2022 was the year in which artificial intelligence took a very important place in the technology industry, however, the source that feeds these digital platforms could have a limit taking into account the large number of uses and requirements to which it is seen. subjected.
An investigation of Epoch AIan organization that studies the development of artificial intelligence, ensures that 2026 is the maximum year for which the current high-quality data centers are designed, which gather the data for the creation of content from these technologies.
This generates an alert on platforms such as ChatGPT, DALL E 2 and Midjourneywhich use that set of information to generate their content via text and machine learning.
It may interest you: Artificial intelligence is the word of 2022, for what reason
How AI data could run out
For an artificial intelligence to create an image, text, video or audio, it needs an instruction, which is mostly written, and with that information it takes data from a base or center, called data setto generate the content.
The problem with this is that the data is being consumed very quickly and is not produced in the same way. In fact, the researchers say that the demand for high-quality data sets to train models of AI language grows 50% each year, while the generation of those sets does 7%. An imbalance.
The collection of information for these data set it is done publicly and on a large scale so that the platform learns correctly. In addition, there are humans involved in the process because there is an important filter to ‘clean’ the data manually and to respond appropriately to user requests.
Those in charge of the study assure that this is a slow and expensive process, and although there are tools, such as the artificial intelligenceusing them to review the models carries a high risk rate that can make the process even more complicated.
2026 would be the limit for high-quality data, while low-quality data for text have a maximum range between 20230 and 2050 and for image data up to 2060. Therefore, they suggest that it should “drastically improve the efficiency of the data or have new sources of data” to compensate for the high demand that is projected there will be, as this technology finds more use cases.
It’s important to note that platforms that use machine learning to produce content, like the ones mentioned above, also rely on the reach of the internal algorithm that progresses based on continuous usage by people.
It may interest you: Apple will sell audiobooks narrated by artificial intelligence
What is ChatGPT
One such platform involved in this situation is ChatGPTwhich is an artificial intelligence created by Open AI to generate coherent responses from user questions in text.
The platform has been used by people as a search engine, but also to create more elaborate documents such as academic essays or work emails.
This growth has led to Microsoft plan to include this tool within Bingits web browser, and by March of this year it will begin tests to integrate it and compete with Google.