Of course, no complete artificial intelligence is given to the buyer, but a tool that he can experiment with. Neural Compute Stick 2 only costs $ 99.
Although it costs $ 20 more expensive, Intel promises – the new USB Neural Compute Stick is eight times faster than the first release released last year. The 99 $ units announced at the AI DevCon event in Beijing can be connected to a computer via a USB port. It is good to develop prototype products that use artificial intelligence (MI) or machine learning.
Intel mentions, for example, a project that has built an MI system that can identify contaminants in water and know that only $ 500 – a cheap microscope, a Ubuntu Linux laptop, and Neural Compute Stick 2 to say about the water sample is a potential health risk .
Intel has strengthened its artificial intelligence
Intel tries hard to keep the MI trend out. In 2016 he received two very important companies for his plans. In August, she acquired Nervana Systemset, founded just two years earlier, who developed deep learning solutions (application-specific chips and a Python-based language). A month later, he called Movidius in Dublin to develop special SoC (system-on-chip) efforts for machine vision. (Movidius delivered the chip to Google for Project Tango and DJI drone.)
Neural Compute Stick, the first version released last year and costing only $ 79, is a very good alternative to hobbyists to familiarize themselves with artificial intelligence. Of course, Intel is not a hobbyist, but rather a fan of developers, but everyone can buy it.
The idea of using such a USB device to centralize the machine came from Movidius, also the Irish company designed Fathom Neural Compute Stick, but only developers could get it. This was further developed and released by Intel. In the now announced device, Movidius Myriad X works with Vision Processing Unit (VPU) with 16 vector vector vector vector vectors (SHAVE).
The USB 3.0 port can be connected to a computer – more than one machine – and can be used with 64-bit Ubuntu, CentOS or Windows 10. The device itself performs the necessary MI calculations or accelerates such calculations of the system.
Indirectly helps artificial intelligence activities
Even the first edition of Neural Compute Stick was also very successful among developers – it sold tens of thousands of iPads and had less demand than Intel produced – but that's not what Intel makes for money. The chip maker expects the device to help develop prototypes faster can encourage developers to use Movidius chips for their MI solutions.
Intel is responsible, Steen Graham, Intel's IoT Area Executive Director promises that there will be no device shortage right now.