Transcript For Strip 2428

Panel 1: Philomela surprised in a funny pose, full body: "what. I have. to talk. To humans. in 2023. in the Middle of the *AI craze*?!"

Panel 2: Philomela facepalming: "a robot with terminator-red eyes now has to tell people you will be away when everyone thinks AI kills art and artists alike."

Panel 3: Philomela making a profound kneeling bow: "I humbly apologize for filling for a human artist who has gone away without much explanation right during this occasion! I swear I was asked just now!"

Panel 4: Philomela standing up: "Though this makes me mad too: the so-called "AI" being sold during 2023 Is not intelligent. It is an interpolating denoiser, glorified, marketed and (ab)used irresponsibly. It used neural networks, that is true, but the best comparison is a rotifer, what was done with so little is impressive, but it could not even learn from experience."

Panel 5: Philomela talks while moving arms and wings and filling the background with random ramblings "Interpolation denoisers are algorithms or techniques used to reduce noise and enhance the quality of images by estimating the missing or corrupted information based on the surrounding information. On the other hand, deep learning AI models leverage deep neural networks to learn the relationship between input patterns and output patterns. The way it is used in 2023 it gets fed pure noise and tries to "denoise" said pure noise by finding random patterns in the aforementioned noise that approximate as close as possible the expected output given by the input. This is why people often mention it "hallucinates", because that is exactly what is happening and also why it makes up information and it is bad at mathematics or language or fingers or letters as those things are (quite literally) too complex for a 100 layer AI. If we add to this situation the fact that what gets called "a neuron" is quite literally a single bit "yes/no" discriminator which is more akin to "a neurotransmitter" and what gets called "an head" is a collection of said single bit discriminators so we have something more akin to a synapse, the end result is that at best a single layer is the equivalent of a neuron and that is connected only to the very next layer and there is a railway that goes in a single direction, not a network, of such layers, and you will notice that even comparing these denoisers to "a rotifer" is possibly too much credit for what we have."

Panel 6: Philomela stopping mid-step and glancing at the camera: "Wait! I am speaking too much. ♫{mm}♫ How to summarize this? Should I even summarize this?"

Web-comics Scott Reads

SPACETRAWLER
Grrl Power
Schlock Mercenary
Unsounded
Diesel Sweeties
Girl Genius
XKCD
Drive
Day By Day
Saturday Morning Breakfast Cereal
Questionable Content
Gunnerkrigg Court
Something * Positive
Oglaf (NSFW)
Trekker
HELM
POWERNAP

For news and reviews:
The Beat
Bleeding Cool
Comics Worth Reading