By AI Trends Staff
Generic AI models save time by compressing a percentage of the work involved in launching an AI application and providing it for reuse. An excellent example is Google Cloud’s Vision AI, which has access to pre-built models for sensing emotions and understanding text.
Some emerging companies seek to build on this trend by supplying ready-made models developed for a particular vertical industry to outweigh the benefits of generic ready-made models in all industries.
“While these solutions are effective in some applications, they are not suited to industry-specific needs right from the start. Organizations looking for the most accurate results from their AI projects simply need to use industry-specific models,” said DJ Das, founder and CEO of ThirdEye Data. TechCrunch. ThirdEye builds artificial intelligence applications for businesses.
Companies have options to produce industry-specific results. “One would have to adopt a hybrid approach – take an open source generic artificial intelligence model and further train it to meet the specific needs of the company,” Das noted. “Companies could also look for third-party vendors like IBM or C3 and use the complete solution right off the shelf. Or – if they really need to – computing teams can build their own models themselves, from scratch.”
In his recent assignment, ThirdEye worked with an electrical company to detect electrical pole failures by analyzing thousands of images using artificial intelligence. “We started using the Google Vision API and found that it was unable to produce the desired results,” which was 90% or better, Das noted. For example, generic Google Vision templates did not recognize the standard font and the different background colors used in the utility’s pole codes.
“So we took TensorFlow’s basic component models and optimized them for the exact needs of the power company,” Das noted. The team spent two months developing artificial intelligence models to detect and interpret tags on electric poles and another two months training the models. “The results show a level of accuracy of over 90 percent,” Das noted.
Considers the need for industry-specific pre-trained models
A similar view was expressed by the CEO and founder of CrowdAnalytix in their recent financial statements Forbes. “THere’s the catch for Google Vision, just like all generic AIs: These generic models know nothing about the particular industry or organization that uses them, ”said Divyabh Mishra.
General artificial intelligence models are trained in general data sets, which are often publicly available and can be used in many applications in different industries. “The result is artificial intelligence, which is undeniably powerful but very limited in its usefulness to businesses,” he noted.
There is a need for a large library of narrowly trained artificial intelligence applications working in specific vertical industries. “We need models that are pre-trained for large data sets for relatively specific use cases: an artificial intelligence market for enterprise-specific solutions that can be implemented directly by the consumer, without a huge data processing team and without having to deal with additional training,” Mishra said.
CrowdAnalytix operates in a mass resource model, with a community of more than 25,000 “solvers” working on projects that the company calls “competitions”. Its website states: “We leverage our community to create a set of ready-made solutions that are then tuned and customized for each client.”
The New York Times is working with Google Cloud to digitize the photo archive
In an example rooted in Google’s investment in pre-built models, New York Times is working with Google Cloud on a project to digitize its photo archive. For more than 100 years, The Times has archived photos in filing cabinets, three levels down the street near Times Square. The archive now contains five and seven million photographs Google Cloud Blog.
“Cemetery is a treasure book of perishable documents that are an invaluable chronicle not only of The Times’ history, but also of almost more than a century of global events that have shaped the modern world,” said Nick Rockwell, Chief Technology Officer New York Times.
“A well-functioning asset management system must allow users to easily browse and search for photos,” said Sam Greenfield, technical director, Google CTO for Google Cloud Office, author of the message. Google brought artificial intelligence technology and expertise to the table to create a useful system for Times photo editors. The system scans the photo and all text information on the back of the photo, allowing the system to further categorize the photo. A photograph from Penn Station is classified into, for example, travel and bus and rail classifications.
C3.ai provides ready-made artificial intelligence applications for the vertical industry
C3.ai, an artificial intelligence software company founded by Tom Siebel, the founder of Siebel Systems, a provider of human resources software, is copying the artificial intelligence packaged software industry. The company offers C3 AI Suite software, which provides pre-configurable artificial intelligence applications for applications such as proactive maintenance, fraud detection, energy management and customer engagement.
C3, in collaboration with industrial service company Baker Hughes, developed the BHC3 AI Suite for the oil and gas industry with proactive maintenance cases. customer story on the C3 website. Over the months, the team deployed proactive maintenance applications extensively based on the C3 website account. “These apps notify instrument engineers when asset components behave abnormally,” the account noted.
“The combination of computer science expertise and the software development expertise brought by c3.ai is really strong,” said Dan Jeavons, Shell Oil’s Chief Information Officer.
The market is developing well for software vendors and consultants with expertise in applying ready-made artificial intelligence models to specific vertical industries.