Rokid, a Hangzhou-based AI firm, launches open-source AI operating system YodaOS with full stack and officially released its code on Github.
Rokid, a Hangzhou-based AI firm, launches open source AI operating system YodaOS with full stack and officially released its code on Github.
This operating system is based on Linux Kernel, which hardware abstraction layer, AI layer and JavaScript framework are based on. FCTC，the Rokid cloud voice algorithm is integrated and applied to the terminal device in order to ensure that the voice interaction experience is consistent between the terminal and cloud under the same computing power.
By integrating YodaOS into terminal devices, Rokid claims that the activation success rate of intelligent voice products will increase by 20% under strong noise and achieve a wake-up speed of less than 300ms. Based on the open concept, YodaOS encapsulates the algorithm in AI abstraction layer facilitating users to develop AI applications and integrates Rokid content, services, and IoT applications. It says YodaOS supports adapting to the mainstream chips and hardware such as Qualcomm, Jingchen, Samsung, MediaTek, and Broadcom.
Previously，EqualOcean reported that Rokid planned to change its business model from smart devices for consumers to smart services for business. As can be seen that product offerings and targeted customers will be changed.
One reason for making changes is that Rokid sees the era of AI is on its way. ZHU Mingming (祝铭明), founder and CEO of Rokid, defined itself as a technology-driven company focusing on human-computer interaction instead of the hardware company.
Rokid was founded in 2014 and launched its first product - voice assistants Rokid Alien in 2016 and has been awarded CES Innovation Award in the same year. Now, Rokid offers two smart speakers for sale, which are Rokid Pebble (CNY 1399) and Rokid ME (CNY 799).
Although the global smart speaker market is booming, the competition is getting tougher. In the first quarter of 2018, global smart speaker shipments increase by 200% quarter-on-quarter and smart speaker shipments increase by 5370% in the Chinese market. China now has the second largest smart speaker shipments after the U.S.
- A quick analysis:
The Chinese smart speaker market is now made up of 5 types of companies.
1. Internet-based company
Alibaba, Baidu, and Xiaomi launched smart speakers as a means to collect user’s data and bring more traffic to products that they backed, and ultimately generate more profit from understanding consumers ‘behaviors through big data analysis.
For example, Alibaba and Baidu launched budgeted smart speakers. Alibaba’s Tmall Genie Fang Tang and Baidu’s Xiaodu are both discounted at CNY 89. Xiaomi’s Xiaoai was firstly selling at CNY 169 then soon discounted to CNY 99.
2. Technology-driven company
Iflytek (科大讯飞) is the leading voice recognition company based in China. Previously, it has worked with JD.com to develop a smart speaker called DingDong.
3. Content provider
Ximalaya FM (喜马拉雅), Chinese biggest audio sharing platform launched its smart speaker in 2017 named as Xiaoya. One of the advantages that Xiaoya has it has a large number of audio resources including exclusive audio content and data of users’ listening behaviors collected from Ximalaya FM App. And it also it is said that Xiaoya’s could recommend more suitable content to users.
4. Traditional speaker manufacturer
Since the smart speaker market is growing rapidly. Traditional speaker manufacturers were about to embrace this trend.
For example, DOOS, a Shezhen-based traditional speaker manufacturer, has collaborated with Baidu to launch a smart speaker based on Duer OS, a conversation-based artificial intelligence (AI) platform developed by Baidu.
5. Start-up company
Startups are also eyeing presence in the fast-growing smart speaker market. Rokid is one of them. Those startups usually aim at a niche market that has not been targeted by giants.
The second reason for making changes is, according to ZHU, he hopes by opening voice technology could bring more talents into this industry to explore possibilities and optimize, and eventually benefit the industry as a whole.
It is not the first time for Rokid to open resources. In 2017, Rokid partnered with Alibaba to develop the first full stack open platform in China which allows developers access Rokid’s technology, hardware solutions, and services such as far-field interaction, multi-noise identification, customized wake-up word, as well as voiceprint recognition.
In summary, tougher competitions and lower profit margin in the smart speaker hardware sections and more possibilities achieved through open source are the key drivers for Rokid to move from a hardware manufacturer to an AI technology company.