“Apple, Nvidia, and Anthropic Turned into Virtual Kleptomaniacs, Stealing Thousands of YouTube Videos to School their AI”
“Apple, Nvidia, Anthropic Used Thousands of Swiped YouTube Videos to Train AI”
“Despite all of its promises, artificial intelligence has so far been synonymous with escalating energy use. Just think about the power needed to run deepfakes or other AI projects, and the tons of carbon emissions that come with it.” This is just another day at the office, bringing up the glaring shortcomings of the burgeoning technology of Artificial Intelligence.
So, the great minds at work are training an AI with YouTube videos, intriguing! It’s as if the invention of fire wasn’t enough, now humanity wants to give machines the power of understanding ‘how to make a smoothie’ or ‘funny cat videos’. It’s meant to allow machines to grasp ‘physics’ better. Alas, the irony is, the ‘physics’ they refer to involves watching drops of water fall in slow motion rather than attempting to unearth quantum dynamics.
Shifting gears, the naming convention of companies, like, Anthropic, expresses intent of shaking the very foundation of, well, ‘thermodynamics’. How’s that for a little comic relief? Names aside, the ‘save the planet’ slogan must have been hidden somewhere in their blueprint when they decided on err… computer chip designs that ‘suck more power’.
And what about Apple jumping in the field of AI computer chips? This move was as predictable as their next ‘slim and trim’ version of an iPhone. Thankfully, they care more about low-power devices. Talk about true power saving. Excuse the pun.
In this grand AI saga, Anthropic’s playbook seems to be borrowing some leaves out of OpenAI’s book– ‘train large language models on the internet.’ Well, just hope there’s a chapter somewhere about filtering out internet noise.
An interesting paradigm was thrown in by Nvidia, preaching about reducing the cost instead of increasing efficiency. A classic, ‘look at the cost, not the value’ scenario. Remember those budget First-Person Shooter games that barely ran, let’s pray it doesn’t turn out the same.
So, here we are, at the forefront of AI, a technology that’s supposed to map human intellect, belching out carbon emissions and scouting YouTube for ‘training’. Well, at least, it’s doing what it’s been universally finalised to do – to learn. It will learn, adapt, evolve, and most importantly, ’emit’. Welcome to the ‘thermodynamically efficient’ yet smoke-billowing future!
(Views expressed here are personal and not necessarily represent that of the tech company)