The typical non-tech companies have nowadays started to explore how they can use deep learning. Well, no restriction to them. The technology isn’t just for titans.
Deep learning, which is a part of a broader family of machine learning methods, recently drove Wisconsin State Cranberry Growers Association to approach a digital advertising company about a project to use the technology to identify pests in cranberry bogs. Initially, it was considered with skepticism because neither an advertisement firm is a right company to handle such a project nor did growers have the technical background to build this sort of deep learning capabilities.
But then Wisconsin State Cranberry Growers Association itself revealed it. It felt that the technology was worth giving a try as the company’s vice president of engineering, Sam Saha said. According to him, there are a number of businesses out there and, they don’t have a tech team to use AI, but still the technology can be used.
Deep learning for SMBs
Most of the use cases of deep learning (or machine learning and artificial intelligence) are introduced to the tech world by tech giants like Amazon, Facebook, and Google. And it continues to happen, which, of course, has created a perception that the advanced technology is just for leading big companies. But now some small and medium businesses have started to prove this perception wrong.
Wisconsin State Cranberry Growers Association’s approach to the project is a proof of concept that shows that how companies with a non-tech background can also carry on and out AI and machine learning implementations.
The team at the company began with building an iPhone application that helps cranberry growers to click pictures of suspected pests.
Initially, the captured images were labeled by employees at Ocean Spray, which was one of the major buyers of cranberries with its largest interest in the health of bogs. The company then got a deep learning model built to help it verify pests in the pictures taken and sent by farmers.
It’s one good example on how AI’s subset based tools, like the deep learning, can also be used by less tech-savvy organizations which have no prior experience in handling or executing AI and ML technologies. Also, it shows how the company took the help of open source tools which means you just don’t need spending like big companies do in developing proprietary technologies.
The deep learning model that the company used in its application is already available through TensorFlow, which is an open-source software library for dataflow programming across a range of tasks. All it needed was training.
So, this is one easy way to get a use case of high-end technology. But things on grounds are contrary to it. First of all, not many enterprises are taking this sort of deep learning use case approach. Instead, there is a common belief that when it comes to DL or ML, companies will need writing models from scratch and, if you are not doing it this way, you are not doing a real AI. It’s what impeded businesses to have benefits of AI and its tools.
The other argument is that businesses are often worried about whether they really need AI. They say they aren’t Google or Amazon and, they don’t know where they start. So, they give up even before attempting solving a problem with it.
AI for the construction industry
Another deep learning use case was brought into light by Bechtel Corp, which is a San Francisco based engineering, procurement, construction, and project management company.
The company is working on a project which uses the reinforcement models to find the fastest route to development. These reinforcement learning models are similar to those which are used by AlphaGo, an AI model that beaten human champions of the game Go in 2016.
The model, on which Bechtel Corp is working, runs a step-by-step simulation of projects, tests out sequences of laying concrete, and installs pipes to find the best sequence.
The construction industry carries a number of things which have historically been less dependent on technologies. Because each project is different and, people have to work in an environment, where there is no set of training data from past projects that a machine learning application can use. Here they are using the reinforcement learning in which simulations actually become the training data sets.
These two are the latest examples how AI and its subset technologies & tools can be used even by those companies, which essentially don’t have a tech foundation.
- Image credit – zdnet.com
- Image credit – ukconstructionmedia.co.uk