[10 December 2023]: "Primitive Hypertext" — pyschopomps; when AI hallucinates; perineuronal nets and rebooting critical periods; AI and the extralinguistic sphere; ethiopian jazz
[10 December 2023]: "Primitive Hypertext" — pyschopomps; when AI hallucinates; perineuronal nets and rebooting critical periods; AI and the extralinguistic sphere; ethiopian jazz
Bismillah. We begin everything with the name of Allah. We say Bismillah to initiate an act to acknowledge the intention and the ethics we carry with all that follows Bismillah.
This is part of the newsletter's “Primitive Hypertext” (Octavia Estelle Butler) strand.
An annotated list of five things I’ve read/seen/heard and want to share. [weekly: every Sunday] Read more about the changes in the newsletter rhythm here.
Hiya!
For the past few Sundays, I have either been on a plane or getting off a plane, and I haven’t been able to make my primitive hypertext offerings. Thank you for your patience, and read below.
I had this word sprawled across my notebook, but I could not remember what it meant or why I had written it down. I think it was on a day when I was thinking a lot about death rituals.
This one was fun! I had to reread it to ensure I didn’t misunderstand.
Brown argues that hallucination is when external perception does not match up to reality (specifically, consequences). First, Brown explains that Chat GPT can only be concerned with collocation (or, the way words combine in a language to produce natural-sounding writing). He continues, reminding us that because the output is bound only by the statistical probability certain words go together and not by their plausibility on Earth, the output runs the risk of being humorously absurd or scarily dangerous (and this reminds me of some summer reading: ”Amazon recently removed a guide on foraging for mushrooms that some readers claimed was generated by AI and could have given incorrect advice about what mushrooms were edible or poisonous.”). There is a gap, Brown says, between “external perception” (the machine learning output) and reality (the world of consequences and dangers activated by following the output’s instructions). Simply, the output is untethered from a local context. The Nautilus pulls an example from this August 2023 article which I am quoting below:
One recipe it dubbed “aromatic water mix” would create chlorine gas. The bot recommends the recipe as “the perfect nonalcoholic beverage to quench your thirst and refresh your senses”“Serve chilled and enjoy the refreshing fragrance,” it says, but does not note that inhaling chlorine gas can cause lung damage or death.In this example, clearly, the AI is hallucinating, according to Brown, because “it’s less that these are factual errors, but fall into a wider problem of groundedness in real-world concerns, death being a rather important one to us carbon-based lifeforms.” I think this returns to some earlier notes on machine learning, where I note some things about machine learning and novelty — basically, in collaboration with these notes and this new article, I am interested in AI-generated text because I want to know more about its edges. Like what are the limits?
Groys, Boris. From Writing to Prompting: AI as Zeitgeist-Machine. E-Flux. August 10, 2023.
More on AI! Similar to the above, but maybe from the other direction? I am interested in this assertion about this “garbage pit into which every new text is thrown as merely an additional piece of garbage.”
Eeeeeeeek! This one made me squeal! So, boom: scientists have focused on the inside of neurons and less so on the viscous coating on the outside, or perineuronal nets (PNNs). These guys are “composed of long chains of sugar molecules attached to a protein scaffolding, and they hold neurons in place, preventing them from sprouting and making new connections.” But what does it mean that they cannot make new connections?: Possible addictions and PTSD memories are “locked in” because the PNNs “form over neurons and lock neural network wiring in place at the end of the critical period.” Like a shell or crust? In neuroplasticity, the critical period is the specific time window when each sensory brain region is more plastic or sensitive to changes, influence, and adaptations, and other studies (this one, on mice) suggest that this time window is different for somatosensory, visual, and auditory sensory areas. The Quanta article explores attempts to fiddle with PNNs by reopening that critical period to have a bit of a do-over for a locked-in memory or habit that is too heavy to carry.
Ethiopian Jazz w/ The Grey Area [Through The Years] (Noods Radio): Tesfa Mariam Kidane - Heiwote Mulatu Astatke - Chifara Mulatu Astatke - Gubelye Seifu Yohannes - Mela Mela Mahmoud Ahmed - Kulun Man Qualesh Mahmoud Ahmed - Fetsum Dink Lij Nesh Alemayehu Eshete - Chiro Adarie Negne Girma Beyene - Ene Negne By Manesh Hailu Mergia & Dahlak Band - Bati Bati Muluken Melesse - Tenesh Kelebe Lie Girma Beyene - Set Alamenme Teshome Meteku - Hasabe
Thank you for reading,
Kameelah 👽
Finally, while I do not organize my finances around paid newsletter subscriptions, wouldn’t it be cool if this wee little newsletter could allow me to take quarterly self-imposed writing retreats? Consider getting a one-year membership at $70 USD :)
How to cite this newsletter: Rasheed, K. (Year, Month Day). Newsletter Title. I Will (?) Figure This All Out Later. URL