Zephyrnet Logo

Why Nvidia’s CTO is Wrong on Crypto OR How Crypto Made AI

Date:

If your presentation is struggling to get much attention amid all the happenings, especially with banks, mentioning crypto is still a somewhat easy way to get some of that attention.

And so soon after a 78 minute presentation by Nvidia’s CEO Jensen Huang last week, where he touched on promises to bring 2 nanometer chips among much else, we get Nvidia’s CTO Michael Kagan go on the Guardian to trash crypto.

“All this crypto stuff, it needed parallel processing, and [Nvidia] is the best, so people just programmed it to use for this purpose.

They bought a lot of stuff, and then eventually it collapsed, because it doesn’t bring anything useful for society. AI does,” Kagan said.

We could of course claim sour grapes after ethereum kicked out not just Nvidia, but all hardware based mining through the upgrade to Proof of Stake.

There was a slightly different tune back then, but there’s a far more serious point to be made, a point you’d expect someone like Kagan to be very aware of.

As it happens there’s a very little known connection, and a fairly fundamental one, between crypto and AI.

Indeed one can even claim the AI we are seeing now is because of crypto, at least in regards to it being out at this point rather than much later.

To make that point, we have someone far more credible than Kagan, someone at the driving wheel of that AI. And that’s none other than Google’s co-founder Sergey Brin.

In a letter to investors back in 2018, which puzzled us back then yet was the first sign of what was to come, Brin said:

“There are several factors at play in this boom of computing… The second factor is greater demand from the GPU-friendly proof-of-work algorithms found in some of today’s leading cryptocurrencies, such as Ethereum.

The profound revolution in machine learning that has been building over the past decade is both made possible by these increasingly powerful processors and is also the major impetus for developing them further.”

It is customary to forget that the crypto space is fundamentally raw code, and the hardware that runs that code.

We drove GPUs through demand, sometimes extraordinary demand, to be far more powerful. Just as importantly, we provided the incentives for many talented men and women – or even ‘kids’ – across the globe to play and tinker with GPUs to make them as efficient as possible to get an edge on mining.

As it happens those GPUs are also useful for other things, like machine learning or ChatGPT models, but these are not really things ‘kids’ can play with, or smart young men that nonetheless don’t run global corporations.

You needed huge resources, certainly in 2018, to develop these models and train them, limiting what can be called open and global participation.

In contrast in crypto the cost was the GPU you use to play gaming, which you can send off to mining, and go google to see how it can be made more efficient, and so you get a whole global community focusing on advancing GPUs, and then you get the likes of Brin or Sam Altman use those advances to now give us Stackoverflow on steroids.

About three months after Brin made the above comments, he went to a crypto gathering where he stated that innovations in this space, like zero knowledge proof, were “mind-boggling.”

That’s pure maths in many ways and has specific applications in this space to scale public blockchains through second layers, but what wider applications it may have in the years to come remains to be seen.

Zk-tech also needs hardware, like asics, though to a far lesser extent than a public blockchain, but the demand for bitcoin mining asics has also led to some advances in semiconductors.

In both instances, it is through opening up these spaces. For asics far more directly as crypto focused startups began designing and producing specialized asics hardware, while for the GPU duopoly more indirectly in as far as the public gained an incentive to tinker with the hardware.

That has benefited all activities that rely on GPUs, and so the dawn of some version of public AI has crypto to thank to some extent.

When someone like Kagen says cryptos are not useful however, he is of course not concerned with these finer details.

For him first of all, he is probably concerned about investors worrying that the overnight vanishment of the ethereum GPU market might affect Nvidia’s profits.

But more widely, these blind biased individuals are frankly more concerned with ignorance. Either not bothering to be informed even on areas that are related to their business, or finding it easier to form cheap opinions on areas they know little about, or indeed in some cases they’re trying to pull the wool over you, or the public.

As, at least where AI is concerned and advances in GPU as well as computing power, crypto has been very useful.

spot_img

Latest Intelligence

spot_img