In many ways, the greatest challenge in expanding the adoption of AI is not making it better. It is making the tech accessible to more companies. You usually need at least some of the programming to train a machine learning system. This imposes the rules for companies that cannot justify a data scientist for a task. A data scientist for the task. Google may have a solution: that just revealed an alpha release of Cloud AutoML Vision Its 1st in a set of tools that trains AI without needing a code. This 1st service trains image recognition systems using a drag-and-drop interface. You just need to upload photos, tag them and start the training process.
As mentioned at AutoML’s preview back in May, Google is actually using “baby” neural networks to develop these systems. It iterates the mini nets with reinforcement training and picks the best one from the bunch.
Not surprisingly, it costs you money: you need to apply for the training, and you will be charged for the training of models and access to them. You will not use it to engage in a hobby. However, that ensures to make AI, and specifically the image recognition, many more accessible. While there are already custom AI options (Microsoft’s well trained AI models for you). Google’s approach is simple and hands-on enough that your favorite website or device manufacturer can take AI with their products in a relatively small effort.
Some practical examples already exist. For example, Disney uses the Cloud AutoML which helps search for products based on its store based on what they look like, not arbitrary tags. You can also find that Buzz Lightyear toy even if it has been miscategorized. Conservationists at the Zoological Society of London, meanwhile, are believed to automatically categorize animals that pass through wildlife cameras. While there will still be a need for advanced, manually programmed AI, It cannot be as essential as it is used.
