Artificial Intelligence
Artificial Intelligence
Out of context: Reply #1637
- Started
- Last post
- 2,578 Responses
- yuekit0
Some people think AGI has been achieved by the new o3 model and that yesterday was a historic day in human history (or at least AI development).
Of course this is complicated by the fact that there is no actual definition of "AGI." But it gets some impressive results in the testing they did.
I was curious about what frontier math is...you can see some examples here
- I take it this isn't public yet? and might be prohibitively expensive to most once it is?_niko
- Yes that's the fine print...in order to get these amazing results they had to use a ridiculous amount of energy. You can see the cost per task in first image.yuekit
- Apparently they trained on 75% of the training set :))
No AGI in sight...or AI!******** - lol...I think training set is just like prepping the AI to take the test.yuekit
- But there's no doubt OpenAI is under a huge amount of pressure to deliver big breakthroughs so yeah it needs to tested independently.yuekit
- This article gives you the backstory...they tried and failed to build GPT-5. Now trying this new approach of reasoning models.
https://archive.is/9…yuekit - Guessing 4o was actually 5, now they are trying to find new ways to scale without more data.
Still no AI, still no AGI. LLMs will never get there.******** - Try to solve the puzzles yourselve
https://arcprize.org… they have 6 as a sample, it's mostly matching visual patterns********