00:00An investigator from Antropic, the company that develops Cloud, one of the main
00:04JADGPT's competitor resigned a few days ago, and his reasons are generating a lot of discussion.
00:09the world of artificial intelligence. His name is Jacob Coxon and his departure has a particularity.
00:14Before joining Antropic, he worked for about three years at OpenAI, where he participated in the
00:19teams that developed models like GPT-4 and GPT-4.5. Now, in a publication on the web
00:25Social media personality X, who has gone viral, says she is leaving because she believes that both OpenAI and Antropic
00:31They are moving too quickly towards increasingly capable artificial intelligence systems.
00:36In a series of posts on his profile, Coxon asserted that both companies are in a race.
00:41towards so-called superintelligence, systems that could eventually far surpass some
00:46human capabilities and, in the scenario that concerns him most, to contribute to improving his own
00:52capabilities. Former Coxon and Antropic boss Evan Hovinger publicly came out in support of him.
00:57Hovinger is responsible for alignment research, one of the areas that study how
01:02to get AI systems to behave in accordance with human goals. Hovinger said
01:07who personally believes there is a greater than 10% probability that intelligence
01:13Artificial intelligence could cause the extinction of humanity within the next decade. This is important.
01:17Put that figure into context. It's Hovinger's personal estimate, not a scientific prediction.
01:22Even Hovinger makes an important distinction. He says he considers the risk of the models to be low.
01:27current systems and that their concern is focused on future systems that are much more capable, especially
01:32Those who can participate in self-improvement processes. The idea is that a system of
01:37artificial intelligence can help develop other, more advanced systems, and that process
01:42It could happen again. This is one of the scenarios that some researchers consider when discussing a possible
01:46superintelligence. But that doesn't mean current models are doing this, nor that a superintelligence exists.
01:51There is no scientific consensus on what will happen. What does exist, however, is a debate within the field itself.
01:56artificial intelligence industry on how quickly these systems should advance and how
02:01This development must be accompanied by security and control mechanisms. Coxon believes there is a problem here.
02:07Additionally, there is competition between companies. Their argument is that Antropic can recognize the
02:12risks, but feeling that it has to keep moving forward because other labs are doing it too.
02:16And something similar, though for different reasons, would happen at OpenAI. Therefore, it suggests that the
02:22Large laboratories should exploit coordination mechanisms to control the pace of development
02:27and even consider, in certain circumstances, a temporary pause in some capacity improvements.
02:33What are your thoughts on this? Let us know in the comments.
02:46Thank you.
02:47Thank you.
02:47Thank you.
02:48Thank you.
02:48Thank you.
02:48Thank you.
02:49Thank you.
02:49Thank you.
02:49Thank you.
02:49Thank you.
02:50Thank you.
02:50Thank you.
02:50Thank you.
02:50Thank you.
02:50Thank you.
02:50Thank you.
02:50Thank you.
