The premise behind Google's Cloud Machine Learning Engine and TensorFlow technologies is to democratize access to machine learning tools and technologies. Additionally, these products are able to be implemented without the help of a PhD-educated data scientist.
At the 2017 Google Cloud Next conference in San Francisco, a breakout session explained how a host of companies in various industries are using machine learning tools. Here are seven companies that implemented Google's machine learning tools to solve problems in their business.
AXA is an international insurance firm. In car insurance, 1% of car accidents require payments over $10,000 from the insurance company, also known as "large loss" car accidents. AXA used machine learning to better understand which drivers would be likely to cause these types of accidents, raising their prediction accuracy up to 78%.
2. Airbus Defence & Space
The issue faced by Airbus Defence & Space was to be able to detect clouds in satellite imagery. The problem existed for around 20 years, and required manual processes to properly tag images. Using machine learning, they were able to decrease the error rate of detected clouds from 11% down to 3%. The company also leveraged the Machine Learning Engine GPU accelerator to speed up the process by 40x.
3. Global Fishing Watch
The mission of Global Fishing Watch is to prevent overfishing around the world. Their previous approach had analysts monitoring small regions to determine what fisheries were being overfished. The company relied on satellite AIS positioning to watch the entire ocean, tracking fishing vessels and using machine learning to determine if they were acting illegally.
SparkCognition used Google machine learning tools to create DeepArmor, a malware detection tool for Android and Windows. The team built classifiers for what constitutes malware, and uses machine learning to go through and determine if malware is present in a device.
One of the largest financial services company in Japan, SMFG partnered with a Google Cloud provider in Japan called JSOL to build a machine learning-based credit card fraud detection tool. The deep learning tools used for fraud monitoring achieved 80-90% accuracy in detecting fraud and alerting the company to take action.
As a food quality and safety company in Japan, Kewpie's problem was properly detecting defective potato cubes that would be used in a stew. Due to the irregularity of food, it was difficult to automate the detection process, meaning they had to rely on skilled workers for hours at a time. Kewpie used Google tools to build a "worker friendly inspection AI" that monitors a video feed and alerts the inspectors when a defect is found, so they can remove it.
7. AUCNET IBS
AUCNET IBS is a car auction service in Japan. The company relies on multiple photos for each vehicle, and they were previously sorted and categorized manually. AUCNET IBS built an image classifier that detects the model of the car and the parts featured in the photo with 95% accuracy.
- Machine learning: The smart person's guide (TechRepublic)
- Should Google be your AI and machine learning platform? (ZDNet)
- Google AI gets better at 'seeing' the world by learning what to focus on (TechRepublic)
- How to Implement AI and Machine Learning (ZDNet)
- Google Cloud adds NVIDIA Tesla K80 GPU support to boost deep learning performance (TechRepublic)
Conner Forrest has nothing to disclose. He doesn't hold investments in the technology companies he covers.
Conner Forrest is a Senior Editor for TechRepublic. He covers enterprise technology and is interested in the convergence of tech and culture.