At the time of writing this post (its Polish version was created in April/May 2023), there’s still quite a widespread excitement about the possibilities of large language models. These possibilities were spectacularly showcased to us by a solution released at the end of 2022 by OpenAI, and our world will never be the same again. …
In 2023, we are witnessing a boom in language models and their practical applications. ChatGPT has sparked interest in replicating its success, and many teams have published the results of their work. A large portion of the new models have been released under the Apache 2.0 license, which allows for their free modification, use, and …
When in 2019 I started this blog, I allowed myself to be a little controversial by writing “AI will change the world more than the industrial revolution.” Of course, prediction is very difficult, especially if it’s about the future (Niels Bohr), but recent developments in machine learning have made me think about how AI will …
In machine learning designing the structure of the model and training the neural network are relatively small elements of a longer chain of activities. We usually start with understanding business requirements, collecting and curating data, dividing it into training, validation and test subsets, and finally serving data to the model. Along the way, there are …
Looking at the advances in technology over the past few years, it’s hard to name two more breakthrough technologies than artificial intelligence and blockchain. The former opened up entirely new possibilities in the fields of data analysis, predictions and robotics. The later one elevated decentralization, transparency and security as a result of the immutability built …
Until recently, a large part of the key concepts in the field of artificial intelligence was not so clearly defined. Some of them, such as Deep Learning, were even referred to as “buzzwords”, term used mainly by marketing and not strictly translated into scientific areas. Now, the basic concepts seem to have taken hold, and …
In the previous three parts of the tutorial, we learned about convolutional networks in detail. We looked at the convolution operation, the convolutional network architecture, and the problem of overfitting. In the classification of the CIFAR-10 dataset we achieved 81% on the test set. To go further we would have to change the architecture of …
Convolutional neural network is one of the most effective neural network architecture in the field of image classification. In the first part of the tutorial, we discussed the convolution operation and built a simple densely connected neural network, which we used to classify CIFAR-10 dataset, achieving accuracy of 47%. In the second part of the …
Deep neural networks are widely used in image and shape recognition. Examples of applications include face recognition, image analysis in medicine, handwriting classification, and detection of surrounding objects. A special type of neural network that handles image processing extremely well is a convolutional neural network. I have to admit that ConvNet is my favorite deep …
Shape recognition, and handwritten digit recognition in particular, is one of the most graceful topics for anyone starting to learn AI. There are several reasons, but the two most important are the ease with which we can use well-prepared ready-made datasets and the ability to visualize these data. From this tutorial you will learn: Okay, …