Diversified Chinese technology giant Baidu has launched a cloud-based tool, EZDL, that it claims makes it easy for anyone to design, build, and deploy artificial intelligence (AI) algorithms without writing any new code.
Baidu says the system is designed to open up AI development to small and medium-sized companies and non-experts.
To train a machine learning model, Baidu says that EZDL requires as few as 20-100 images and/or 50 audio files assigned to each label, and training may take as little as 15 minutes to an hour.
According to Baidu, potential applications for EZDL could include:
- Image content identification: custom identification, showing business-specific image-recognition results
- Industrial sorting: custom conveyor belt product identification, for sorting or quality-inspection purposes
- Image censoring: custom image-censoring rules, such as identifying smoking during live broadcast – a telling comment, given China’s censorious online environment.
Algorithms generated by EZDL can then be deployed in the cloud and accessed via an API, or downloaded in the form of an SDK that supports a range of different operating systems, including iOS and Android.
The company claims that more than two-thirds of models get accuracy scores higher than 90 percent, according to a report in Venturebeat.
However, beta users should be wary of an implicit problem with the system: up to ten percent inaccuracy. More, they should also consider the security implications of uploading sensitive images to an online AI tool.
“Through our Global Business Unit, we are devoted to sharing Baidu’s core technologies, capabilities, and resources with the global community,” said Youping Yu, general manager of Baidu’s AI ecosystem division.
“It is our hope that by opening up the platform and democratising AI capabilities, we will help developers and businesses around the world achieve greater success.”
Internet of Business says
Modular development has been a major trend this decade, but Baidu’s system is merely the latest evidence of another trend: the democratisation of AI.
Just last week, for example, Microsoft released an AI Web design tool that can generate functional HTML from freehand sketches, while in August Gartner predicted that democratised AI will be commonplace over the next decade.
Gartner’s latest Hype Cycle research, which draws together insights on over 2,000 technologies, suggests that AI will be “virtually everywhere” within the next 10 years, but – significantly – it will be open to the masses, rather than being the sole, expensive preserve of large enterprises.
A range of factors will contribute to the emergence of mass-market AI, including cloud platforms, open-source projects, and the maker community, eventually “propelling AI into everyone’s hands”, according to the company.
Gartner’s research also says that AI-based Platform as a Service (PaaS) solutions, conversational AI, and deep neural networks are set to become enterprise technology staples, alongside mobile robots and driverless vehicles.
“Technologies representing democratised AI populate three out of five sections on the Hype Cycle, and some of them, such as deep neural nets and virtual assistants, will reach mainstream adoption in the next two to five years,” said Gartner research VP, Mike Walker.