Thread 106945073

11 posts 4 images /g/

Anonymous 10/20/2025, 4:50:48 AM No.106945073 [Report] >>106945131 >>106945164 >>106945300 >>106945341 >>106946981

I have it on good authority that OpenAIs latest unreleased experimental LLM just hit AGI, and is capable of even reasoning at a PhD level in some domains.

Anonymous 10/20/2025, 4:51:53 AM No.106945084 [Report]

>corporation says what agi is
what IS agi? you think you know but you don't

Anonymous 10/20/2025, 4:56:57 AM No.106945131 [Report] >>106945141

>>106945073 (OP)
can it learn on its own? can it actually make something new and interesting?

Anonymous 10/20/2025, 4:58:28 AM No.106945141 [Report] >>106945152

>>106945131
Can you?

Anonymous 10/20/2025, 5:00:01 AM No.106945152 [Report]

>>106945141
yes

Anonymous 10/20/2025, 5:01:15 AM No.106945154 [Report]

It isn't capable of continual learning beyond its context window. But it can learn several books worth of data.

Anonymous 10/20/2025, 5:02:31 AM No.106945164 [Report] >>106945209

>>106945073 (OP)
>even reasoning at a PhD level in some domains.
free-tier chatgpt can do that right now (in the humanities)
specify which domains

Anonymous 10/20/2025, 5:08:28 AM No.106945209 [Report]

>>106945164
Some STEM domains.

Breakthroughs are in physics, chemistry, and engineering. Basically a wide swathe of MSE

Anonymous 10/20/2025, 5:24:06 AM No.106945300 [Report]

1746992285990719.png md5: 83d8f50b...

>>106945073 (OP)
> openai allowing pornchats
> shilling on 4chan
> "just 2 weeks until AGI"
really inspiring confidence here

Anonymous 10/20/2025, 5:28:48 AM No.106945341 [Report]

>>106945073 (OP)
They can barely make an LLM that just regurgitates whatever you train it on, incapable of originality or thinking for itself, and you want people to think they might be close to AGI?
>>>/x/

Anonymous 10/20/2025, 10:13:13 AM No.106946981 [Report]

>>106945073 (OP)
> t. trust me bro, the source is my uncle who works at nintendo