web analytics
Home » Technology » Expensive AI Development To Save Moore’s Law

Expensive AI Development To Save Moore’s Law

Chip development is being supported so strongly by the AI ​​boom that Moore’s Law, which has been declared dead several times, will remain in place for a long time to come. At least one of the most renowned experts in the IT industry is convinced of this.

Developing and training AI models is very expensive. The new GPT-4 probably cost around 100 million dollars. This was mainly provided by Microsoft in the form of resources in its data centers, in which tens of thousands of graphics chips worked on the evaluation of training data for months. GPT-5 will then probably be much more expensive again, with costs in the billions already being expected.

This is an approach from which AI development will give a new impetus to chip designers. Because the development of new processors that follow Moore’s law is also eating up more and more money. So users are needed, for whom the higher performance brings enough advantages to be able to cope with the significantly more expensive purchase prices. This is certainly the case with AI development, as Robert X. Cringely – behind whom the industry expert Mark Stephens is hiding – explained.

Incidentally, this effect is just right for corporations like Microsoft and Google to keep the competition at bay. After all, a startup cannot afford the high expenses for better processors and AI development in general. The financially strong large companies can thus build up a trade barrier without violating antitrust law.

AI sees through chip designs

The second side of the story is, of course, the contribution that AI systems can make to chip development itself. The architectures have long been very complex, but still follow simple logical patterns. For the AI ​​algorithms, they are basically ideal to see through. Cringely is therefore convinced that there will soon be a significant upswing in the use of AIs in the development of chip designs.

In his view, significant impetus will come from China in particular. Because the country is banned from the most advanced technologies from the West by embargo measures, it is not easy to build up your own skills to play at the top. However, it is quite easy for the Chinese state to invest a lot of money in AI hardware, which then produces algorithms that can even make large leaps in chip design.