Blog - Articles around AI, ML, DL and more

For an internal product prototype we have traced OpenAI’s Whisper 3 model from Huggingface and made it usable under JAVA via DJL.

In today’s digital business world, AI-powered communication platforms like ChatGPT are essential for tasks such as answering complex code questions or creating top-notch texts for offers. However, in companies dealing with sensitive customer data, using ChatGPT can lead to a data protection dilemma. While ChatGPT offers an option to prevent the use of chat conversations for training purposes, it comes with certain limitations. Moreover, as of June 2023, there is no way to manage multiple team members or users through a company account. Each user must register individually and use their own email, phone number, and credit card. If you want to use ChatGPT+, for example, you cannot pay for all users with one credit card. Individual invoices also end up with individual users, creating an organizational and accounting nightmare. We at DIVISO have also grappled with this issue and went in search of a solution.

In this part of the series of articles on MLOps, we start with information that will be familiar to most of you: With the basics of Git. However, to give a different perspective on the well-known tool, these basics provide the basis to highlight the function and benefits of Git for machine learning (ML) and the difference in managing training data.

With Machine Learning Operations (MLOps) we ensure that data is efficiently and strategically integrated into business processes through regular and automated training, thus contributing to increased revenue. The challenge is to establish and maintain these automated processes.

In our real-world example, we used a “feed-forward neural network” to recognise handwritten numbers. This is probably the most basic form of a NN. In reality, however, there are hundreds of types of mathematical formulas that are used – beyond addition and multiplication – to compute steps in a neural network, many different ways to arrange the layers, and many mathematical approaches to train the network.

Developers who wanted to explore neural networks and deep learning using the JVM, and especially Java, had little choice so far. Those who wanted to focus exclusively on Java could not get around DL4J until now. If it had to be the JVM, but not necessarily Java, the MXNet Scala Frontend was also an option. Finally, if a little Python didn’t scare you, you could try a hybrid solution, combining TensorFlow and Java just like we already explained in previous articles.

So far, we have generally steered clear of the areas of text comprehension and text generation by ML in our practical examples for the basic understanding of AI. For good reason, we have focused primarily on two types of problems: classification of images and prediction of numerical values.

AI and especially Neural Networks or Deep Learning have been the technological hype topic for some years now. However, since the subject is quite abstract – one could say it is uncharted territory for most people – we want to clear up some mistakes that we often encounter in our work.

In our past articles we mainly covered the basics of current AI research and tried to shed some light on them in a way that is understandable for non-IT scientists. We are now proceeding to the probably “hottest” current AI topic: Neural Networks (NN).

After our first presentation of Amazon’s new Deep Learning Framework for Java, DJL, we now want to introduce the basics of Deep Learning under Java with DJL step by step in a series of beginner posts. This is not about quickly copying code snippets, but about really understanding the framework and the concepts.