Meta’s Lips Are Sealed When Asked If It Can Train AI Models On Images From Its AI-Powered Smart Glasses
Meta just unveiled the final look of its highly anticipated smart glasses. Powered using AI technology, the offering was highlighted as having discreet cameras on the front. Meanwhile, the AI system can even give rise to specific keywords like ‘Look’.
This means the product carries the capability to collect hundreds of images, either taken deliberately or otherwise. However, the firm refuses to comment on whether or not the images taken will remain private or not.
When Facebook’s parent company was asked if it has plans to have AI models trained on this data from its latest Ray-Ban offering, it refused to comment. The company remained hush at all times, despite the world knowing that it does take data from public social media accounts for this purpose.
For now, the company’s head of AI wearables says this is the least of their concerns. They refused to comment on that publicly as they felt it was private.
The reason why this is causing a lot of concern among the masses has to do with the simple fact that the new Ray-Ban has the tendency to take tons of clicks passively. This means the data in storage is diverse and to use that without consent can lead to many issues.
Meta also is gearing up to launch another key feature that entails videos in real-time. After it gets activated with specific key terms, the product can stream a long series of pictures into a multimodal AI model. This means it can answer queries about the environment more naturally.
This is a lot of pictures it’s going to get exposed to and most of them might not even make the user consciously aware of what is being done at this point in time. In case the smart glasses are being used for scanning material inside your fridge, it end up clicking dozens of images and then uploading those to the cloud’s AI model or database.
Now the question is what happens next because Meta refuses to say anything on this front.
When you wear the smart product, you are also wearing a camera on the face. As seen through Google Glass, it’s not something that people might be comfortable with and that is why the company needs to be more clear. Users are waiting for reassurance that their data remains private and not used for training AI models.
During the training of its own models, Facebook’s parent firm was so clear about how it was using every public post from Meta’s apps for training purposes. According to them, it’s all available publicly and the world needs to accept it.
Tech giants have taken on a new definition of what’s up for grabs publicly for AI training and what is not.
The world you get to see through AI glasses isn’t available publicly. But we cannot say anything right now as Meta might be using the footage for its own benefits. What is interesting is how many other AI model providers are delineating clear rules on their training methods and whether or not user data is utilized.
For instance, Anthropic says it would never train using a client’s inputs or outputs from the AI model. The same is the case with OpenAI which says it never uses API inputs and outputs.
For now, we’re still going to wait for Meta’s full explanation and clarification on this front and when we get it, we’ll keep you posted.
Image: DIW-Aigen
Read next: Fortnite Maker Epic Games Sues Samsung and Google For Anticompetitive Behavior
