3 Alarming Challenges in Personal AI Development

Smartphone manufacturers are now releasing their newest models with the adoption of AI techniques, bringing the era of living with personal AI right around the corner. But we’ve seen thousands of sci-fi movies and novels featuring threatening robots and AI, haven’t we? Furthermore, we also discussed how AI can threaten your privacy in our previous posts. So why don’t we take another look at the possible threats, this time focusing more on personal AI?

Better Knowledge, More to Disclose

Just as a simple common sense, if you know a lot about a topic then you also have a lot to talk about it. The same rule applies to AI. Since the only goal of AI model in your phone is to study you, it’s very eager to gather as much information about you as possible. For your personal AI, the topic is you and you only. Therefore, you are inevitably faced with privacy issues. Still, we cannot avoid the rapid change in technique and want to stay up-to-date with it. Because you know, it’s way too convenient. Embracing innovation doesn’t mean throwing caution to the wind. It’s about smartly navigating the tech landscape.

Know Your Personal AI

It’s about the balance in the end. We want technology to help us without causing any harm. Then we need to know how it works first. Normally, you train an AI model with billions of data to gain general knowledge. If the model works well with unseen data then we move to the next step. We fine-tune the model with more specific data so that it can learn some details while keeping the pre-gained knowledge in mind. Namely, we personalize it. Your personal information goes to the model at this step. That is, it’s more about customization than performance itself.

The More You Fake, The Safer You Are

As discussed in our previous posts, you don’t necessarily need to provide the REAL data to train an AI model. That way, you can be both safe and convenient.