07—09/22
Transmedia—DHBW
This is not only evident in the success of social media, where the focus is on observing others and self-staging, but also indicates that we initially only perceive what we can observe directly.
„The invisible stimulates the imagination more than the obvious“(2). This can be applied to the entire digital world and associated digitalisation: I cannot observe how many resources it takes to send an email, edit an image or use artificial intelligence - unless a small, colourful wheel turns, telling me that I have overused my computer and that it is now taking longer than a fraction of a second for once.
To make these invisible resources a little easier to understand, I draw a comparison between an architectural masterpiece of the past, the Egyptian pyramids, and an information technology masterpiece of the present, artificial intelligence.
When you receive a postcard with a picture of the pyramids on it, you might think „oh, I‘d like to go on holiday again“ or „they look really cool“ if you haven‘t already had your fill of the motif.
There are various theories as to how the pyramids were built - but most of them have been disproved or at least have not yet been proven. Which is why a certain myth surrounds the architectural wonder of the world is surrounded by a certain myth.
What resources were needed to build it? „2.6 million stone blocks with a minimum weight of 2.5 tonnes per block“(3) are estimated to have been used in one pyramid. The granite stones in the King‘s Chamber of the Pyramid of Cheops weigh 50 tonnes each.
The simple question of how these stones were transported and stacked has not yet been answered: wooden sledges for stone blocks and ramps and levers or machines (lifts, winches, cranes) for the granite stones? At the very least, it is clear that people alone cannot expend this energy. Thousands of labourers are said to have worked an elevenday week, ten days of ten hours each and one day off. However, they were not enslaved, as was previously assumed, but were probably well paid. There are different theories for the motivation to work:
„It cannot be assumed that a dictatorial pharaoh could have forced a large part of his people to co-operate in this mammoth project.“(4) More likely are religious ideas, occupational therapy or „as an overall societal journeyman‘s piece for eternity.“(5)
This can be wonderfully transferred to an AI-generated image of fictitious pyramids. Looking at it makes you think „holiday“ or „wow“, provided you have not yet had your fill of AI-generated images.
A myth is often woven around AI. The terminology artificial intelligence alone gives laypeople the impression that this computer could have a consciousness, which brings scifi fantasies to life. On the other hand, the non-publication (or incomplete publication) of the source code and the data used for training also deliberately mystifies the subject further.
So here again the question arises: What resources were needed for this?
OpenAI trained its GPT-3 model with 45 terabytes of data. Nivida ran 512 V100 GPUs for nine days to train the final version of MegatronLM, which corresponds to around 27,648 kilowatt hours (kWh).(6) These are figures that, like the 2.6 million stone blocks of the pyramids, are hard to imagine for a layperson. This is not hard physical labour, but dull, monotonous work with a high number of repetitions. These are needed to assess whether an AI-generated image is photorealistic, how an object is properly cropped, or what people look like when they perform activity X.
Like the Egyptians, these workers are not enslaved, but in contrast to Egypt, they are paid significantly less: for example, the average wage of so-called „crowdworkers“ from Amazon Mechanical Turk is around 2 US dollars per hour(7).
The crucial difference between the pyramids and the AI-generated image, however, is that we have already gathered a lot of knowledge about the pyramids and done a lot of educational work on the resources.
If we didn‘t know that the pyramids were built around 4500 years ago and without the most modern technology known to us, they would be far less impressive than the buildings in New York, Dubai or Singapore and certainly not a wonder of the world.
With AI-generated images, however, we generally lack the knowledge and imagination about the massive resources that were required to create them. In view of the circumstances described above, we can at least question how we as a society want to deal with AI and the associated products and processes in the future.
„We live in an observation society“(8). This essay is therefore not intended to call into question the use of AI or to determine its significance in comparison to the pyramids. Rather, the aim is to open our eyes to the superficially invisible in order to enable a reflective and socially acceptable positioning in the first place. Especially as this will only become more relevant in the context of the increasing digitalisation and further development of artificial intelligence.
(1) Meyer, Kim/Velbrück GmbH Bücher und Medien: Das konspirologische Denken: Zur gesellschaftlichen Dekonstruktion der Wirklichkeit, Konstanz, Deutschland: Velbrück Wissenschaft, 2017, S.228.
(2) Giesen, Bernhard: Zwischenlagen: Das Außerordentliche als Grund der sozialen Wirklichkeit, Weilerswist, Deutschland: Velbrück Wissenschaft, 2010.
(3) Bolten, Götz: Antike: Pyramidenbau, in: Antike - Geschichte - Planet Wissen, 21.04.2022, https://www.planet-wissen.de/geschichte/antike/pyramidenbau/index.html (abgerufen am 31.08.2022).
(4) Ebd.
(5) Ebd.
(6) Vgl. Labbe, Mark: Energy consumption of AI poses environmental problems, in: SearchEnterpriseAI, 26.08.2021, https://www.techtarget.com/searchenterpriseai/feature/Energy-consumption-of-AI-poses-environmental-problems (abgerufen am 31.08.2022).
(7) Vgl. Kotaro, Hara/Abi Adams/Kristy Milland/Saiph Savage/Chris Callison-Burch/Jeffrey Bigham: A Data-Driven Analysis of Workers’ Earnings on Amazon Mechanical Turk, in: arXiv, 2017, doi:10.48550/ARXIV.1712.05796.
(8) Meyer, Kim/Velbrück GmbH Bücher und Medien: Das konspirologische Denken: Zur gesellschaftlichen Dekonstruktion der Wirklichkeit, Konstanz, Deutschland: Velbrück Wissenschaft, 2017, S.228.
Frame-by-Frame Animation mit PyTTI-Tools Collab Notebook
RGB, Mute, 4 sec Loop
This project was developed while studying at the DHBW Ravensburg in the class of Transmedia held by Florian Tscharf.