A OpenAI afirmou nesta sexta-feira (30) que vai limitar as informações relacionadas a armas que o ChatGPT pode fornecer a adolescentes, mesmo em perguntas que parecem inofensivas. As medidas visam prevenir a violência no mundo real.
A medida surge num momento em que as empresas de IA enfrentam escrutínio crescente em relação aos riscos que os chatbots representam para jovens, e ao papel destas empresas na deteção de ameaças.
As restrições se aplicam mesmo em conversas em que não houver indícios de violência. Na prática, o ChatGPT poderá encerrar conversas automaticamente, e a OpenAI poderá pausar temporariamente as contas após várias conversas deste tipo.
A desenvolvedora também afirmou ter reforçado seu sistema de previsão de idade para direcionar usuários que, segundo estimativas, têm menos de 18 anos para a experiência voltada para adolescentes. Além disso, planeja notificar os pais caso a conta de seus filhos adolescentes seja suspensa por discussões sobre violência.
A empresa também está limitando, para todos os usuários, conversas envolvendo ataques em massa, incluindo o fornecimento de informações sobre as armas utilizadas e como foram obtidas.
A OpenAI afirmou que os adolescentes passam, em média, menos de 15 minutos por dia no ChatGPT, em seu primeiro relatório sobre o uso da plataforma por adolescentes.
Um relatório divulgado no início deste mês pela organização sem fins lucrativos Common Sense Media constatou que a versão para adolescentes do ChatGPT frequentemente alertava os pais tardiamente ou não os alertava quando contas de adolescentes mantinham conversas sobre suicídio, automutilação ou transtornos alimentares.
Una empresa de Singapur de centros de datos, Megaspeed, compra chips a Nvidia, la empresa más valiosa del mundo gracias a sus sofisticados dispositivos que impulsan el boom actual de IA. El Gobierno de EE UU sospechaba desde 2024 que Megaspeed revendía esos chips a Alibaba y otras empresas chinas de manera irregular. Porque desde 2022 Washington restringe la exportación de los modelos más potentes a China.
OpenAI CEO Sam Altman said the AI giant is sorting through how to continue research while making sure those in specialized fields "get the best version of the future that they can have."
Minh Connors/Bloomberg via Getty Images
OpenAI CEO Sam Altman is conflicted about how AI models are changing fields like mathematics.
Altman said that he had some "sadness and melancholy" when he saw what OpenAI's models could do.
Prominent mathematicians have blasted AI companies for pursuing some of the field's most complex problems.
OpenAI CEO Sam Altman said he feels conflicted about OpenAI solving some of math's biggest problems.
"As I said, I don't think it was bad for AI," Altman told Stripe President John Collison during an episode of "Cheeky Pint" that was released on Thursday. 'And I think, on net, the world deserves to understand how the universe works. But I do have empathy for the method."
A number of prominent mathematicians have blasted AI companies for pushing frontier models to solve some of the most complex math problems. OpenAI pulled out of a Caltech "math hackathon" after current and former Caltech mathematicians and others in the field said the event was "likely to have destructive impacts for the mathematical community."
Separately, 28 Fields Medal winners published their own open letter, warning that "the push by AI companies to solve mathematical problems as a benchmark is detrimental to the science of mathematics, and to the mathematical community."
OpenAI received considerable attention after announcing that its models had solved a 90-year-old, unresolved problem related to the Navier-Stokes equations, a set of formulas that predict how liquids and gases move. The Clay Mathematics Institute had offered $1 million to anyone who could determine whether those formulas would always behave as we would expect in the physical universe, part of a series of difficult mathematical questions known as the "Millennium Problems." OpenAI said that its AI models found an example where the equations do not, and end up with a "singularity" where a vortex starts spinning infinitely quickly.
Altman said OpenAI is "working through" how to continue scientific research while also making sure those in highly specialized fields "get the best version of the future that they can have."
"I would like to figure out a way to do this that is as sensitive to the communities that have driven these fields forward for centuries as we can," he said, adding, "It's a hard thing."
Collison tried to assuage Altman's feelings, reminding the OpenAI CEO that "the longshoremen union weren't happy about the port automation."
Altman said part of the reason he feels differently about the whole experience is based on an OpenAI Slack channel that was named "Model Name-Hold My Beer" after the meme about if you think this is crazy just wait. (He did not explain if this was a released model or not.)
"People were just throwing that at the hardest problems they could," Altman said. "And people didn't think the model was going to do these things."
The AI model tackled "people's hardest research problems from their academic career." Altman said he was so excited, "I couldn't go to bed."
In addition to the euphoria, he said other feelings set in.
"I also had much more sadness and melancholy, and a feeling of something ending than I was expecting," Altman said.
The OpenAI CEO said it was "beautiful" that any mathematician could solve even the most formidable problem if only they worked hard enough. Altman said AI is changing that equation.
"I mean, certainly, most mathematicians never solve a Millennium Prize problem," he said. "But the idea that you might be able to."
It’s almost hard to remember, but just a few years ago, a lot of very smart people thought that science was in terminal decline. Respected researchers were writing papers with titles like “Are Ideas Getting Harder to Find?”, “Papers and patents are becoming less disruptive over time”, and “Slowed canonical progress in large fields of […]
Login
