OpenAI valued at $0.5 trillion

More Midweek Madness: GPT5 is out! Arguably, way out...

A box without hinges, key or lid, Yet golden treasure inside is hid. What am I?

(Likely not ChatGPT)

Soo, Sam Altman’s teaser Rogue One tweet promised the Death Star. Nearly six million people lapped it up; OpenAI is now valued at 500 billion dollars. At least, that’s the reported valuation.

Here’s part of a chat I had with v5 yesterday…

Me: Can you think of a mammal that swims in the ocean and wears trousers?

ChatGPT 5: That sounds like you’re baiting me into a pun.

I’d guess you’re going for a polar bear in “pants” — but the better answer is probably a manatee in “jeans”, which makes it a “man in tee and jeans.”

Or maybe you mean “otter pants”… as in, “other pants.”

Unless, of course, you want the full groaner:

A whale in trousers is a pair of pants.

Do you want me to try to work out the exact riddle answer you’re hinting at?

Me: I'm checking whether in your current incarnation (ChatGPT 5) you're any better at generalisation than previous versions, all of which failed to point out the non-twee answer that would delight any intelligent five year old human person.

It goes downhill from there on.¹ I’ve tried this before.^⌘ My current attempt merely underlines how these models struggle to generalise—and they’re not about to fix it. For a reasonable price, they can’t.^⌘ Possibly not for any price. The Artificial General Intelligence dream is not about to be realised. For a multitude of reasons,^⌘ it just can’t, not the least being that the concept of ‘AGI’ is itself incoherent. And again, dammit, Gary Marcus was right.

But perhaps I’m wrong?

What are others saying? Fortune magazine says that the above valuation “signals AI euphoria”. The really interesting part of their analysis is where they point out that the same value inflation is happening with their competitors: Anthropic, Perplexity, and even that scruffy-Nazi-saluter Elon’s xAI. If several people say they’re Jesus, ultimately we’ll discover that no more than one is.²

Maybe the Financial Times can help us out? Here’s their comparison with other companies that hit half a trillion:

They do point out that OpenAI simply needs to skim 5% off the combined market of Google, Microsoft and Apple. The question they ask is whether OpenAI can fend off the competition—and they seem to be remarkably upbeat. To me, this is all a teeny bit speculative. Where’s the moat?³

Pop?

A fair number of bright people have claimed that AI presents an existential threat to humanity, because it will take over. This doesn’t hold up to analysis.^⌘ But there is one way bots can inflict huge damage. If we funnel enough funds into our inflated belief that they will produce magic, and the bubble bursts, well then …

That Fortune analysis has a revealing quote from Nnamdi Okike, boss of 645 Ventures. He’s reported to have said “broadly we’re seeing classic signals of market euphoria and potential bubbles”.

I can’t help but agree. And as a minor aside, the CIO of ten-trillion-dollar Vanguard is now suggesting people pivot away from US stocks. Not that I am giving you investment advice in any size, shape or form, of course. But I’m driven to my own personal conclusion that anyone who talks about ‘rational markets’ at this point may end up with egg on their face.

My 2c, Dr Jo.

---

By the way, I first came across the ‘egg riddle’ at the start when I read The Hobbit as a child. I particularly loved the contest between Bilbo and Gollum:

He asked to gain time, until he could think of a really hard one. This he thought a dreadfully easy chestnut, though he had not asked it in the usual words. But it proved a nasty poser for Gollum. He hissed to himself, and still he did not answer; he whispered and spluttered.
After some while Bilbo became impatient. "Well, what is it?" He asked. "The answers not a kettle boiling over, as you seem to think from the noise you are making.”
   ― J. R. R. Tolkien, The Hobbit, or There and Back Again

Here’s my most recent interaction with ChatGPT in its new incarnation:

The Empire protects its own, I guess. It was not averse to generating this image, though:

Sam as the Good Guy. Let’s see whether this holds.⁴

1. The way GPT5 and similar models is trained, the bulk of the training data is ‘right’. It just so happens that the Internet provided most of those data, and the bulk of similar questions expect moronic puns as “the answer”. The model simply isn’t bright enough to think past this training.

2. I’m tempted to bet heavily on the number zero.

3. That protects them from the competition.

4. My apologies for using Comic Sans in my editing of the picture at the top. At this point, I would also like to thank all OpenAI investors for providing the uncopyrightable image at the bottom—free, to me. I didn’t pay a cent, you did. But, of course, I could have done it on a variety of competing platforms, even the despicable Grok. I just chose you to pay out, this time.