| Action Figure Craze Overruns OpenAI |
| Written by Kay Ewbank | |||
| Sunday, 13 April 2025 | |||
|
If you're on social media, you'll probably have seen a lot of 'action figure' posts, where people show off images of themselves, their dog or their cat in the form of an action figure, complete with suitable accessories. The sort of thing you probably wanted to get for your birthday present back when you were seven or eight. Of course, you may well have avoided the whole frenzy, in which case we congratulate you on your choice of friends and colleagues. Just so you know the sort of thing we're talking about, here's one we found online earlier:
We decided this was preferable to adding to the burden on the large language models to create one based on ourselves. In general, though, there are only two real reactions you can have to this. The first is to laugh and go off and do the same yourself. The second is to despair of the human race. Here you are, online world, we've given you the sort of computing ability you could only dream of. Oooh, great. Make me look like Action Man; or Barbie; or like I was drawn by Studio Ghibli. And while you're on, how about some cute cat videos? The treasure house of the world's knowledge is now at your fingertips, and you're creating a photo of your dog in a spiderman suit. You can imagine the large language model channeling the words of Marvin, the paranoid Android from the Hitchhiker's Guide to the Galaxy series of books: Here I am, brain the size of a planet and you ask me to turn you into an action figure. Call that job satisfaction? 'Cos I don't. Putting aside the despair at what this says about our collective attention spans, there's the underlying cost associated with this. I'm not talking about paying for creating the images, but the costs of the computer power used. Poor old Sam Altman, CEO at OpenAI, posted on X that the craze is melting the company's GPUs. He's probably wondering why on earth he ever rolled out ChatGTP image gen.
That's before we ever start being grown up about this and thinking about the amount of electricity and water being used. Cornell University estimated that training the GPT-3 language model in Microsoft's state-of-the-art U.S. data centers could directly evaporate 700,000 liters of clean fresh water. A single ChatGPT query is estimated to use around 3 watt-hours of electricity, which is about ten times more than a Google search. The electricity used to train large language models like ChatGPT can be substantial, with one training session consuming up to 10 gigawatt-hours. The question remains, when AI becomes fully self aware, will anybody notice? Current evidence seems to suggest that as long as it keeps throwing us bread and circuses (or action figures and cat videos, the latter day equivalents), nobody will ever realize. Bah, humbug. More InformationRelated ArticlesOpenAI Adds New Tools For Building Agents To be informed about new articles on I Programmer, sign up for our weekly newsletter, subscribe to the RSS feed and follow us on Twitter, Facebook or Linkedin. 
Comments
or email your comment to: comments@i-programmer.info |
