Owly - An AI powered iOS mobile app that blocks spam messages

@jin_wang007 With GPT-4 at the helm, I'm curious, how does it distinguish between a spam message and an important, albeit unsolicited, one? Also, Jin, since you've had it running for a while, how does it adapt to new types of spam that evolve over time? Congrats on the launch and good luck!

@anastasia_chakileva_ Hi Anastasia, thanks for the questions. Q: how does it distinguish between a spam message and an important, albeit unsolicited, one? A: There are two aspects on this. First, the iOS system only allow messages from an unknown source to be filtered. The messages from your friends or anyone in the contacts will not go through the check at all. Any important notifications from a known source, such as an emergency notification from your government, will also skip the check. Second, if the message is from an unknown number, GPT has the ability to identify typical features in a spam. For example, a spam usually comes with a suspicious link for you to click and it usually sounds very urgent by using terms such as "Quick Action Needed", "Immediately" etc. Q: how does it adapt to new types of spam that evolve over time? A: The GPT model the app uses keeps learning from the internet so it kinda evolves itself over time. In the past few months, the model had at least 2 big updates.

@saifrkhaan Hi Saif, thanks for the questions. One very important thing is that, only messages from an unknown source is allowed by the iOS system to be processed by the app. The messages from anyone in your contacts will not even reach the app. At the moment, these suspicious messages from unknown sources will be sent to the server but the server never stores them. It removes the messages as soon as it's done processing. The GPT-4 model is updated by OpenAI every now and then. It keeps learning from the internet so any new spam messages will be picked up.

@jin_wang007 thanks for clarifying!

@abhilash_chowdhary Thanks!

@korzonek Hi Tomasz, thanks for the comment. The filtering system does support Polish. So if you receive a spam message in Polish, it'll stop it just fine. The app itself doesn't support Polish (instructions on how to set it up etc but it's straightforward enough and a one-off thing). Once you set it up, it's likely you won't need to open the app anymore so it should mostly be fine.

@dne Yeah, to be honest, it's really hard to come up with the subscription models. On one hand, I think people might not be willing to pay unless they know the app works. That's why there is a free tier. But on the other hand, GPT costs money. The more accurate you want, the more you need to pay. Therefore I had to limit the free version. That's where the "1" comes from. I don't particularly like that but I couldn't figure out a better model.