelligence. Users have the option to select a preloaded Microsoft avatar or upload footage of a person they wish to replicate in terms of both voice and appearance. In a blog post released on Wednesday, Microsoft highlighted the tool’s potential applications in constructing “conversational agents, virtual assistants, chatbots, and more.
The post states: “Customers can opt for either a prebuilt or a customized neural voice for their avatar. If the same individual’s voice and appearance are utilized for both the custom neural voice and the custom text-to-speech avatar, the avatar will closely resemble that person.”
The company emphasized that the new text-to-speech software is introduced with several limitations and safeguards to prevent misuse. “As part of Microsoft’s commitment to responsible AI, the text-to-speech avatar is designed with the intention of safeguarding the rights of individuals and society, promoting transparent human-computer interaction, and countering the proliferation of harmful deepfakes and misleading content,” the company explained.
Customers can upload their own video recording of avatar talent, which the feature uses to train a synthetic video of the custom avatar speaking,” the blog post reads. “Avatar talent” is a human posing for the AI’s proverbial camera.
Customers have the option to upload their own video recording featuring avatar talent, a term referring to a human posing for the AI’s figurative camera, which the feature then utilizes to train a synthetic video of the custom avatar speaking, as stated in the blog post.
The revelation swiftly drew criticism, with concerns raised that Microsoft had introduced a “deepfakes creator,” potentially facilitating the replication of a person’s likeness to say or do things they haven’t actually said or done. Microsoft’s president himself expressed deepfakes as his “primary concern” regarding the growing influence of artificial intelligence back in May.
The company, in response to the criticism, countered by stating that the customized avatars are currently a “limited access” tool, and users must apply for and be approved by Microsoft to use it. Additionally, users are mandated to disclose when AI is employed in creating a synthetic voice or avatar.
Sarah Bird from Microsoft’s responsible AI engineering division emphasized in a statement, “With these safeguards in place, we help limit potential risks and empower customers to infuse advanced voice and speech capabilities into their AI applications in a transparent and safe manner.
The creation of the text-to-speech avatar maker marks the latest addition to the array of tools introduced by major tech firms seeking to leverage the recent boom in artificial intelligence. Following the widespread success of ChatGPT, developed by the Microsoft-backed firm OpenAI, companies such as Meta and Google have swiftly introduced their own artificial intelligence tools to the market.
The ascent of AI has sparked increasing apprehensions about the technology’s capabilities. OpenAI CEO Sam Altman cautioned Congress about the potential for AI to be exploited in election interference, emphasizing the necessity for the implementation of safeguards.
Experts highlight the specific threat deepfakes pose to election interference. Microsoft recently introduced a tool enabling politicians and campaigns to authenticate and watermark their videos, ensuring their legitimacy and mitigating the dissemination of deepfakes. In a similar vein, Meta unveiled a policy this week mandating the disclosure of AI usage in political ads and prohibiting campaigns from employing Meta’s generative AI tools for ad creation.
