If you're using Alexa on a regular basis to assist you with your tasks, you might want to choose what you say to the assistant. It seems that companies dealing with AI have an ongoing theme of using materials they shouldn't to train their AI models.
Alexa is Taking Notes
Some of those who use Alexa as a virtual assistant might be happy about the fact that Amazon is adding generative AI capabilities to it. The retail giant even posted a video of users being amazed by the advancement of the AI assistant.
In an unsurprising turn of events, it turns out that the conversations that users have with Alexa are used as training data for Amazon's AI models. The company has already admitted to the fact, stating that they were using it for product development and improvement.
The company did not deny it. In fact, an Amazon spokesperson said that the customer voice recordings used to train company algorithms were not new, saying that they've "always believed that training Alexa with real-world requests is essential to delivering this experience," as per Gizmodo.
AI systems require a lot of training data including real-world conversation to make their interactions seem more natural. The issue comes when it turns out that private prompts and conversations are also included in the mix.
Amazon has not disclosed a specific number of users that are affected by the training method. The company's spokesperson only provided a vague answer saying that only an "extremely small fraction" of voice recordings are being used.
Read Also: Amazon's Astro Robot Can Now Detect Pets, Recognize and Alert Owners of Issues at Home
Bad Habit of Eavesdropping
Amazon is not new to landing in hot waters regarding privacy issues. The company has actually had this kind of problem with Alexa before. Reports revealed that the retail giant's employees were listening to Alexa's customer recordings before as well as transcribing them.
In some circumstances, the staff would transcribe artist names and link them to certain musicians in the database, while others record commands to determine what the automated system understood and how it responded to the command.
Amazon says that it takes the security and privacy of its customers' personal information seriously and that it only annotates an extremely small sample of Alexa voice recordings in order to improve the customer experience, according to The Guardian.
The company assured that it has strict technical and operational safeguards, and has a zero-tolerance policy for the abuse of its system. Employees cannot access information that can identify the person or account that provided the recording.
As a general response, it all boils down to advancing its service. In Amazon's words: "This information helps us train our speech recognition and natural language understanding systems, so Alexa can better understand your requests, and ensure the service works well for everyone.
Amazon created a feature to avoid or reduce backlash from their methods, which allows users to stop staff from listening in on their voice commands. If the users opt for this, then the company won't be able to use their conversations to train AI models.
Related: Amazon Intends to Invest in OpenAI Competitor, Anthropic
