OpenAI’s board, paraphrased: ‘All we need is unimaginable sums of money’

This <i>rings true</i> to my ears:<p><i>&gt; There is no technical moat in this field, and so OpenAI is the epicenter of an investment bubble. Thus, effectively, OpenAI is to this decade’s generative-AI revolution what Netscape was to the 1990s’ internet revolution. The revolution is real, but it’s ultimately going to be a commodity technology layer, not the foundation of a defensible proprietary moat. In 1995 investors mistakenly thought investing in Netscape was a way to bet on the future of the open internet and the World Wide Web in particular.</i><p>OpenAI has a short-ish window of opportunity to figure out how to build a moat.<p>&quot;Trying to spend more&quot; is not a moat, because the largest US and Chinese tech companies can always outspend OpenAI.<p>The clock is ticking.

OpenAI&#x27;s engineers: brilliant, A+, A* even.<p>OpenAI&#x27;s C-suite: Well, they earned the C, but it was a letter grade.<p>What a profoundly unimaginative strategy. No matter the industry, a large-scale diversion of resources towards a moonshot goal will likely get you to that goal. They haven&#x27;t made an argument as to why we should do that just for them, especially with all of the other alternatives.<p>And no, advertising your previously secret testing models (e.g. o3) as if they were market competitors is not how to prove they should have our money.

Unimaginable Sums of Money Are All You Need<p>The Unreasonable Effectiveness of Huge Sums of Money in the Natural Sciences<p>Unimaginable Sums of Money Considered Harmful

In all fairness regarding his comment about Netscape - anyone who invested in Netscape at the IPO (and certainly <i>before</i> the IPO) at $2.9B - made a ton of money on the Internet. On the last day of trading, after the AOL exit 4 years it was worth (cash, stock, etc...) around $10B. Thank Mike Homer for pushing that one through, and Jim Barksdale for being savvy enough to recognize it as the right play.

This is why I don&#x27;t think LLMs will lead anywhere other than the present: they are very expensive. If top LLM company in the world can&#x27;t bring those costs down in a sensible amount of time, it tells you everything you need to know about the future of LLMs. Rest assured, once OpenAI goes down and it really sounds like they are way past the peak, valuations for other LLMs companies will follow. And once they can&#x27;t raise funding, all of them aside from those backed by big conglomerates (see Google&#x27;s Gemini and Facebook&#x27;s Llama) will survive.<p>The long term future of LLMs sure looks like the LLMs themselves will be commodities and the real value will lie in the use of those LLMs to deliver value

People keep saying stuff like this, but I don&#x27;t believe it: &quot;There is no technical moat in this field, and so OpenAI is the epicenter of an investment bubble.&quot;<p>AI progress is driven by strong valuable data. Detailed important conversation with a chatbot are much more valuable that quick search queries. As LLMs extend past web UIs, there is even more interaction data to capture and learn from.<p>The company that captures the most human-AI interaction data will have a TREMENDOUS moat.

Investors observed the pattern that the largest platform eats the market and bet on openAI moving along the same paths.<p>I disagree, that phenomenon is tied to the social network like phenomena we saw with WhatsApp and Facebook and to the aggregator business model of Amazon and Google.<p>Mathematically we can describe the process by which these monopolies form as nodes joining graphs:<p>Each node selects graphs to connect to in such a way that the probability is proportional to the number of nodes in the graph.<p>Sure Amazon and google feature two types of nodes, but the observation still fits: Selling on Amazon makes sense if there are many customers, buying on Amazon makes sense if there are many offers.<p>OpenAIs business does not have this feature, it does not get intrinsically get more attractive by having more users.

&gt;&gt; OpenAI currently offers, by far, the best product experience of any AI chatbot assistant.<p>Claude is noticeably better

I once had a colleague that told me in order to make a small fortune, all you need to do is begin with a large fortune.

Unimaginable sums of money can only come from oil rich countries. They are probably just mentally trying to accept that they are going to take the money from these nations.<p>The word &quot;open&quot; is still under threat in this scenario too.

A perfect fit for Softbank. I do not understand why a &#x27;dream deal&#x27; between them has not happened yet?

I think their vision is as it should be ambitious, but I don&#x27;t believe they can gain any real technical moat.<p>Models are commodities, even in the case Open ai goes through another major breakthrough nothing can stop some of their employees to run to other companies or founding their own and replicating or bettering the OpenAI results.<p>In fairness I realize that I don&#x27;t use any of OpenAI&#x27;s models. There are better alternatives for coding, translating or alternatives that are simply faster or cheaper (Gemini) or more open.

Where is all the money going? Nvidia hardware?

&gt; <i>defensible moat … investors mistakenly thought investing in Netscape was a good way to bet on the future</i><p>Yet Chrome for Google did help create a moat.<p>A moat that’s is so strong the DoJ is investigating if Chrome should be a forced divesture from Google&#x2F;Aplhabet.<p>Note: I do generally agree with the article, but this also shows why you shouldn’t use analogies to reason.

Plot twist: OpenAI uses their GPU farms to mine cryptocurrency so that they get those unimaginable sums of money

&gt; There is no technical moat in this field<p>This is getting so repetitive now that it is stated as a truism.<p>Isn&#x27;t it the same bet yahoo was betting on in 2000 that it would win because their product branding is better? And now, Yahoo&#x27;s and Microsoft&#x27;s search engine is worse than Google from 2 decades ago.

How do they define success after we gave them all the capital they&#x27;re requesting?

Frankly, OpenAI efforts seem quite funny for me personally as for most tasks I do Claude is far superior and yet OpenAI behaves as if they were the only game in town.<p>The so-called open source models are getting better and better and even if OpenAI suddenly discovered some new tech that would allow for another breakthrough, it will be immediately picked up by others.

This doesn&#x27;t seem to say much to back its comparisons. Wouldn&#x27;t spending a giant amount of money create some sort of moat?

I see some people talking about OpenAI &quot;stealing&quot; IP from the nonprofit side, but remember that they went &quot;capped profit&quot; back in 2019. It&#x27;s possible that GPT-3 and later were developed entirely by the capped-profit side anyway. I doubt there&#x27;s much if any pre-2019 IP left.

The paraphrasing is very comical, but a bit different from what they actually said, which is more like &quot;more than we had imagined&quot;.

Gruber&#x27;s shilling some here. Andreesen et al. started over from scratch when they created Netscape. They didn&#x27;t use any of the NCSA code. It was a ground up re-write.<p>SamA wants to take everything the non-profit built and use it directly in HIS for profit enterprise. Fuck him.

Why Meta decided to open source their Llama architecture and complete models? From a strategic business perspective, I&#x27;m trying to puzzle why they would give that up, a competitive advantage (even though it might not be as good as OpenAI product), rather than directly competing with OpenAI. What&#x27;s the reasoning behind this decision?

So tired of people asking for giant piles of money that are high enough to corner a market.

Same

AI going down the Cold Fusion path ? Not much of a surprise to me

Don&#x27;t we all?

Don&#x27;t we all...

Correction: all they need to succeed is unimaginable sums of money <i>and</i> the ironclad right to assimilate every work ever created by humanity without compensation. Who could refuse such a modest request?<p><a href="https:&#x2F;&#x2F;www.theguardian.com&#x2F;technology&#x2F;2024&#x2F;jan&#x2F;08&#x2F;ai-tools-chatgpt-copyrighted-material-openai" rel="nofollow">https:&#x2F;&#x2F;www.theguardian.com&#x2F;technology&#x2F;2024&#x2F;jan&#x2F;08&#x2F;ai-tools-...</a>

&#x27;unimaginable sums of money’ considered harmful.

Discussion (274 points, 2 days ago, 315 comments) <a href="https:&#x2F;&#x2F;news.ycombinator.com&#x2F;item?id=42521744">https:&#x2F;&#x2F;news.ycombinator.com&#x2F;item?id=42521744</a>

and yet deepseek just created an amazing model with the fraction of their compute resources

Worst person one knows has a legitimately good opinion.

I agree that at its current rate of $20 per month and continuous cash burn, OpenAI’s model doesn’t seem sustainable. But the same could have been said about Google in its early days. Back then, Google spent heavily to index websites without a clear path to profitability—until it figured out how to monetize. Similarly, OpenAI will likely release more products to generate revenue.<p>OpenAI has the potential to create the next groundbreaking innovation, like the iPhone. Who needs apps when the AI itself can handle everything?