What does artificial intelligence have to do with steel and nuclear tests of the atomic age in the Second World War?Nothing in concrete terms, but there are interesting moments that John Graham-Cumming, former Chief Technology Officer of CloudFlare, have brought to launch Lowbackground Steel.aia website that aims to catalog Save the content created by people before the massive arrival of generative artificial intelligence.
OK, but steel and nuclear energy? The conceptual connection is simple: in the same way in the era of Manhattan ProjectThe Trinity test and the subsequent bombing on Hiroshima and Nagasaki, Nuclear explosions of polluted world steel production(Due to a larger amount of radionuclides in the air), the presence of content generated by the AI increases to contaminated and cannot be distinguished from the works produced by humans authentic.
After 1945, the scientific community, in the light of the need to create sensitive and reliable tools, worked for Recovery “pure” steel, produced before 1940, from maritime wrecks prior to the war conflict: This material was known as the name of
“Steel with low backrest”From where Graham-Cumming has drawn inspiration to baptize the preservation service of the real creative expression of humanity.
The arrival of the generative, textual, images and video models has made it gradually more complex to check whether a content is fully produced by the human being, indeed, the infection by the AI has already led to the closing of Wordfreq, an important research project that suggested tracing the language evolution in more than 40 different languages. The need to maintain the fruit of the human intellect also comes from the risk of the possible “collapse of the models”, that is, the progressive breakdown of the quality of the models themselves when they continue in training with the help of data generated by other IAS, in that kind of “digital canibalism” that we had already had the opportunity to speak last year.
But how it works Lowbackground Steel.ai? The portal collects and reports pre-II content archives, such as I dump the Wikipedia Dellrageto 2022 (before the release of the chatgpt), the collection of books of public domain books of Project GutenbergThe Library of Congress Photographic Archive E Github Vault CodeAn “immediate” of the World Open Source Code Buried on the Svalbard Islands in 2020. The WordFREQ project is also included, “Frozen” before the IA made its methodology impossible. The site accepts reports from other sources of pre-AI content via the Tumblr page.
Graham-Comping emphasizes that the project does not want to be a position against artificial intelligence, but is planning Document human creativity before the Machines era. The site was made in March 2023, but was held on the way so far.
Just as the return to the natural levels of radiation for many applications has restored the less fundamental steel with low background steel (in some cases, especially for the construction of particularly sensitive equipment it is still used), it is possible that the evolution of the AI can follow the same fate to the pre-AI content in the coming years. However, the protection of the fruits of human creativity and intellect remains fundamentally in the perspective ofHistorical heritage to allow the authenticity of content, the evolution of language and communication in the futureFor the penetration of the IA.
#Hiroshima #Chatgpt #unexpected #link #artificial #intelligence #steel #nuclear