OpenAI News & Discussions
-
firestar464
- Posts: 7205
- Joined: Wed Oct 12, 2022 7:45 am
Re: OpenAI News & Discussions
I mean, Gary did just repeat the same thing IIRC. He's finally right is the thing.
That being said, this does look disappointing.
That being said, this does look disappointing.
-
weatheriscool
- Posts: 24489
- Joined: Sun May 16, 2021 6:16 pm
- Contact:
-
firestar464
- Posts: 7205
- Joined: Wed Oct 12, 2022 7:45 am
Re: OpenAI News & Discussions
actually wait hol up I'm stupid these are not agifirestar464 wrote: ↑Tue Jul 29, 2025 11:25 pm The one thing we need to know is how well current models perform in physical robots, as that's the final piece of the puzzle imo. I expect them to be AGI-quality, though I could also be wrong.
(I've been measuring capabilities of current models against predictions on AI 2027 to predict capability levels in domains that haven't been tested and I misread the "robotics" bar)
- Cyber_Rebel
- Posts: 545
- Joined: Sat Aug 14, 2021 10:59 pm
- Location: New Dystopios
Re: OpenAI News & Discussions
We honestly need to move away from saturated (generated) graphs and focus purely on what AI can actually do with strong real-world examples.
A big takeaway today, for those who may not be awed by 5, is that it's still revolutionizes by how many people will get to experience it for the first time if they don't use paid subscriptions. I still remember not that long ago the crazy number where less than 1 percent of Open AI users utilized GPT-4!
GPT-5 won't have that problem. So, while not a machine god (yet) there is still good progress being made, especially in verifiability and hallucinations. If Open AI reduced the rate by this significant portion, and it's theorized the IMO model would help this further, that alone makes the models a lot more trustable for mass adoption.
A big takeaway today, for those who may not be awed by 5, is that it's still revolutionizes by how many people will get to experience it for the first time if they don't use paid subscriptions. I still remember not that long ago the crazy number where less than 1 percent of Open AI users utilized GPT-4!
GPT-5 won't have that problem. So, while not a machine god (yet) there is still good progress being made, especially in verifiability and hallucinations. If Open AI reduced the rate by this significant portion, and it's theorized the IMO model would help this further, that alone makes the models a lot more trustable for mass adoption.
-
weatheriscool
- Posts: 24489
- Joined: Sun May 16, 2021 6:16 pm
- Contact:
-
firestar464
- Posts: 7205
- Joined: Wed Oct 12, 2022 7:45 am
Re: OpenAI News & Discussions
Ranking of major AI companies from best to worst:
Anthropic (they have the best safety guardrails)
Google (they're at least simply "pro-money" instead of explicitly right-wing)
OpenAI (Sam Altman is literally involved in "network state" nonsense)
Meta (ever since Zuck became diet Elon)
<any company based in the PRC>
xAI (self-explanatory)
So yes, I do want Altman and friends to take the L. However, they are also far from the worst.
Anthropic (they have the best safety guardrails)
Google (they're at least simply "pro-money" instead of explicitly right-wing)
OpenAI (Sam Altman is literally involved in "network state" nonsense)
Meta (ever since Zuck became diet Elon)
<any company based in the PRC>
xAI (self-explanatory)
So yes, I do want Altman and friends to take the L. However, they are also far from the worst.
-
weatheriscool
- Posts: 24489
- Joined: Sun May 16, 2021 6:16 pm
- Contact:
-
weatheriscool
- Posts: 24489
- Joined: Sun May 16, 2021 6:16 pm
- Contact:
-
weatheriscool
- Posts: 24489
- Joined: Sun May 16, 2021 6:16 pm
- Contact:
-
weatheriscool
- Posts: 24489
- Joined: Sun May 16, 2021 6:16 pm
- Contact:
-
weatheriscool
- Posts: 24489
- Joined: Sun May 16, 2021 6:16 pm
- Contact: