Machine Learning and AI has become the number one priority for big players like Google as we move from a mobile first world to an AI first world. When we think about the last five for ten years, large organizations and industry players were not really considering mobile seriously.
In fact the success of any Mobile App Design Company right now is linked to the fact that the big players were not taking mobile seriously and it swept them away by storm. We can hope we don’t make the same mistake with AI as well.
Mobile App Development and AI Key Factors
Machine learning is already a reality and we see it everyday form image recognition in Google Photos to content recommendations on YouTube and other 1000plus projects at Google that are already using neural networks. You must probably be thinking:
Yeah, that’s amazing but how does this apply to me?
I am not Google, I don’t have the resources.
I don’t have the skills to do it. How should I get started?
This blog presents a slightly different angle and tries to give insight on how as mobile app developers, other business owners, we should be approaching machine learning and what are the initial aspects that you should be thinking about before taking on a machine learning project.
Consider the Problem
First of you have a problem, this might sound a bit silly, but this is one of the most common mistakes that we observe in the industry. People are looking at how great the AI potential is and try to randomly apply this to their business.
The right way is to assess your problems, not the simple problems but in fact we are talking about the mission critical problems and try to see how machine learning can help you solve these problems.
Applying efforts to find problems is really important as with machine learning you might end up building a great model or a successful machine learning feature, but in the end it is useless because it is not solving the problems that exist within your company.
Consider the Data
Once you have the problem figured out you need data. You need a lot of quality data large volumes of it, you also need to be able to process it and access it at scale.
This means that you might have to go through the process of improving infrastructure, you have to clean your data and you might also have to think about placing the data into categories or labels.
Consider Skill set
The next step is to identify the people that are going to execute this project within your company and if you don’t have anyone please don’t fell discouraged. Nowadays there are a lot of resources available online to everyone to train your teams, resources like Coursera, Udacity, YouTube, GitHub and many more.
But in an ideal world the AI infrastructure tends to recommend is the combination of two types of skill. At one hand we have data scientists, the people that are able to conceptualize mathematical models taking into account the requirements of the initial problem.
On the other side we have data engineers, the people that translate all of the information into code, are able to train the model and run it into production.
If you are a growing business it is encouraged that the more you grow the more you should think about structuring your skillset into one centralized team that can then defuse their knowledge into the different divisions and processes of the organization.
Consider an Approach the Fits Requirements
A lot of companies in the machine learning field believe in what is called the open research. Big corporations like Google and DeepMind believe in open research. This means that a lot of research findings, massive libraries of developed models are published for free and are available online as open source.
Considering all of this, as a developer you have to choose the extent to which you would want to rely on existing models and to what extent you would actually develop your own models in the execution of a certain project.
Sure at the beginning of the project it is encouraged to tap into as much as you can into existing libraries but down the road you might realize that this doesn’t exactly fit into your requirements. So you might have to train your own AI model with your own data or even build your own custom model.
Consider the Best Solution to Run Your Model
Every AI app development project is a trade-off between simplicity and flexibility that helps determine what kind of technological solution you are going to be using. There are amazing custom models and already developed models that can work for you.
If we take the example of Google, Google has different pre-developed AI algorithms solutions that one could consider. Solutions like Google Cloud’s machine learning API that can give you Cloud Vision, speech, translate and natural language APIs, as well as Action Google. If this is not enough one of the most commonly used frameworks used by developers is TensorFlow.
This tool has become the most widely used resource for the development of AI projects. This tool can be used to build custom models, model training, and you can run it on different processing units and on different cloud platforms. There is also a lite version available for this tool so that it can be run on mobile devices.
The point being there some amazing already built solutions available for mobile developers that can easily handle data processing and development, so one does not have to look far. The thing of utmost importance is to know the extent and need of any AI project at hand and the rest becomes achievable.
