Artificial intelligence (AI) is experiencing a renaissance; it’s no longer a hot buzzword — it’s getting integrated into our favorite apps, powering our productivity tools, and leading the vision of the future for tech giants like Google, Apple, Microsoft, and Amazon.
In 2017, after its rebranding as an “AI-first” company, Google followed up with an announcement that it’s been using AI to build other AI. It also announced a strategic plan to open an AI center in Beijing, China’s tech capital. Continuing with this trend into 2018, the tech titan announced its newest AI innovation: an easy-to-access interface that anyone can use to create their own AI.
The product, called AutoML Vision, is a cloud-computing platform that creates a unique computer vision system for your needs. And it’s currently open to the public. Google touts this technology as having endless possibilities: radiologists can train an algorithm to find signs of lung cancer; realtors can teach an algorithm the difference between a bedroom versus a living room to speed up image captioning; retailers can use an algorithm to classify shirts as V-neck, scoop neck, crew neck, and more. London developers at the Zoological Society are already using Cloud AutoML to track images from camera traps in the wild.
At the moment, Google’s only offering computer vision on their platform. But their plan is to offer more technology for speech, language, video, natural language processing, and others in the future.
With Google’s new product, businesses can use the platform to create fine-tuned algorithms focused on solving their particular needs. It’s perfect for small businesses with low budgets and unique use cases as well as students interested in machine learning. Of course, humans must manually label the data so the system can learn from it. For businesses with larger budgets, Google offers human labelers. Your AI won’t be very accurate without you feeding it at least a few hundred hand-labeled images.
Not Truly Automated… Yet
Right now, the main barrier for companies of all sizes is budget. Algorithms will eventually automate most of our basic work and play a bigger role in our lives. But right now, it’s often too expensive or time-consuming to apply AI automation, especially for small businesses.
Automating data science and opening machine learning up to the public is a grand idea, but when humans still have to do so much work to get the algorithm to work accurately, it is often cheaper to just keep having humans do the work manually. Google, in this case, is training and tuning the model without any extra effort on your part except for labeling images.
Right now, though, Cloud AutoML is only being released to developers. There is no specific pricing for business-level AI, but it’s likely there will be two fees: one for training the algorithm and another for utilizing Google’s API. The higher the strain on the API, the higher the pricing structure, traditionally.
We don’t know how accurate Google’s cloud-computing algorithm platform will be when it launches or what sort of business value it’ll create for business owners. Google promises higher accuracy with its inclusion of advanced machine learning techniques, like transfer learning, learning2learn, and neural architecture search technology.
Supply and Demand
Diane Greene leads Google’s cloud-computing group. She encourages experimenting and urges, “You can still build a highly accurate machine learning model” without “a Ph.D. in machine learning.”
And while we’re on the subject of Ph.D.s in machine learning, most of the 10,000 worldwide experts in machine learning have been hired by Google, Amazon, and Microsoft. Frankly, most businesses cannot compete with the tech giants’ business tools.
There is a much higher demand for AI experts than the supply of existing experts. Google’s chief scientist for AI and machine learning, Fei-Fei Li, elaborates, “AI and machine learning is still a field with high barriers to entry that requires expertise and resources that few companies can afford on their own. Today, while AI offers countless benefits to businesses, developing a custom model often requires rare expertise and extensive resources.”
This opens a great business opportunity for the tech leaders: sell AI to other businesses. Microsoft’s already selling a computer vision algorithm that can apparently be applied to anything and integrated into nearly anything.
An AI Democracy
And although Google’s Cloud AutoML isn’t the most unique, first of its kind, or even the largest technology of its type on the market, it is still working hard to bring AI to the masses. Maybe it thinks new experts will be born from the product’s audience. Li writes that her main mission at Google is to democratize AI. She plans to accomplish this goal by lowering the barrier of entry and opening AI tools to researchers, developers, and businesses.
Whatever Google’s end goal is will inevitably affect the evolution of AI. Even Li admits that Google’s only just starting to make AI more accessible and that more work needs to be done. But right now, we’re excited that Cloud AutoML is trying to level the playing field. Google’s new AI release helps bridge the gap and open a dialogue about AI between enthusiasts and experts.
As machine learning and AI penetrate the mobile app development market, it’s important that consumers have a better understanding of the algorithms driving their favorite app features. Request your own access to the platform through this link, and let us know in the comments what you’d train your Google computer vision algorithm to detect!Tags: AI, AI App Developer, AI App Development, AI applications, artificial intelligence, AutoML, business tool, deep learning, democratizing AI, enterprise solutions, Google, London app developer, machine learning, small business, tech, tech news, technology advancement