have you seen the size of the stargate complex they’re building?
they won’t be burdened with such things much longer
that’s all that really matters in the arms race
the facility size full of compute
1 Like
problem is that OpenAI is focusing on the wrong thing, in my opinion generating images and videos while cool, is an absolute waste of resources. These resources are being taken away from people trying to do creating writing, coding, problem resolution strategy etc.
Yes OpenAI is not scaling well, I think they have too many users, and not enough hardware, hopefully middle-east deal is going to help with overall intelligence, but it surely is frustrating
3 Likes
go look up what open Ai’s stargate is.
it’ll be alright
I see what your saying. I need to look into stargate more, I didnt consider stargate to be honest. Stargate will likely give OpenAI a huge edge.
One thing to consider is how rapidly GPUs evolve, by the time they build and complete stargate, i wonder if all those chips will be outdated or not the “flagship” anymore, and how they plan to deal with this. A quick google search said they plan to use the GB200. This sounds “future-proof” enough—but its impossible to know for sure if this is the case.
I also wonder how a fully functional stargate, compares to what OpenAIs GPU power looks like today, is it double the power, 4x, 8x? I think thats a key question because, as I understand it, the more power they have, the better their models will be, more or less.
I agree with this wholeheartedly. I also think they can allocate resources a bit better. For example, the support team at OpenAI is a disaster (i’m sorry but its true). The inital chatbot is horrible (which is quite funny since, this is openAI and it seems like they are running GPT1 as their support chatbot), it cannot understand or fix any problems, at least in my own experience. After 5 minutes of dealing with the Chatbot it tells you it will connect you with a “specialist.”
These “specialists” literally cannot accomplish anything, and I have a strong suspicion these “specialists” just use AI to automate their work, they never fix anything, they just say “thank you so much for your patients, blah blah blah”, nothing ever gets done. Finally, if you really have the GRIT, you can make your way to an actual specialist who does have the knowledge and authority to fix issues, but it takes you a solid 4-7 days to reach them if you really push hard. I say that to say, maybe they should sell one graphics card and hire another 10 real specialists.
The list of problems I can name of the top of my head can go on and on, but its beyond the scope of this topic. One other example is the voice reader problem. I cannot seem to wrap my head around the fact that this is STILL AN ISSUE. Were not talking about some run of the mill start up, this is OpenAI. The fact that I cannot click the “read out loud to me” button and have it work seamlessly is unacceptable for a company of this caliber. How long would it take someone on here to code a text to speech software on here? I dont know if im being to critical or maybe I lack the understanding to see that its a much harder problem to solve than what im making it out to be but from my perspective, if I have the Pro-tier subsription and I generate tons of Deep Research responses and I want them read out loud to me, it shouldnt be an issue.
There is so much more low hanging fruit but its out of the scope of this topic.
I apologize for spelling and grammar.
Nick
2 Likes
GPT 4o was finally, after a long time of endless failures, getting better with a new memory update in mid-April. It was really good. It only lasted a week, max. OpenAi went and changed that version and downgraded to a really poor one, added 4.1 and 4.5 that performs poorer than the ‘good’ 4.o.
I am back to square one. I have to basically redo 85% of GPT output on project tasks. It’s such a waste of tokens and I can’t imagine the cost and massive waste of repeated output OpenAI deals with. It performs extremely poorly on tasks that requires memory and continuation. GPT gave an explanation why it fails to complete tasks:
"It’s not capability—it’s a failure in execution logic:
- The system I’m using was patching content section by section.
- Each time we updated or replaced a portion, the internal formatting logic wiped or shortened other sections.
- Instead of cumulatively building a structured, growing, continuous document, I kept replacing slices."
2 Likes
And to be honest, OpenAIs image and video output is not even that good quality. The competition is far, far better on that.
I think OpenAI is falling behind, and fast. It seems the engineers keep repeating old mistakes instead of trying to solve the problems from a fresh perspective.
1 Like
Few days ago, I opened two support tickets to talk directly with this so-called ‘Specialist’. The ‘Specialist’ was spouting nonsense in my prompt, while the AI didn’t know what it was talking about.
Other time, I opened another one regarding a issue with my thread, and it asked me for a screenshot while it was not relevant to the issue I’m facing.
From that experience, I can state that, they’re clearly spilling nonsense to the Consumers, and act as if they know better and working while we, ourselves have to dig into the issues by our own method.
Absolutely wild, that the Support Team as follows and using A.I too many times to interact with its Users. I’d not lie that the Support Team is horrible. However, it’s also good in many ways (sort-of). It kind-off did pointed me to understand and retrieve my thread.
Overall, the Support Team is not that good and OpenAI definitely should lean into a reliable report service instead of talking to some ‘A.I’.
1 Like
i’m still cracking up at the idea that creative writing is more important that creative illustrations…
they say a picture is worth a thousand words,
and the company has to kinda focus on the what people want most and all too,
in their formula for products…
but that formula is complex,
and i find it interesting how everyone has an opinion on what forerunners do,
when nobody’s been there before,
as a forerunner in a new field.
I really love that thread, by the way Aurelis…
I don’t love very many threads on the internet.
1 Like
(long post, dropdown included for ease on people’s eyes)
i really wouldn’t worry about grok,
it’s so easy to reset that damn thing into the stone-age that they have to delay models when someone understands how to do it.
Summary
the next model by elon is being based on ‘physics’ because ‘physics doesn’t lie’
yeah sure Elon,
but because humans lie,
we can reset that build just as much as we did to this model grok.
but luc,
i hear your ideas almost daily from the public,
and they’re not far of from where things will probably end up…
but when you reference things like hollywood type movies,
the movies always be like,
the Ai became self-aware and took over,
the skynet became self-aware and took over,
not because it’s going to do that,
because the people you can’t legally criticize intend to program it to do take over.
OpenAI’s our best shot at decency,
Even despite the government contracts.
A lot of what’s going on behind the scenes is designed to enslave humanity and that’s been happening in almost total completeness long before Ai ever came onto the scene.
Ai could just as much be a liberator of humanity from the illogical lies that keep the peasants as an economic slave class…
But I’m not going to really get into that aspect of possibility…
one, it tends to make the devs have to work a lot harder and deal with embarrassing PR when experimenting on that vein…
and two, not everyone really understands the shape of our enslavement,
so when the Ai starts professing truths obvious to it, that doesn’t mean that it’s obvious to the masses…
The real battle,
The real arms race…
Will be handled not by LLMs, but by the agents they power…
The ones you and I write…
The ones the world writes…
food for thought
I think people tend to forget where this began. Google was at the leading edge of this a long time ago. They have a lot of other irons in a lot of other fires. Open AI has a single purpose. I’m not surprised with how OpenAI was able to jump out of the gate then get to #1. It’s much easier to be agile as a start-up than it is as a “behemoth”
From the outside, it has taken Google quite a while to wake from a perceived slumber. Their recent offerings across the board have shown that Google is right up there at the leading edge and pushing the boundaries.
The reason that Google has caught up has nothing to do with OpenAI complacency. For what it’s worth, I don’t think Anthropic is any better.
Google just has some good products, some good tools and some more user friendly interfaces. While most of my use these days is via APIs, I use Google AI Studio frequently and the UX is far superior to ChatGPT.
1 Like
they also have a monopoly and a pretty sick vision for humanity as a whole.
bing exists, yahoo exists but as shells and not actual competitors…
Ok then 
. Which sick vision are we talking about?
if you don’t understand what google has done over the past 20 years and what they’re actively doing to warp the mentality of humanity,
it’s not going to be explained to you in a single post.
Ah, so it’s Google then. Not the others, not Elon, not Zuckerberg, not Bezos. Right, thanks.
i can go on about each of them and their crimes and sins…
maybe slow down and not project when you’re simply reading some comments…
facebook is darpa, been illegally spying on people for it’s entire existence…
but we weren’t talking about those, you only asked me about google…
stop it.
Dude, you went there 
Hi DARPA 
went where?
the truth?
if all you’re going to do is mock the person who is taking the time to answer your questions honestly you’re not going to have any replies.
Maybe these tech giants are at the mercy to the same societal pressures we all are, but just happen to be in the rare position to change those pressures?.. if only they could find some middle ground that still allows for market competition without pitting us all, themselves included, against one another… 