Putting AI into hardware seemed like a great idea when it was trending a while back, but it has become more clearer with time that people are not exactly going to want to learn how to use a completely new device, especially if it can be done on their phone. This Finnish startup, Binit, however, seems to be bringing something interesting and perhaps very useful to the table—trash, rather—with their new AI hardware.
)
Seeing how AI for sorting trash on a commercial scale has gained a lot of attention in recent times, Binit's direction is a bit different as they are mostly targeting household trash, says Binit founder Borut Grgic, for it is "untapped territory."
Just like a sleep tracker, Binit will track trashing habits. And they are using Large Language Models [LLMs] for household waste object recognition with a wall-mount device equipped with a camera that's designed to be in the kitchen and around where the "trashing" happens.
"We're producing the first household waste tracker," Borut Grgic says to TechCruch. "It's a camera vision technology that is backed by a neural network. So we're tapping the LLMs for recognition of regular household waste objects."
With the camera device in place, the user can simply "pass the object in front of the camera," then they would have to stabilise it for a little bit for the camera to capture all the angles in its image. Then that information is sent to the cloud, and Binit will analyse it with AI and give feedback to the users.
Binit was going to develop and train their own AI model for trash recognition, but they realised that their accuracy was low, about 40%, and decided on using OpenAI's LLM for image recognition. Surprisingly, their accuracy jumped up to 98% with OpenAI's LLM.
With 98% accuracy, Binit's founder Grgic wonders why it works so well and says that he isn't sure if OpenAI already had images of trash as part of their training data or if the LLM is just that good with sorting because of the immense data it has been trained with. "It's incredible," Grgic claims. "It's even able to tell, with relative accuracy, whether or not a coffee cup has a lining because it recognises the brand," he continues.
Many of the prospects for this innovative AI hardware benefit the users in ways that they may not realise, as well as the company. Binit has the potential to become a data provider for companies that need to know what people are throwing away.
Now, it may seem like you do not need some costly gadget to tell you what you already know—wwhat you are throwing away and that you should try to produce less waste—but a different angle may provide a new perspective.
Grgic says "it's habits." We "think that we are aware of it but don't necessarily act on it."
"We also know that it's probably good to sleep, but then I put a sleep tracker on and I sleep a lot more, even though it didn't teach me anything that I didn't already know." He adds.
The idea that you're more aware of what you're doing, seeing the analytics, and interacting with a gamified user interface for the trash you toss away is a psychological response to better handling our waste.
Some argue that an app is sufficient for the function the hardware is providing and that using a phone's camera can work, too. Binit knows this, and so they are also providing the scanning feature through their app for free. That is, both options will be available.
The price point the company is looking at for the AI hardware is about $199, which they claim is the "sweet spot" for smart home devices. They also intend to bring in premium features that you'll need a subscription for.
Posted Using InLeo Alpha
