800 pieces of DIY AI hardware, the whole process is open source, netizen： the new Linux era is coming.

Goose sister! AI hardware, you can do it yourself!It only takes one hour and the cost is less than 800 yuan.. It can record people’s conversations and environmental sounds to a private database at any time, and customize the exclusive AI assistant. andThe whole process is completely open source.All the data collected are controlled by the users themselves.

In the promotional video, the author shows his private GPT formed after wearing this device for several days. The author said that he has been discussing this hardware with others in the past few days.

When he asked about his big model, he could see that the model had learned what the author said these days and introduced his wearing equipment in great detail.

Moreover, the whole system can DIY by itself.The hardware cost is only $100, about 720 RMB. However, the brand AI in the market is expensive at six or seven hundred dollars, and some of them only pack software services for one year. The key is that they are not in stock yet.

This system has caused a burst of cheers from the open source community. Some people say that the Linux time for AI to wear has arrived.

So, how does such an AI hardware DIY?

Structurally, the whole system worn by this AI mainly consists of the following four parts:

Let’s look at the hardware first. The current hardware consists of a Coral AI development board ($80) and a Wifi module ($20).

The plastic box outside is optional and costs $10.

Because the two boards come from the same manufacturer, the hardware assembly process is very simple, just plug them together.

Next, we need to register a database account and get the API.

Then install the database client and docker on your own computer or server, and configure the database and App server according to the tutorial. The detailed process of this part can be read on the author’s GitHub project page.

Prepare the software and hardware, and then write the program for the board.

After connecting to the computer with the type-C data cable, it is also a step-by-step operation to complete the compilation and upload. In this process, you need to write the WiFi name and password to the device.

Then, such an AI wearable device is done. Then, as long as you wear it with you, you can build your own large model database according to your own conversation.

If you are interested, you can try it yourself! (See the project portal at the end of the article)

$100, which is far from the lower limit of hardware cost-

The author plans to try to replace Coral development board with cheaper raspberry pie soon.

The raspberry pie Zero W with integrated WiFi module costs only $15. If it succeeds, the cost will be directly reduced by 85%.

In addition, the SD card can be inserted, so in addition to reducing the cost, the use of raspberry pie can also localize the cloud database.

At the same time, the author also plans to add Bluetooth and other communication methods in addition to WiFi, so that it can truly become a portable device and upload data anytime and anywhere.

Since Coral’s WiFi module also supports Bluetooth, this operation only needs to update the software without replacing the hardware equipment.

In view of the problem that there is no battery and external power supply is needed, the author has initially chosen a battery device, and it is expected that there will be a solution soon.

Waiting for the party, you may wish to continue to expect a full version ~