A study presented as entirely written by Grok 3, Elon Musk’s artificial intelligence chatbot (IA) is brandished by climatosceptics on social networks, but independent researchers warn against the credibility of such a method.
Entitled “Critical reassessment of the planetary warming hypothesis linked to CO emissions2“, The article published in” Science of Climate Change “rejects in particular the conclusions and projections of the reports of the UN climate experts (IPCC), based in particular on studies widely disputed for years in the scientific field.
An anti -tin applaud
Many climatosceptic accounts shared the study after its publication at the end of March, including the American biochemist Robert Malone, at the origin of multiple false information about vaccination during the Pandemic of COVVI-19.
“The use of AI for research funded by the State will normalize and standards will be developed for its use in scientific journals,” said Robert Malone. The study rings the end of “climate scam”, he even proclaimed on X, collecting more than a million views.
False neutrality
However, there is a scientific consensus making a link between the consumption of fossil fuels and warming, as well as the growing intensity of meteorological phenomena, such as heat waves and floods. Specialists thus warn against a false sense of neutrality under the guise of an artificial intelligence presented as “author” of a scientific article.
These “great language systems do not have the capacity to reason. These are statistical models that predict words or sentences on the basis of what they were trained. It is not research, ”recalls Mark Neff, professor of environmental sciences.
Doubtful co -authors
The article maintains that Grok 3 “wrote the entire manuscript”, with the help of co -authors who “played a crucial role in the orientation of its development”. Among these co -authors is the astrophysicist Willie Soon, a notorious climatosceptic that has received more than a million $ in funds from the fossil fuels through his career.
Certain studies referenced by Grok 3, however questioned by scientists when they were published, were added in the analysis at the request of the co -authors, according to the article. “We do not know how the authors asked the AI to analyze” the data and sources presented in the article, underlines Elisabeth Bik, Dutch microbiologist installed in California and specialized in scientific integrity.
For Ashwinee Panda, an expert in artificial intelligence, it is impossible to verify if the AI has carried out an analysis without external interference: “Anyone can claim that AI wrote this, alone, and that it is not biased,” he says.
An unrecognized newspaper
Neither the newspaper nor its publisher seem to be part of a scientific ethics committee. The article was also submitted and approved for publication in just 12 days, a very short period of time, note experts. “That an AI can plague bogus articles” is by no means a surprise for Gavin Schmidt, NASA climatologist. And this analysis presented as new “has as little credibility” as the references it uses, he says.
“The use of AI is only the last stratagem to give a false impression of renewal in the climatosceptic argument”, abounds Naomi Oreskes, historian of science at the University of Harvard.
AFP contacted the authors of the article about the research process and writing the study, but has not received an immediate response.
