Fighting Audio Deepfakes & AI’s Role in Education
▼ Summary
– A new technique called “machine unlearning” can teach AI models to forget specific voices by creating a modified version of the original model without the unwanted data.
– Machine unlearning offers a potential solution to prevent audio deepfakes, which are used for fraud by copying someone’s voice.
– OpenAI, Microsoft, and Anthropic partnered with a major US teachers’ union in a $23 million initiative to integrate AI into K–12 classrooms, starting with teacher training this fall.
– The initiative aims to help teachers use AI for tasks like lesson planning and report writing, but skepticism exists about its educational benefits.
– Critics question whether AI companies can provide unbiased advice on AI use in classrooms due to a lack of clear evidence on its net benefits for students.
The rise of audio deepfakes poses a growing threat, but emerging “machine unlearning” techniques could help combat voice cloning scams. This innovative approach focuses on teaching AI models to selectively forget specific data, such as a person’s voice, rather than just filtering unwanted outputs. By retraining models to exclude sensitive information, companies may soon have a powerful tool to prevent unauthorized voice replication.
Unlike traditional methods that screen prompts or responses, machine unlearning creates a modified version of the original AI model, one that never absorbed the problematic data in the first place. If successfully implemented, this could significantly reduce the risk of fraudsters using AI-generated voices to deceive victims.
Meanwhile, major tech firms are pushing AI into education, sparking both excitement and skepticism. OpenAI, Microsoft, and Anthropic recently partnered with a leading U.S. teachers’ union, investing $23 million to integrate AI into K–12 classrooms. The initiative includes training educators on using AI for lesson planning, grading, and administrative tasks, with pilot programs launching this fall.
However, critics question whether these companies can provide unbiased guidance on AI’s role in schools. With limited evidence proving AI’s educational benefits, some worry the push prioritizes corporate interests over genuine learning outcomes. As classrooms increasingly adopt these tools, the debate over their effectiveness, and potential drawbacks, continues to intensify.
(Source: Technology Review)
