Aardvark Daily

Please visit the sponsor!

They're coming for your RAM and your cash

The universal justification for any form of control on the internet used to be "won't someone think of the children".

Now a similar approach is being used to try and convince you that if you've got more than a few bytes of RAM in your computer then you're just being greedy.

Sam Altman of OpenAI has gone on record stating that he thinks people shouldn't have VRAM in their personal computers.

Instead of being so selfish, that VRAM should be used to further help the build-out of AI datacentres so that everyone can share that memory for the greater good (translation: for OpenAI's hope of making a profit, eventually).

It strikes me that Mr Altman is finally waking up to the reality I presented in a recent column where I pointed out that the future of AI as an industry is in the provision of locally hosted LLMs, not in the old "mainframe" model of centralised processing through AI datacentres.

Altman clearly recognises that the single biggest threat to the profitability or even just the survival of his own AI business is the ability for people to build and use their own LLMs and AI agents, running in-house on their own hardware.

In an attempt to squash this killer-competition to his centralised "AI as a service" model, he wants the supply of high-bandwidth memory (VRAM and HBRAM) to be directed almost entirely towards AI hyperscalers like his own business.

The rationale being proposed is that when it's used in a privately owned computer, that RAM is only used for a very small percentage of the time but if it's instead used in an AI datacentre then it will be utilised 100 percent of the time. That, claims Altman, is a far more efficient use of valuable resources such as this form of memory.

Nice try Sammy boy, I see what you tried to do there.

The sad (for him) reality is that the AI as a service (AIAAS) model is about to crash and burn, likely right about the time that Anthropic's IPO due to launch later this month, fails to win over investors.

If you're read SEC filing for the Anthropic IPO then you have to wonder why *anyone* would invest a single red cent in this company. Here is the most sobering of the declarations:

Seriously, the company is admitting to having staggering losses with the very real possibility that it may never achieve profitability? Why would you throw money at this.

But wait, it gets worse!

Elsewhere in the filing it goes on to mention that their offering could effectively be used by bad actors to create horrific weapons and engage in attacks on a massive scale. In fact they even use the phrase "and may pose catastrophic or existential risks to humanity".

Come on, seriously... who would fund that?

I sit here in utter bewilderment that anyone would take a business to the market with so many warnings, at a time when faith in the profitability of AI as a whole is in question.

If this IPO succeeds and meets its target goal then I will have lost all faith in the critical thinking abilities of the human race.

Perhaps they're relying on people to simply ask AI whether they should invest... and perhaps they've already trained their AI to respond positively to the proposition?

Ah well, lets see what the next few weeks bring.

Carpe Diem folks!

Please visit the sponsor!

Here is a PERMANENT link to this column

Rank This Aardvark Page

Change Font