AI with machine learning models are good to solve common problems. But to get the most out of this technology, custom models are needed. This is where Google wants to help.
The tech giant has several AI under its sleeves. And during its Cloud Next conference in San Francisco, it has revealed the beta version of AutoML Vision Tool.
What it does, is allowing developers with no knowledge of AI, or even coding fluency, to create machine learning models for image and object recognition.
Google made the tools legible to those outside the software engineering and AI fields by using a simple graphical interface and common well-known UI like drag and drop..
AutoML Vision that leverage Googles' cloud computing offering, was first announced during Google's I/O conference. After existing its alpha period, the company is making the tool available to everyone.
Google revealed that since January 2018, around 18,000 people have expressed interest in AutoML Vision.
The idea behind Cloud AutoML, is to provide organizations, researchers, and businesses who need custom machine learning models, the simple and easy way to train them, Google said.
With that in mind, Google expands AutoML to include AutoML Natural Language for predicting text categories using natural language processing and AutoML Translation, which allows users to upload their own language pairs to achieve better translations for texts in specialized fields.
Along with AutoML Vision, all three are available in public beta.
"AI is empowerment, and we want to democratize that power for everyone and every business — from retail to agriculture, education to healthcare," explained Fei-Fei Li, chief scientist of Google AI, in a statement. "AI is no longer a niche in the tech world — it’s the differentiator for businesses in every industry. And we’re committed to delivering the tools that will revolutionize them."
Besides the Cloud AutoML services, Google is also updating existing APIs, which include:
- Cloud Vision API, so they can recognize handwriting, support PDF and TIFF files, and recognize where an object is when located in an image.
- Cloud Text-to-Speech to include multilingual access to voices generated by DeepMind WaveNet technology and the ability to optimize for the type of speakers.
- Cloud Speech-to-Text so it can identify what language is spoken as well as different speakers in a conversation, word-level confidence scores, and multi-channel recognition.
Going beyond the above, Google also unveils Contact Center AI, which is a machine learning-powered customer representative built with Google’s Dialogflow package that is able to engage with callers over the phone. The company is marketing it as a toolkit for conversational agents.
When deployed, incoming calls are scanned by Google's natural language processing to suggest answers to common problems. If the agent can't solve the callers' problem, it will forward the callers to its human agent, by presenting the agent with all the relevant information to the call at hand.
Rajen Sheth, the director of product management for Google Cloud AI, said that the extension of AutoML is yet another step toward the company’s vision of democratizing AI. "What we are trying to do with Cloud AI is to make it possible for everyone in the world to use AI and build models for their purposes," he said.
While Cloud AutoML can certainly help developers create custom machine-learning models they need, Google didn't exactly detail how the company employs the methods. What this means, putting aside Google's Cloud AutoML ability to create unique models with ease, users are left with just a standard AI tool.
In other words, Cloud AutoML is just another AI training tool available in the market.
Google should at least be more transparent about how well its techs perform. This is considering that others have also opted for a hybrid approach that allows users to upload their own data to customize the existing models.