[Jackson__]

They've already claimed that there will be no "GPT-5" LLM, and that instead what they want to call "GPT-5" is a fusion of their various models like 4o, dalle, their video model, etc. That in and of itself is a move that makes it quite clear to me they've hit a wall on the intelligence side.

Add these purchases, and it seems like they are extremely desperate.

[mmaunder]

Models are getting smaller, faster, cheaper to make, reflecting on their own output, adding modes and running in more places. But they’re not getting much smarter because they can only be as smart as us and each other, because that’s where their training comes from. OpenAI is strongest in a world where models cost billions to train. A world filled with cheap open source models is their worst nightmare. This is what’s happening. So they have to pivot into being a product company and away from being a model company.

[rstuart4133]

> But they’re not getting much smarter because they can only be as smart as us and each other,

That doesn't look to be true in general. AlphaGoZero didn't learn off smarter humans or smarter AI's (at all - it only trained against itself), yet it became better at playing some games than any existing AI or human.

To me it looks like the same thing has happened for LLM's in the one area they are truly good at: natural language processing. Admittedly they only learned to mimic human language by begin fed lots of human language, but they look at least as good at parsing and writing as any human now, and much, much faster at it. And admittedly they have plateaued at natural language processing. But that's not because of any inherent limitation in the level of intelligence an AI can achieve. It's because unlike playing Go there is a natural limit on good how you can get at mimicking anything, which is "indistinguishable".

The other things LLM's seem to be good at a lossy compression of all the text they have been trained on. I was floored when I ran a 16GB locally, and it could tell me things about my childhood town (pop: under 1000, miles away from anywhere). It didn't know a lot, but there isn't a lot out there about it on the internet, and it still astounds me it could compress the major points of everything it read on the internet down to 16GB. The information it regurgitated was sometimes wrong of course, but then you only expect to get a overview of a scene from a highly compressed JPEG. The details will be blurry or downright misleading.

What they are attempting to tack onto that is connecting the facts the LLM knows into a chain of thought. LLM aren't very good at that, and the improvements over the past few years look to be marginal, yet that is what is being hyped with the current models.

None of that detracts from your main point, which I think boils down to the rapid advancements in proprietary models have stalled. Their open source competitors aren't far behind, and if they have really stalled open source will catch up.

But that's only true for the natural language processing side. The shear compute required to keep a model up to date with the latest information in the internet means the model with the most resources behind it will regurgitate the most accurate information about what's on the internet today. Open source will always lose that race.

[yakbarber]

That's always been the case and was obvious to many from the start.

It really wont be that long until we see some ~GPT4 llm embedded locally in a chip on the next iPhone release...

[lukan]

Are you aware, what hardware is currently needed to run GPT4?

Something bigger than a smartphone usually.

So small mobile optimized LLMs will come, or are rather already there - but if they would manage to make the big GPT4 modell run on an iPhone, that would be a pretty big thing in itself, way larger than GPT5.

[petra]

But llms are relatively rarely used, and on the other hand, perf/latency is important to ux, and perf is variable(simple question, complex question, visual work).

Those demand are better fullfiled at the cloud.

[mushufasa]

Userbase and customer relationships are valuable. If someone else creates GPT5, but doesn't have a large user base, then OpenAI the company could buy that invention. Or, as we saw with deepseek in January, fast-follow with a comparable model within a reasonable amount of time.

Brands have value. If someone has logged into ChatGPT for two years daily, they have built a habit. That habit certainly can be disrupted, but there's a level of inertia and barrier -- something else has to be 10x better and not just 2x better.

When DeepSeek came out, I tried it out but didn't fundamentally switch my habit. OpenAI + Claude + Gemini instead caught up.

[danpalmer]

> Userbase and customer relationships are valuable

Which of these does Jony Ive's company have?

[srik]

The comment also includes brand value in the next paragraph and Jony has loads of that.

[danpalmer]

Does io have any brand value?

They're not acquiring Jony or Jony's design firm. They're acquiring the remaining portion of a joint venture. You could even say that LoveFrom is divesting from the joint venture.

[hansworst]

Following that logic, they’ll have to keep spending quite a bit to get to the user base of the current hyperscalers, some of which are already ahead of OpenAI in terms of LLM performance.

[kaliqt]

OpenAI would not be able to as every other company and governments even will make bids and OpenAI is not well loved to get favor to tilt the scale back in their direction.

[ivape]

There is a space to make a suite of products that synergize entirely. Glasses, watches, buttons, clothes (yes, clothes), and home devices/computers/tvs. The reason they are in a spearhead position is because unlike like Google and Apple, they don't need to maintain a legacy paradigm. They don't have to introduce new tech and make it work with old tech, while also maintaining usability familiarity (e.g You can't just change iOS and Android).

They take zero risk while attacking user fatigue (people just get bored of stuff). The current leaders take all the risk following OpenAI because everyone will complain about the changes no matter what they do, and just come up with a reason to switch. This is a human phenomenon that is truly fucked up, the same as when a partner in a relationship is ready to move on no matter what you do.

[grensley]

More like they see the future as more multi-modal, and they're probably right to think that is the best value approach vs. throwing more money at large language models.

[agentcoops]

I'm not so sure it's desperation. As an alternative hypothesis, we might simply view it as an attempt from a temporary position of strength to secure their tremendous lead as the primary consumer access point to intelligence. I don't think it's much of an exaggeration to suggest that this is one of the most important open questions at the moment -- one which will likely be relatively winner-takes-all (in contrast to the more commoditized B2B/API side) and where the winner likely won't be decided based on the intelligence side alone. The questions also aren't entirely separate since the winner, here, will have such incomparably valuable usage data...

Unlike most successful startups, OpenAI is not faced with the possibility that the giants (Apple, Google, Microsoft) decide to look their way, but the reality that these are their real competitors and that the stakes are existential for many of them (trends indicating a shift away from search etc). The most likely outcome remains that one if not all of the giants eventually manage to produce a halfway-decent product experience that reduces OpenAI to a B2B player.

[lelanthran]

> I don't think it's much of an exaggeration to suggest that this is one of the most important open questions at the moment -- one which will likely be relatively winner-takes-all

That makes the presumption that we are currently in a `winner-takes-all` scenario, and I'm not convinced that that is the case.

I'm not sure what the criteria is for a winner-takes-all scenario, but it is not at all evident to me that there is one now, or ever will be.

There is, as everyone says, no actual moat here: Google search had a moat, Windows Desktop had a moat, Apple phones had (and still have) a moat. LLM output currently has no moat, not even performance (both speed and accuracy) because the productivity difference between no-LLM and poor-LLM is about 100x the difference between poor-LLM and good-LLM.

My prediction is that the price of LLM usage will slowly but consistently climb until it reaches the floor on LLM cost-to-suppliers. Right now we are all (myself included) being subsidised by VC money. When the supplier has to actually turn a profit, there's no moat that they can use to keep out newcomers, because the newcomers need only a fraction of the money spent by (for example OpenAI) in order to compete.

Maybe Google has a moat, in that they have everything in-house, from the user-facing product to the tensor-processing hardware? That's as close to a moat that I can think off.

[threeseed]

> secure their tremendous lead as the primary consumer access point to intelligence

Yes because the only way to get access to intelligence is via ChatGPT which continues to lie and hallucinate on a regular basis.

Definitely can't get it via the web, books, videos etc.

[jhayward]

I don't think your conclusion of "hitting the wall on intelligence" is warranted.

It makes more sense to believe that scaling has hit the wall on available text data to train on, and that to continue scaling, along with whatever emergent properties arise they need much more data than exists as text.

There are orders of magnitude more data as video, audio, and images and this is what they intend to use to continue scaling.

[simmanian]

> and that what they want to call "GPT-5" is a fusion of their various models like 4o, dalle...

Do you have a source? I ask because I read the opposite.