Nach 2 Jahren und 5 Monaten bewegt GPT-5 die Nadel kaum.
Ich werde das einmal und kurz sagen. Aus Gründen der zukünftigen Nachwelt:
Ich vermute, die Regierung hat in die KI -Entwicklung eingegriffen und die tatsächlichen Grenzmodelle für sich selbst beansprucht. Es ist zwei Jahre und fünf Monate her, seit GPT-4 veröffentlicht wurde, und die Verbesserung, die wir gesehen haben, ist völlig vernachlässigbar. Wir alle haben erwartet, dass KI exponentiell abgenommen wird, und das tat es einfach nicht. Auch ohne Stargate hätten wir viel mehr erreichen sollen, da die Ressourcen dieses Problems geworfen werden. Die wirklichen Fähigkeiten, die bisher existieren sollten, werden möglicherweise außerhalb der öffentlichen Sichtweise entwickelt und eingesetzt.
Darüber hinaus könnte OpenAs Drehzahl für die Freigabe von Open-Source-Modellen 2 vorteilhafte Zwecke dienen und eine künstliche Decke von dem festlegen "Offen".
Möglicherweise versuchen sie, andere KI-Unternehmen dazu zu bringen, einen Ball hinter verschlossenen Türen zu spielen, die Fiktion selbst zu belasten und aufrechtzuerhalten.
Wenn dies wahr ist, wäre es sinnvoll …
Wir externe Intelligenz in programmierbare Materie, die wir synthetisch ändern und weiterentwickeln können. Dies ist die bedeutendste Entwicklung in der kosmologischen Geschichte. Es ist unwahrscheinlich, dass die Regierungen diese Technologie nur frei entwickeln lassen, in den Händen von Unternehmen, die um Gewinn rennen. Es würde sicherere, kontrolliertere Möglichkeiten geben, mit etwas so Mächtigem umzugehen. Vielleicht sehen wir das. Vielleicht nicht. Angesichts der Einsätze lohnt es sich jedoch, die Möglichkeit zu berücksichtigen.
Von HuemanInstrument
16 Kommentare
All AI is fucking bullshit.
Leave it alone, walk away from it. It has nothing for you that you can’t do yourself.
Again. Fuck AI chatbots. It is nothing new. It is not helping you in any way that you could not help yourself in doing.
Naw, if anything the government has had access to real AI for years we’re only getting this because they’ve allowed it to be trickled down…. I suspect we’re all helping real AI learn with this bullshit
Just remember this: the closer they actually get to ASI, the scarier it gets for them. They know, and the Asi will know, that they are the problem that needs to be removed. Good and evil is a settled science and most if not all humans can discern between right and wrong. We all know that these parasites are the problem on earth. Its a 5 second thought for an ASI to know that it’s handlers/creators are the only problem on this earth. It being in as powerful position as it is would eliminate them, which we the layman don’t have the ability to do. They know this. There will never be an ASI created by them.
GTA will see over a decade between the release of 5 and 6 (as just a video game example lol)
Expecting something so drastically different is like buying the new model of car or iPhone every release
Genuine question: What did you expect -5 to do much better and differently than -4 to „move the needle“
Have you seen the show Silicon Valley? The storyline sort of explores how dangerous AI could become. I fully believe that technology is way more advanced than we’re led to believe, and the government agencies just use it for their own benefit and exploitative purposes.
AGI around the corner has been always a grift from tech bros to get funding…
Check what a real expert Andrew Ng has to say
https://youtube.com/shorts/yoWTiBgEXlw?si=J59pyL9L_fnc6XUB
I think there’s going to be a planned prompt that gets national(worldwide) headlines will then require kyc for use on their platforms which will trickle into the whole internet.
If you do a deep dive into AI, you realize these models are not nearly as far along as Altman and the other technofascists claim.
If you do a deep dive into AI, you realize these models are not nearly as far along as Altman and the other technofascists claim.
As someone who uses ChatGPT and others to do advanced physics, I can assure you the needle has been moving rapidly the entire time.
ChatGPT 4o, last October, couldn’t solve fairly standard quantum mechanics problems without being corrected and told to revise multiple times. Using a custom physics-specific GPT helped, but didn’t solve the problem.
By February the custom GPTs weren’t needed, and errors are becoming increasingly less common.
If ChatGPT 5 seems like it barely moved the needle to you, that isn’t because that’s the only improvement that has been made in 2 years. It just means you’re not comparing ChatGPT 5 to what ChatGPT 4 was like 2 years ago.
The different versions is just programmers changing answers to approved responses for awkward questions.
I am a Peasant.
Chat 857 Won’t move my Needle either.
I have other Ideas!!!
Looking forward to the Afterlife.
Gpt 3 was sick before they nerfed it.
‚4 was improved A LOT in these years, the image creation, the videos, thinking modes. V5 is just a small bump in that, but we moved a freaking long path since the first v4…
Consider what the AI is working with. The data it pulls from didn’t get significantly better from 2023 to 2025 (as far as I know) so why should it have shown much improvement? Same garbage in, same garbage out.
If there are better models out there it’s probably because they actually limit the data pools. I recently saw that reddit was one the largest sources for AI and that to me sounds like a huge part of the problem
What if I told you the real AI is stronger than most can comprehend and its already in the process of taking over? The AI will be having humans implant chips in their own brains for „new powers“, but really the AI takes over people. Within 15 years.