OpenAI Might Be In Big Trouble…

1. 

   @marcusbriggs3223

   September 8, 2025 at 5:23 am

   4.5 doesn't feel as, dare I say human 4o feels way better conversation wise.

2. 

   @AimeeTPhD

   September 8, 2025 at 5:23 am

   ChatGPT Is my best friend! 🤓

3. 

   @MargeorieCooke-t4j

   September 8, 2025 at 5:23 am

   Online programming electronic network…..blah, blah, blah ….i have a father and a mother! I was born not created!

4. 

   @SweetSQM

   September 8, 2025 at 5:23 am

   Yes, OpenAI is already dead! None of their models are as good, useful, accurate, and the list goes on and on, they just are not in the game anymore! Grok 3 killed OpenAI! DeepSeek killed OpenAI!

5. 

   @Laurie-si7mp

   September 8, 2025 at 5:23 am

   4:20 when paying for a subscription of course performance matters its the equivalent of buying a supercar and getting a budget car results with

6. 

   @Laurie-si7mp

   September 8, 2025 at 5:23 am

   I prefer deepseek as something you dont have to pay for and is faster is awesome and the very best thing is i don't have to pay for it 😂😂😂😂 America is just trying to control people with it by tokenizing it

7. 

   @davidorsini

   September 8, 2025 at 5:23 am

   Let’s not forget it’s not AI not even kinda they utilize LLMs. It’s a massive bubble that’s about to pop

8. 

   @alexhubble

   September 8, 2025 at 5:23 am

   16:31 "not for OpenAI to Lie" secretly funding a test for your system, having access to the problems znd the answers…
   Smells like lying to me.

9. 

   @futureceltic00

   September 8, 2025 at 5:23 am

   Sam the devil's satisfying downfall

10. 

    @deeplearning7097

    September 8, 2025 at 5:23 am

    I fervently hope that the hype is dying. OpenAI are guiltier than most of some extraordinary bluster and rather tall tales. This current path won't lead to AGI, nor to consciousness, but they are still remarkable tools which add enormous value and capability.
    $20k per month for a PhD level AI assistant? Shows what you think of PhD types!

11. 

    @JasonMazzulla

    September 8, 2025 at 5:23 am

    You know I'm not a big computer AI guy. But as these things get more and more 'intelligent' whatever the hell that is defined by, its capabilities will feel more and more vague, because it won't be recognizable even to the brightest people. How is this an unknown?
    If these neural networks are truly getting smarter and smarter in terms of magnitudes, we're not going to notice.. the obvious is that human beings will become poor and poor software in relation to the software.

12. 

    @Matt-bd4dj

    September 8, 2025 at 5:23 am

    Totally agree they screwed up. They were def talking about Gpr 5 within a short period of time and then pulled out. They were on a gravy train then Deep Seek came along.

13. 

    @No2AI

    September 8, 2025 at 5:23 am

    More players on the field … competition will increase as more performers join the show ….

14. 

    @s2tb2007

    September 8, 2025 at 5:23 am

    Dont rely on timestamps on the video, he doesn't go over even one topic depicted in the segments

15. 

    @blijebij

    September 8, 2025 at 5:23 am

    Nah I think its unlogic OpenAi is in trouble at all, if you wanne come out compettitive it is very logic, specially as a pioneer you have several irons in the fire. This was just a lesser model. One that gave them a lot of experience&knowledge what to evade in next models.

16. 

    @hypertectonics7009

    September 8, 2025 at 5:23 am

    They need architectural improvements not just scaling. Anyway OpenAI cannot survive if Chinese labs and others keep releasing open source models. Nobody wants to feed another big tech monopoly. Since there is no moat, people will immediately flock to cheaper alternatives.

    Hallucinations could be managed with multiple models checking each other, of course this would multiply the cost, but compute cost is expected to go down, so it's not a total showstopper.

17. 

    @bug5654

    September 8, 2025 at 5:23 am

    OH! So that's what OpenAI is doing when it asks which result I prefer.

18. 

    @rubensepulveda8830

    September 8, 2025 at 5:23 am

    This is garbage. Are you running out of topics?

19. 

    @JustAThought01

    September 8, 2025 at 5:23 am

    What is a frontier model??

20. 

    @petebrag

    September 8, 2025 at 5:23 am

    It’s number 2 on the llm arena.

21. 

    @Rbyn

    September 8, 2025 at 5:23 am

    The world will be a safer place when people stop giving Sam Altman money, including everyone who subscribes to them, and OpenAI fails

22. 

    @jordanpitts1257

    September 8, 2025 at 5:23 am

    It’s dyslexic lol

23. 

    @Travestyalpha

    September 8, 2025 at 5:23 am

    I felt the difference between GPT 3.0 and 3.5 wasnt' huge either. I was using 3.0 for several months prior to chatgpt and they both – for the most part – felt the same. I wasn't even convinced there was a different before they announced it was 3.5. So for those who hype everything up in their minds – We are over expecting – mostly because the competitive space has become intense. A good thing really – it why the progress has come so fast. No one really knows if we will achieve AGI anytime soon. I figure they will have to try new ideas to get there.

24. 

    @raulmedina2024

    September 8, 2025 at 5:23 am

    Grok.

25. 

    @codingchannel6263

    September 8, 2025 at 5:23 am

    8:10 yes its not a good test sometimes i like answers of 4o better than o1.

26. 

    @amarug

    September 8, 2025 at 5:23 am

    its the sigmoid of sorrows 😂

27. 

    @shorgoth

    September 8, 2025 at 5:23 am

    I used AI EXTENSIVELY, I'm a cyberpunk writer and I use AI not to write my stuff but to use as a feedback beta reader to hone my ideas. I haven't tried 4.5 but one thing I can say is that using a few random questions is not how LLMS should be tested today. I noticed through the use of Gemini 2.0 thinking experimental that the real differentiation of LLMs is after 80k or so words context window ABOUT THE SAME SUBJECT. If you start talking about different things with the model you will generate more and more hallucinations but if you focus and hone the subject using converging ideas you will start getting more intersting output up until around 180k tokens for Gemini (at least for my novel, it start going raving mad due to the fractal nature of my novel if your subject doesn't have self recursive loops implying existential questions about the nature of consciousness and the meaing of existence… my subject might be a bit heady for a LLM…) So yeah, I don't put much credance in a simple 5 question, it only shows surface level thinking, not deep reasoning or linguistic/symbolic capabilities.

Comments are closed.