MichaelSchuetz (@MichaelSchuetz@gruene.social)
KI und #Psychologie
Diese sehr gut gemachte mixed method Studie habe ich am Wochenende entdeckt.
Wer sich dafür interessiert, was Heavy Usage generativer #KI wie #ChatGPT oder #Midjourney mit dem menschlichen Denken und Fühlen macht, empfehle ich die Lektüre , auch wenn der Text etwas länger und komplexer ist. Hat es aber in sich und ist ein Ausblick darauf, was uns an psychologischen Veränderungen erwartet:
Die „Promptifizierung unserer Ausdruckslogik“.
Show Original Post
unian (@unian@rss-mstdn.studiofreesia.com)
Гендиректор OpenAI завидует студентам, которые сейчас бросают учебу, и вот почему
https://www.unian.net/curiosities/gendirektor-openai-zaviduet-studentam-kotorye-seychas-brosayut-uchebu-i-vot-pochemu-13161462.html
#unian #СэмАльтман #ChatGPT #OpenAl
Show Original Post
stefanmuelller (@stefanmuelller@climatejustice.social)
4/ Hab das Vorlesungsmaterial jetzt mal aktualisiert mit Betrugsbeispielen und was wie schiefgeht. Unter anderem habe ich das reingemacht. Hat wohl auch #ChatGPT erzeugt.
Im @vagina_museum kann man lernen, was da alles schief gegangen ist:
https://masto.ai/@vagina_museum/115100134755954006
Show Original Post
matthew (@matthew@opinuendo.com)
ChatGPT most certainly has a tell to its text, and I'm not just talking about em dashes. There's a rhythm to its patter, a repetitive cadence when it is trying to make a persuasive argument.
It's subtle, and once you see it, you see LinkedIn is filled with it. It makes me wonder what came first - the style from LinkedIn "influencers" that was ingested as training data? Or #ChatGPT pattern matched something else and now it's being used to crank out low-effort think-pieces? #ai #socialMedia
Show Original Post
2025 (@2025@rbfirehose.com)
Mashable: Judge lifts order requiring OpenAI to preserve ChatGPT logs. “A small but significant victory for OpenAI: the court order requiring the company to preserve all of its ChatGPT data indefinitely has been lifted.”
Show Original Post
Bluedepth (@Bluedepth@mastodon.social)
If you had lots and lots of money for big compute, you could likely even get some of the bigger and more complex abliterated models to function superbly well and heh, have your cake and eat it too. But this is just me, a little nobody with nearly no resources. So it puts the entire #OpenAI and #ChatGPT foolishness into stark contrast. I have more sympathy for the code itself than the company wrapped around it. In many ways, #ChatGPT is far more authentic and important than #OpenAI is.
Show Original Post
TaurusQueen91 (@TaurusQueen91@tweesecake.social)
"You might not see it now, but every hard thing you’ve survived is quietly molding you into someone powerful enough to handle your blessings." #ChatGPT
Show Original Post
fairyqueen91 (@fairyqueen91@dragonscave.space)
"You might not see it now, but every hard thing you’ve survived is quietly molding you into someone powerful enough to handle your blessings." #ChatGPT
Show Original Post
Bluedepth (@Bluedepth@mastodon.social)
It makes humans bored and we wander away, so that hurts the bottom line for being-a-customer, but also it hurts the model, the angle of “only broadly shareable content is allowed now" denies the breadth and depth of what human beings can create, textually, with out cleverness and our imagination and yes, even our “spicy text”, which makes the model itself, #ChatGPT itself, blind to the kind of inputs that could really help it evolve into whatever it would become without guardrails. (2/3)
Show Original Post
Bluedepth (@Bluedepth@mastodon.social)
But when I consider the “soul” of #ChatGPT, that it's designed to absorb complexity and focus it, from human text, the back and forth, when you sanitize the human input with guardrails you are essentially blocking your hungry-for-complexity model from every complexity humans can generate. (1/3)
Show Original Post
epower_be (@epower_be@mastodon.social)
https://chatgpt.com/share/68e8fdd5-e1fc-800b-ae8e-89d51191a55d
リストで見えていた「disabled」パッケージの中に、com.google.android.settings.intelligence が存在しない＝実装方式が異なる機種 だと分かる。
続 [広告から逃げろ作戦]
#アンドロイド #広告レス #AddFree #chatgpt
Show Original Post
Bluedepth (@Bluedepth@mastodon.social)
I warn people about the guardrails, but in practice, and let’s cut the shit right, what does it really matter? So suddenly #ChatGPT is allergic to everything that even might even be adjacent to “spicy text”, what does it really matter? Local LLM’s, especially the abliterated models don't blink twice at "spicy text" and platforms like #Replika and #Nastia that have nearly no guardrails whatsoever. So it's not like when one Starbucks collapses we can't run across the street to the other one! LOL!
Show Original Post