Today, artificial intelligence has penetrated every aspect of our lives. It is now hard to complete a day without hearing any news about artificial intelligence. On the side of the companies, this technology is important and great enough to justify hundreds of billion dollars of investments. So how did we get to this point? Of course, thanks to Alexnet, which was developed in 2012 and started a deep learning age. Now, Alexnet’s original source code, representing this historical turning point, has finally become public after years of negotiations.

This important development took place thanks to the cooperation between the Computer History Museum and Google. This revolutionary code, which was written by Alex Cislikhevsky and first introduced in 2012, is now open to access via Github.

He had it before chatgpt

Alexnet has gone down as a nervous network that radically improves computers’ ability to recognize and classify. In fact, fundamental theories such as the idea of ​​neural networks and backpropagation algorithm have been known for decades for decades. However, the two critical elements were missing: Hardware that will perform a huge amount of processable data and complex calculations quickly and efficiently to train these neural networks. Initiatives such as Stanford’s Imagenet project and Nvidia’s Cuda GPU programming finally met these important elements.

On the other hand, Alex Cislikhevsky, the leading names of artificial intelligence Nobel Prize -winning AI scientist Geoffrey Hinton and later established with the OpenAI Ilya Sutskever developed Alexnet. Alexnet in the Imagenet competition held in 2012, was by far the first place compared to other methods. This success proved the potential of deep learning and opened the door of a new era in artificial intelligence research. Alexnet was an unseen depth (8 layers) until that period (Convolutional Neural Network-CNN) and only two GTX 580.

The CNNs had shown promising beginnings in performing tasks such as recognizing handwritten numbers until then, but did not lead to transformation in any sector. However, they developed the design of neural networks, including CNNs, including Alexnet, and found that increasing the number of artificial neuron layers on the Nvidia GPU chips would lead to theoretically better results. Sutskever thought that when sufficient horsepower and educational data were provided, the theoretical work could be scaled to a much larger nervous network. As a matter of fact, it was. Sutskever is now working at SSI and argues that a new method is necessary, saying that this scaling is over.

However, this success of Alexnet at that time did not immediately create great excitement. While some researchers in the artificial intelligence community evaluated Alexnet as an ordinary development, Meta Platform’s Chief Artificial Intelligence Scientist Yann Lecun, who was known as a legend in the field of deep learning, who had previously learned from Hinton and pioneered CNN Engineering in the 1990s, was one of the first names that realized the importance of this study. Lecun predicted that Alexnet would be a turning point in artificial intelligence – and he was not mistaken.

Thanks to Alexnet, there have been revolutionary developments in many areas such as image recognition, speech synthesis, music production and creative writing. Chatgpt, which was released in 2022, is considered the summit of this evolution. Now the open source of Alexnet, which is extremely simple for today, is an important situation in terms of transparency and information sharing in artificial intelligence research. It is also worth mentioning that this process is not easy. Computer Historical Museum conducted challenging negotiations for five years with Alex Cricfrey Hinton and Google’s legal team. Because Alexnet’s intellectual property rights belonged to Google because Google had acquired Dnnresearch, the venture company of Google, Hinton, Crichevsky and Sutskever.