By Martín Piqueras, Professor of Obsiness School and member of Gartner, consulting company and research of the ICT -based ICTs in Stamford, United States.
When the AI occupied the first chaatgpt headlines, Sam Ullman himself said the models they had used to develop it were not scalable. That is, developing it had cost a lot of money and a lot of computing capacity, and there would be enough of either of the two things to take it to the next level. That is, then it was identified that it would be necessary for artificial intelligences to evolve in a more economical way that had been used until then, that it was a gross force model.
Since then, multiple organizations have been working thoroughly to develop AI models in a much more frugal way. This is not only desirable for economic reasons, but also for reasons of environmental sustainability; For example, there are not enough raw materials to manufacture as many chips as it would be desired to reach new paradigms.
If we add to all this that Asian giants also have their great technology in constant innovation (the so -called “digital dragons”, such as Tencent, Alibaba and others), we have the perfect formula so that, while some are developing artificial intelligences at once of muscle, computation and carrier, others are trying in a more economical way and, hence, the results we are seeing.
What can happen to the investments made in the United States? Will both models coexist or will one win? In my opinion, there is space for different types of AI: on the one hand, those of muscle, which will need very computing capabilities and that, perhaps, in the future are supported by quantum computing and new paradigms that require less processing capacity; And, on the other, the most affordable and sustainable AIs that can solve smaller problems, for example, operate autonomously on a smartphone, on the computer of a car or in a surveillance chamber.
It is possible that, in the short term, investments as large as have been made in the US are difficult to sustain. However, they will also be necessary for certain types of computing based on brute force, as in the discovery of new medicines or the design of new molecules.