As part of my role at Verhaert, I delivered an internal talk on transformers. The session began with a brief overview of neural networks before delving into the theory and implementation details of transformers. I explained the inner workings of transformers and their applications in various domains, including natural language processing and computer vision. Additionally, I discussed how we can leverage pretrained models from HuggingFace to enhance our projects for clients. It was an engaging session that explored the power and potential of transformers in the field of AI.
In the newspace era, earth observation data has become commercially available for everyone. Numerous opportunities now present themselves through this data. From digital farming to arctic shipping, lots of industries stand to benefit from this. In making these opportunities a reality, AI is playing a crucial role.
The release of ChatGPT has been dominating headlines and rightfully so. The latest chatbot by OpenAI is so good, you would sometimes forget you were chatting with a computer. The core technology behind this marvel? Reinforcement Learning from Human Feedback! In this blog, I explain how this technique can be applied to language tasks.
Continuing on the theme of sustainability, my colleague Jente and I wrote a perspective sharing different techniques to reduce the energy consumption of neural networks. The key takeaway? Many different techniques exist, but we should prioritize using them from the beginning of the development cycle.
Join me and my colleague Bart as we discuss the current issues of AI, particularly deep learning technologies, due to their high power consumption and associated costs and carbon emissions. We explore new techniques like resource-aware AI to address these challenges and pave the way for energy-efficient AI systems.
Performing maintenance is critical to keep appliances, vehicles, manufacturing machines, etc. working correctly. By employing artificial intelligence, you can gain better insights into these machines. With these insights, you can optimize performance, ensure quality, and even predict when the machine will break down!
The backpropagation algorithm can be quite unstable. In this blog post, I discuss some improvements, including the use of minibatches and the Adam optimization algorithm, to enhance the training process in deep learning models.
The backpropagation algorithm is a key factor in enabling deep neural networks to learn. In this blog post, I dive into the inner workings of backpropagation and explain how it helps optimize the parameters of deep learning models.
Deep learning is a powerful technology driving many advancements. In this blog post, I dive into the fundamentals of deep learning, exploring the concepts of neurons, layers, and activations. Through Python coding examples, we'll gain a solid understanding of the basics.
An engineer at Google claims that an AI has become sentient. In this blog post, I discuss why this claim should have us concerned, but not for the reasons that immediately come to mind. Join me as we explore the implications of the "Sentient AI" narrative.
Have you ever wondered how some intelligent products excel at specific tasks? Are you interested in applying continuous learning to your own products? In this blog post, we explore the world of continuous learning and provide a comprehensive roadmap for implementing self-learning capabilities. Whether it's a simple lamp or a complex modern car, our guide has you covered.
Earth observation systems generate massive amounts of data, which can be both a blessing and a curse. In this talk, I discuss with my colleague Daniel the unique role that data science, machine learning, and artificial intelligence play in making NewSpace data actionable. Discover how AI can revolutionize earth observation and enable us to extract valuable insights from the vast amounts of collected data.
In this blog, we explore the concept of hybrid AI and its role in distributing AI models between edge devices and the cloud. Discover how deep neural networks can be leveraged to achieve a balanced approach, allowing for efficient and effective AI processing at both the edge and in the cloud.
Discover how to make smarter decisions when designing AI for smart home devices. This blog provides a 4-step guide for choosing between edge and cloud intelligence, enabling you to create devices that can function even when the internet connection is unavailable. Explore the benefits and considerations of both approaches and gain insights into optimizing your product's intelligence.
Learn how edge AI can revolutionize smart homes and appliances. This talk, presented during the Innovation days at Verhaert in 2021, explores the limitations of current smart devices and appliances, where the cloud is responsible for most of the intelligence. Discover how edge AI can make devices smarter, enhance their capabilities, and address privacy concerns. Watch the talk to unlock the potential of true smart homes and appliances.