As the industry moves towards no-code artificial intelligence model training platforms, AI is being positioned as a tool for the masses. It, therefore, comes as no surprise to find that the industry bellwethers are releasing no-code or low-code platforms to build custom machine learning models that can be used with ease and security.
Joining the big tech giants is China’s internet major Baidu that launched an AI platform designed to make building custom ML models easier and it rules out the need for algorithmic programming. Known as EZDL, the service platform enables developers to build custom ML models with a drag-and-drop interface, Yongkang Xie, tech lead of Baidu EZDL, said in a company statement. He further emphasised developers can build deep learning models which are specific to their business needs only in four steps.
- There are two reasons for this — firstly, these platforms combine intelligent, decision-making algorithms with data, which enables developers to create a business solution. These algorithms have functionality for image recognition, natural language processing, voice recognition, recommendation systems, and predictive analytics.
- The platforms offer pre-built algorithms and simplistic workflows with features like drag-and-drop modelling and visual interfaces that can easily connect with data and accelerate bringing services/applications to the market.
- Secondly, building low-code or no-code ML platforms allow market leaders in this space like Google, Microsoft, AWS, IBM and Salesforce to emerge as the central AI ecosystems and will further build alliances to scale the use-cases and delivery as well, HfS research indicates.
- Another perspective is that low-code AI platforms can help organisations and startups which lack a formalised AI practice build applications and solutions quickly without any coding experience.
- From a developer’s/researcher’s perspective, AI platforms tackle some of the most pressing challenges in the workflow, for example, manual assignments or time-consuming tasks which can be automated. For example, data scientists are required to perform multiple experiments to compare the performances between different models and to find better hyperparameters.
- Over time researchers end up changing the model structure or tweak the hyperparameters to improved model performance. With AI platforms, it is easy to experiment with hyperparameters. For example, AutoML is able to predict the performance of experiments based on previously run experiments and automatically optimises the hyperparameters based on the performance predictions.
- AI platforms allow for execution of the entire ML pipeline, beginning with data preprocessing to finding the best model, whenever the dataset is uploaded
- These no-code or low-code AI platforms are geared towards improving the developer’s productivity with smart suggestions for application logic flows. There are also built-in checks for consistency and quality, such as application scalability
- Companies which lack formalised AI practices are following a project-centric or incubator type approach can make use of these platforms to work on domain-specific use cases
- Platform approach also increases agility and provides a simple, user interface to business and technology teams
- So, what’s in it for IT bellwethers AWS, Google or Microsoft? One analyst pointed out that the launch of no code AI model building platforms is a step towards AI democratisation. But more than that, it helps companies win over programmers or end users who will build their own models.
According to Mendix, a leader in low-code platforms and Mendix Assist provides AI-assisted development approach, a low-code environment can resolve many technical challenges. From agile software creation to user-friendly interfaces, these platforms foster greater collaboration among data scientists, developers and IT teams. They also resolve the skills shortage and provides the business and technology teams with a collaborative user environment and also increase the agility across the teams.
The expanding ecosystem requires an interplay between computing power, a set of algorithms and platforms that are geared towards domain-specific use cases. In most cases, use cases determine the technical approach and it is here that AI platform solution providers differ from the rest. For example, Google’s AutoML is known for its accuracy in image recognition tasks. Another case in point is EZDL is that currently focuses on the three core aspects of ML tasks — image classification, sound classification and object detection. These tools are largely aimed at talent-starved startups and organisations that want to get started with their AI use cases and build their own models.
A report from ResearchAndMarkets forecasted the low-code application development market will surge from $4.32 billion to $27.23 billion by 2022 and grow around 45 percent annually during the period. According to a Forrester report, the current size of the low-code market is around $4 billion. The market for low-code development platforms is expected to grow by 55 percent by 2020. Besides, rapid adoption of low-code platforms will drive an annual growth by 50 percent.
Try deep learning using MATLAB