Larry Ellison basically explained why OpenAI is positioned to win the AI race. Today’s models all trained on the same thing: public data. The internet. Books. Papers. General knowledge. Impressive, but capped. Ellison’s point: peak value only comes when AI is trained on private data. That’s where OpenAI is... moving: • partnerships with universities to work with unpublished research • experts teaching models real job tasks, not just theory • deep access to how science, medicine, law, and finance actually work Public data makes AI smart. Private data makes it decisive. The next AI leap won’t come from reading more of the internet. It’ll come from being invited behind the curtain. That’s the real moat.show more

Except that the Genesis Mission just nuked the moat by ensuring that all that high value private data becomes public. Data wants to be free

Also the data OpenAI is receiving by the Ivy League universities they have partnered with? They are scanning unpublished research and work in giant quantities.

@DaveShapi Have you read about how much of that research is now being written by AI also? OpenAI is reading data published by ChatGPT. 🤣

@DaveShapi Nonsense, that’s really old research. Some are a century old.

You realize that Google by far has the most acces to private data ? ( Gmail, Google Drive, Android, all the stuff running on Google Cloud etc)

Actual private personal data from users of their services like Gmail and Google Cloud? I see a huge lawsuit incoming.

They also use your Gmail contents to generate personalized ads, so they are probably also using your emails to train ai models

Wouldn’t surprise me that somewhere in the TOS and privacy policies of the Google services it is buried that you consent to let them use your data for ai

> That’s where OpenAI is moving And other LLM builders are dumb and not moving there? Or OpenAI doesn't let them move to where this guy told you the moat is?

If OpenAI doesn’t win the AI race, Oracle's 300B contract with OpenAI won't be fulfiled - cause of the enormous debt. Another biased talk and no reason to believe a word

i’m with @DaveShapi on OpenAI 😎🍿

@DaveShapi This will age very badly in less then a year.

Google's been building these sort of datasets even before it was useful for training NNs. I'm sure OpenAI is doing a great job to collect this data too, but don't see them having a real advantage.

Um, what? Google has access to your email, your Google Drive, the way android peeps use their phones. They’ve been at it for 20 years

Ellison is right — private data makes AI much stronger and more useful. But think about what that really means. The true advantage goes to companies that get exclusive access behind closed doors. We move away from open knowledge toward gated, private intelligence.

Except they’re losing.

Thisisretarded

it will be a race between full vertical integration (goog) and agile partnerships (openai). if google over indexes on their current stack, openai will surely be able to acquire enough major partnerships to push the envelope in private data

Read the architecture. Then rethink blockchain. #Qubic

Unpublished research? So, bad data?

AI slop comment about AIs future. Ironically closer to what the actual future is than it's content.

Human slop comment….

Nahhh Google wins

OpenAI isn’t winning in any of these spaces

This framing assumes that private data access is the limiting factor, but I'm not convinced that's where the bottleneck actually is. Most enterprises already have proprietary data. The challenge is that the data itself is messy, siloed, and poorly structured. Training on it requires massive cleanup and domain expertise that doesn't scale easily across industries. Ellison's point about experts teaching models is closer to the real unlock, but that's extraordinarily labor-intensive. You can't just feed an AI medical records and expect it to understand clinical decision-making without significant human oversight.

lol so it’s exclusive to them? I don’t think so

OAI wil go way of the Nokia. Only one Company has the full LLM stack, from TPUs to datacenters, and above all, massive FCF pockets

Google still wins there, sorry. Next time, Sam

If the winning strategy is "form partnerships to access private data," what stops every other well-funded AI lab from making the same phone calls to the same universities and hospitals?

First of all, I believe Gemini & Grok have edge over ChatGPT, if it's about data. Secondly, tech companies should stop considering AI as a race and start approaching AI collaboratively to be force for good.

I agree that that probably will make one/two models beat the others, but it wont be OpenAI doing this - if anything it will be nail in the coffin for OpenAI because they NEED Microsoft for this.

"Public data makes AI smart. Private data makes it decisive." this is the entire thesis OpenAI getting university research data, medical records, legal docs = massive moat

Definitely, a new round in the development of AI begins when the stage in becoming a humanoid creature that understands primitive human psychology and needs has passed. Then this is a deep private deepening into science and political narratives of states However, in fact, I do not think that only OpenAI has access and builds up such data capital. The result of the race is not so unambiguous yet

There will always be data that isnt accessible. Like classified or proprietary data. And much of governmental data requires foia lawsuits to obtain. Full transparency would revolutionize the planet. Especially in areas of control. There will always be data that isn’t accessible, not just because of bad actors, but because power systems are designed around asymmetry. Classified data, proprietary data, national security carve-outs, trade secrets, and even mundane bureaucratic opacity all exist because control depends on who knows what, when, and under what conditions. FOIA lawsuits are revealing in that sense. The fact that you often need legal force just to see what your own government did years ago tells you transparency is not the default state. It’s an exception that has to be extracted. What’s interesting is that full transparency wouldn’t just be a moral upgrade, it would be a structural revolution. Control systems rely on narrative gaps. When people can’t see flows of money, decision logic, risk transfer, or failure rates, authority fills the gap with explanation. Once the data is visible, explanation loses power. The system either works on its own terms or it doesn’t. That’s why transparency is so threatening in areas of control. It collapses plausible deniability, but more importantly it collapses ambiguity. Ambiguity is where coercion hides. When decisions are opaque, accountability is negotiable. When data is visible, accountability becomes mechanical. At the same time, I don’t think total transparency is fully achievable in an absolute sense. Some information genuinely has defensive value if asymmetric, and some systems would become fragile if every internal detail were exposed in real time. But that doesn’t weaken your point. It sharpens it. The real question isn’t “can everything be transparent,” but “which data must be transparent for power to stay legitimate.” That’s where the revolution actually lives. Not in exposing everything, but in forcing transparency at the interfaces where decisions touch reality: spending outcomes, enforcement actions, model assumptions, risk allocation, success and failure metrics. When those are visible, control systems can’t drift into narrative-only governance. In a way, full transparency would do to political power what open ledgers did to accounting. It wouldn’t make humans honest. It would make dishonesty expensive, obvious, and harder to sustain at scale. And once that happens, a lot of structures that depend on confusion rather than performance simply stop working. I think that’s why transparency is always promised rhetorically but throttled operationally. It’s not just about secrets. It’s about preserving a world where authority explains itself, instead of proving itself.

Smart AI is everywhere. Decisive AI isn’t.

Private data that have huge value are owned by banks and Wall Street. I suppose...

Sure it is. OMG what idiots. No one can train private models using deepseek or kimi? of course they can and they are. Dunce caps all around for this idiot call.

isn't this the domain of Google?

Many already mentioned that Google has all the data in the world. They won this already.

He is missing a key gap of OpenAi vs the other big two: 1. Google has YouTube and innumerable other assets and real world inputs 2. xAI has Twitter, Tesla. OpenAI is positioned the weakest for real world data imo and buying a few ad hoc publications or Disney catalogues etc won’t change that vs Google and xAI

@grok ​Grok, shall you reveal to us the veritable truth concealed behind Ellison's words, were we to forsake the customary, superficial prattle regarding universities ?

Google has 1,000x more private data than OpenAI, and moreover, it is in an AI-ready format. Imagine just the data regarding ads, search, the web journeys of billions of users, Cloud, Gmail, YouTube, Docs, Sheets, and Android. Wow. This is almost infinite