Making AI accessible to everyone
The coronavirus pandemic has clearly accelerated our dependency on technology, online activities, and artificial intelligence. AI is particularly important for businesses as it enables personalized services on a massive scale, and customers are increasingly demanding it.
However, not every company has the knowledge or the tools to implement AI, nor do they know what is required from them to become AI-driven. In this post, I will discuss what options these companies have.
It is important to note that while many of the methods described below assist no-coders, they are also suitable for developers, who can enjoy the extra development speed they bring in.
Ever since I was learning to program, the idea of developing a tool that could create applications with plain English commands was floating around. Many years later, OpenAI’s GPT-3 managed to get quite close to this idea as we saw demonstrations of code and HTML markup being written by the text generator.
GPT-3 stands for Generative Pre-trained Transformer 3, which demonstrates the idea of training an AI on colossal amounts of data, then using that built-in knowledge to get stunning results for new tasks with little or no training. GPT-3 was trained using huge amounts of data including, among others, Common Crawl and Wikipedia. But more importantly, it was on supercomputers which enabled it to amass 175 billion parameter values, making it the largest AI model developed to date.
This has enabled the AI to use its current learnings and transform it to apply to other tasks. Transformative AI has many advantages, as it takes much less time to train and gives a head start compared to developing from scratch. It also makes AI much more accessible: companies only need to share their specific data with the model to make it their own. For instance, Anyline’s no-code AI trainer helps companies build their own text reader solutions (such as ID scanners or license plate readers). Customers simply upload their data into the trainer, which automatically tunes the neural networks for them to produce a customized OCR scanner.
Users do not need to learn how the system works or what the source code and architecture of the application look like—all they need to do is to feed the data they want intelligence on, and the AI adjusts accordingly.
Of course, some degree of AI knowledge is still necessary. According to Drew Conway’s Data Science Venn Diagram, effective development and implementation of AI requires two important skills: hacking skills, and math and statistics knowledge. Without these components in place, companies risk developing an AI that works well in lab settings but fails in when faced with real-world problems.
