Gary Marcus posted that he recalls repeated past claims naming GPT-5, o3, and Jensen Huang's reference to GPT-6 as AGI, none of which proved accurate. Other users on the platform discussed definitions of AGI, with one noting models improve faster than the term's meaning. Another criticized arguments from Yann LeCun on the topic. A user described feeling gaslit by changing labels such as ASI. The posts contain no first-party confirmation of any new model release or capability threshold being met.
76.1K
6 posts, first seen 15h ago
Posts on X debate whether recent models qualify as artificial general intelligence amid shifting definitions.
Gary Marcus posted that he recalls repeated past claims naming GPT-5, o3, and Jensen Huang's reference to GPT-6 as AGI, none of which proved accurate. Other users on the platform discussed definitions of AGI, with one noting models improve faster than the term's meaning. Another criticized arguments from Yann LeCun on the topic. A user described feeling gaslit by changing labels such as ASI. The posts contain no first-party confirmation of any new model release or capability threshold being met.
Jun Song@jun_songI know this might be an unpopular take. AGI stands for Artificial General Intelligence, meaning human-level general intelligence across multiple domains. And right now, current AI is still worse than humans in countless areas. Just because a model can use Blender or handle…
Rov@ProxyJuice"no bro its not AGI its ASI, its about performance and cognitive abilities" i really feel like im getting gaslighted. when this tech came out back in 2023, they said that it can do tasks as humans but AGI is when AI can do tasks that human can do.
Gary Marcus@GaryMarcusI am old enough to remember when people told me GPT-5 was going to be AGI. I am old enough to remember when people told me o3 was AGI. And I still remember yesterday when Jensen told me GPT-6 was AGI. And when Elon Musk told me he had have a million cybercabs by the end of…
Pierre Spagnou@PSpagnouThis argument from Yann is really stupid. I do not understand why this guy can still be trusted. He has been so wrong from the beginning on LLMs.
Siddhartha Saxena@siddsaxwhat if we’re already living through the beginning of AGI and just don’t realize it because the models keep on getting better faster than the definition of “AGI” does
muizz.@0xMuizzWell, we have not achieved AGI yet, but I heard that OpenAI has a new model and it has been causing a little bit of excitement in their office. Anthropic also has Mythos which was banned, all these are pointing to the fact that we might reach AGI sooner than ever. And I think…
Jun Song@jun_songI know this might be an unpopular take. AGI stands for Artificial General Intelligence, meaning human-level general intelligence across multiple domains. And right now, current AI is still worse than humans in countless areas. Just because a model can use Blender or handle…
alvaro lordelo@alvaro_lordelo@jun_song Depends on which kind of humans you're measuring, compare to Einstein.. But compare to common person it's already much more smart.
am.will@LLMJunkyThat doesn't make it general, it makes it jagged A general intelligence can take abstract concepts and apply it to things it's never seen before A general intelligence can invent concepts that have never been created And many many other things Bots have been smarter at…
Chahat Sharma@Chahatusharma@jun_song depends where you draw the line though. a year ago computer use wasn't even on the table, now you're saying it's human-level
Lux Veritas@LuxVeritasAeter@jun_song Counter Point: Lots of dumb humans that aren't very intelligent, ergo the bar is low for AGI.
John Collins@Yinielin@jun_song Depends on what humans you choose. The average human? You really don't think AI can do multiple tasks better than the average human? Not saying we have AGI, I just dont think you have an accurate view of where human intelligence actually is.
#2
#2