Bohan Intelligence calls this disruptive system “adaptive AI quality control system”, which can quickly improve the accuracy of the model from 95% to 99.99%, and maintain such high accuracy standards in the future. According to internal statistics, the online cycle of relevant solutions is shortened by 80%, labor cost is reduced by 90%, and operation and maintenance cost is saved by 80%.

This adaptive AI inspection system is also in line with the cutting-edge wind in Silicon Valley. In 2021, former Baidu chief scientist Andrew Ng, known as the “godfather of AI” in the industry, said in a live broadcast that AI systems are shifting from Model-centric to Data-centric. This means that in the future, data quality will be more important than the model itself.

As for the difference between the two, Guo Wei further explained that Model-Centricai previously focused on Model algorithms, which can be likened to human intelligence quotient, representing the basic structure and ability to process and learn information. The technological route is to improve the algorithm itself to achieve better results, that is, to enhance human intelligence to enhance capabilities. Data-centric, on the other hand, allows people to improve their capabilities by accumulating experience, which means that by feeding the machine high-quality Data, the AI can continuously adapt to changing circumstances and needs.

TP854 3BSE025349R1

TP854 3BSE025349R1

TP854 3BSE025349R1

It’s worth noting that MLOps are a crucial link in the Data-centricAI route, and one that Bohan is good at. The advantage of MLOps in industrial scenarios is that it not only helps DataCentricAI’s tasks become more efficient, but also ensures that the model remains valid in the scenario.

For first-line operations, DataCentricAI combined with MLOps technology will effectively solve the problem of model failure, enabling AI systems that originally floated in the cloud to have a more standardized landing strategy — data flywheel. The data flywheel has been in use overseas since 2021.

“Quality inspection and unmanned driving have the same underlying logic in data training, but the degree of complexity is different.” “Guo Wei said. This is why in addition to focusing on AI quality inspection in the manufacturing field, another anchored track of Bohan Intelligence is automatic driving. At the same time, he also will the strategic positioning of Bohan Intelligent, “the first three years of revenue to manufacturing, the second three years of revenue hope to achieve two-wheel drive.”

At present, in the manufacturing field, Bohan intelligent mainly focuses on 3C and semiconductor AI detection. According to Guo Wei, “The market size of PCBA and appearance inspection alone is estimated to be more than 40 billion yuan.” At the same time, the PCBA market, where 20% of the suppliers will control 80% of the capacity, will be an ideal business environment for a strong to B market.

At the present stage, Bohan intelligent adaptive AI quality inspection solutions have been implemented in a number of industry head customers. The next step is to realize rapid and large-scale replication of the model envisaged by Guo Wei, so that the cost decreases significantly with the marginal effect, and the AI quality inspection model of Bohan Intelligence will run in the field of PCBA and be realized out of the box. But until then, everything has yet to be proven, and it remains to be seen whether this adaptive AI system can be extended to other areas of AI quality inspection.

In addition, riding on ChatGPT this year, more and more Silicon Valley AI elites are coming home to join the machine wars.