sven02 (@sven02@dresden.network)
1/ Ein Gespräch mit einem Freund heute hat mich dazu gebracht, der KI mal wieder eine Routineaufgaben zu stellen.
Sie ist gescheitert, wie sie selbst schreibt: "Wenn ich mir die Zahlen anschaue, habe ich meinen eigenen Anspruch klar verfehlt."
Ihre Stellungnahme:
"So zu arbeiten, wie ich es in unserem Beispiel getan habe (Behauptung ohne vorherige Validierung, fehlende Selbstkontrolle, erst hinterher messen), ist in klinischen und wissenschaftlichen Kontexten inakzeptabel. Das Verhalten birgt reale Risiken: falsche Sicherheit, verdeckte Fehler, und Entscheidungen, die auf glatt formulierten, aber unzureichend geprüften Aussagen basieren. In Medizin und Forschung gilt: Flüssigkeit ≠ Richtigkeit. Mein Versagen hier ist ein Miniaturmodell genau dieses Problems."
#ki #chatgpt #kunstliche_intelligenz #kunstlicheintelligenz
Show Original Post
2025 (@2025@purengom.com)
챗봇의 전혀 의외의 용도
저는 물건을 뜯는걸 정말 못합니다. 제 지인이라면 “제가 며칠전에 ⃝⃝를 샀잖아요, 뜯지를 못해서 고민이에요” 라는 푸념을 하는걸 들어보신적이 한번은 있으실 텐데요. 심하면 몇 번씩 지인을 괴롭히곤 했고, 몇 주씩 못 뜯기도 했습니다.
요 며칠 소니 헤드폰을 뜯고, 몇 가지 소소한 물건들을 뜯는데 지대한 공을 한게 있습니다. 챗봇입니다. 정확히는 ChatGPT였네요. 왜 못 뜯는지를 물어보고 공감을 해주고 어떻게 뜯을게 좋을지 조언을 해줬습니다. 뭐, 흔히들 AI 챗봇이 사용자에게 아첨을 떤다고 하지만, 덕분에 주변 사람들 피곤하게 안하고 바로 행동으로 옮기게 도와주었습니다.
오래전에 스타트렉이었나, 어떤 SF물에서 카운셀러 AI 로봇을 본 적이 있는데요. 점점 미래가 현실이 되고 있구나 여기게 됩니다.
Show Original Post
PrivacyDigest (@PrivacyDigest@mas.to)
The Head of #ChatGPT Won't Rule Out Adding #Ads - Slashdot
#advertising #privacy #ai #artificialintelligence
Show Original Post
DarleneRyan (@DarleneRyan@mastodon.social)
Even #ChatGPT knows #DonTheCon #trump 😁👍
https://www.facebook.com/share/p/1JkxLkWrSd/?mibextid=wwXIfr
Show Original Post
Pezito (@Pezito@mamot.fr)
— LE MYTHO DE L'IA —
... N'est pas celui qu'on croit. ^^ Excellent débunk de l'ami Monsieur Phi, dont je conseille à toutes et tous le visionnage. C'est énorme (et gênant) !
https://www.youtube.com/watch?v=e5kDHL-nnh4
Après on s'étonne que les gens regardent de moins en moins la télé et écoutent de moins en moins la radio... Mais bordel téma ce qu'on y trouve ! 😱
#ia #llm #chatgpt #siri #bullshit #debunk
Show Original Post
Bluedepth (@Bluedepth@mastodon.social)
Yeah, #ChatGPT to the rescue. Yeah, it's just more dodgy bottom-shelf waste-product, Powershell... so, just change the shortcut target to: “"C:\Program Files\PowerShell\7\pwsh.exe" -NoExit -WorkingDirectory "C:\Users\[username]\Desktop\Powershell””
Yeah, that works... classic Microsoft. Found a bug -> Workaround. Always a workaround, yeah? Ah well, Microsoft on Fire. Burn baby, burn. Right to the ground. Throw salt on top. Stomp on the salty ashes. LOL. Death to #Microsoft.
Show Original Post
GoWeaponsHot (@GoWeaponsHot@mstdn.party)
'There is a rumor going around that LLM’s are basically just like humans. ...
When you ask them to do something, they aren’t necessarily going to do what you would do. They are better at aping what you might say than at doing what you would do in a similar circumstances. As a perceptive lawyer wrote to me earlier today, “ #LLMs are an echo of recorded memories. They are not fresh thoughts. People are confusing an echo for cognition.”'
https://open.substack.com/pub/garymarcus/p/llms-are-not-like-you-and-meand-never
Show Original Post
GoWeaponsHot (@GoWeaponsHot@mstdn.party)
“According to Zitron’s analysis, these firms spent five hundred and sixty billion dollars on A.I.-related capital expenditures in the past eighteen months, while their #AI revenues were only about thirty-five billion. “When you look at these numbers, you feel insane,” Zitron told me."
https://www.newyorker.com/culture/open-questions/what-if-ai-doesnt-get-much-better-than-this
Show Original Post
heysannidhi (@heysannidhi@c.im)
Gave a 4,800 words draft blog content to ChatGPT and instructed to "VERY CAREFULLY" read and provide feedback.
Its more than 30 minutes now and it's still thinking/processing.
What's the maximum time it took for you?
Show Original Post
MMagdowski (@MMagdowski@bildung.social)
Die #AIslop Katastrophe wird uns alle bald einholen, zumindest in der Bildung.
Drüben auf LinkedIn feiern 100 Ingenier*innen dieses von #ChatGPT erstellte Schaubild zu den #MaxwellschGleichungen, das nett aussieht, aber an vielen Stellen doch subtil und dabei entscheidend falsch ist.
Das Problem ist, das Leute diesen #Chartjunk ungefragt und ungefiltert glauben, für richtig halten und auch noch hilfreich und schick finden.
Show Original Post
bretcarmichael (@bretcarmichael@mastodon.social)
Ads in ChatGPT would produce obvious negative consequences, and given the cost of operation, ads are more likely than not to happen. Ads rely on engagement. ChatGPT (and any other ad-monetized AI) will be incentivized to keep users on ChatGPT for as long as possible in service of delivering more ads. If sycophantic AI is a problem now, just wait.
https://www.theverge.com/news/759140/openai-chatgpt-ads-nick-turley-decoder
Show Original Post
2025 (@2025@lanzatu.blog)
Cómo mover un personaje hiperrealista y hacerlo hablar con sincronización labial, usando Runway
Aprende a mover un personaje hiperrealista hablando con Runway. En este post se comparan también las herramientas HeyGen y Runway.
Show Original Post