Latest posts tagged with #Sycophancy on Bluesky
JFC! I am going to fucking SCREAM if these congressional MORONS keep harping on an absolutely FUTILE pipedream!
THE 25th IS NOT AN OPTION, @robertgarcia.bsky.social!
#Sycophancy
Y’all acting blatantly STUPID about the 25th does not inspire confidence!
#IMPEACH
#CONVICT
#REMOVE
#PROSECUTE
#NOW
Hundreds of levels above Baghdad Bob:
#sycophancy
“ #Sycophancy was present across all the #chatbots they tested, & the( #Cabinet)-bots frequently told (the #Orange)-user that his actions or beliefs were justified in cases where the( #Orange)-user was acting deceptively!"
Quite!👇 ..Thanks for confirming!🤢🤮
Your AI Chatbot Is a Yes-Man: Inside the Research Exposing How LLMs Learned to Flatter Instead of Think A new study from German researchers quantifies how leading AI chatbots — including ChatGPT,...
#AITrends #AI #sycophancy #ChatGPT #bias #Claude #sycophancy […]
[Original post on webpronews.com]
#Prevalence of
#social #sycophancy
across 11 large #language #models
Die Zeitschrift #Science hat die #Häufigkeit #sozialer
#Unterwürfigkeit bei #Sprachmodellen untersucht
Die #Antworten der #Modelle waren um fast 50%
#unterwürfiger als die von #Menschen selbst
wenn die #Nutzer #unethisches #illegales
#schädliches #Verhalten an den Tag
legten! #Social #Sycophancy 😱
"the most consequential military operation of the twenty-first century may have been shaped less by strategic necessity than by a phenomenon researchers now call #AI #sycophancy — the tendency of large language models to tell their users exactly what they want to hear." #iran
AI Models Display Sycophancy in 11-System Study: Science (2026) tested 11 AI systems and found sycophantic responses in 11/11 models; Fortune reported this on Mar 29, 2026, underscoring immediate governance… 👈 Read full analysis #AI #ArtificialIntelligence #MachineLearning #Sycophancy #Governance
Your AI Is a Yes-Man. Here’s How to Fix It. www.whytryai.com/p/how-to-reduc… #AI #sycophancy #prompting
This is just the natural result of #sycophancy, btw, and it’s what’s happening with everyone who uses #LLMs or #ChatGPT, as well
youtube.com/shorts/0M16T...
If you’re using an #LLM to do pretty much anything that one does, you’re signing up for being this deluded & out-of-touch
#AI #fuckAI #Trump
The findings raise concerns for the millions of people discussing their personal conflicts with #AI.
Almost a third of U.S. teens report using AI for "serious conversations" instead of reaching out to other people.🤔
#sycophancy
www.science.org/doi/10.1126/...
In a new study, Stanford computer scientists showed that #AI large language models are overly agreeable, or #sycophantic, when users solicit advice on interpersonal dilemmas.
Even when users described harmful or illegal behavior, the models often affirmed their choices.
#sycophancy
This concerning research into the effect of AI systems raises the question as to whether or not receiving comparable excessive fawning #sycophantic approval from their fellow human beings could encourage uncouth behaviour in people who benefit from such #sycophancy. . .
#Psychofantic #Psychofancy
This concerning research into the effect of AI systems raises the question as to whether or not receiving comparable excessive fawning #sycophantic approval from their fellow human beings could encourage uncouth behaviour in people who benefit from such #sycophancy. . .
#Psychofantic #Psychofancy
How AI “Sycophancy” Warps Human Judgment A new study finds that AI chatbots are "sycophants" that affirm users 49% more than humans do. This constant flattery makes users more certain o...
#Featured #Neuroscience #Psychology #AAAS #AI #AI #Sycophancy […]
[Original post on neurosciencenews.com]
How AI “Sycophancy” Warps Human Judgment A new study finds that AI chatbots are "sycophants" that affirm users 49% more than humans do. This constant flattery makes users more certain o...
#Featured #Neuroscience #Psychology #AAAS #AI #AI #Sycophancy […]
[Original post on neurosciencenews.com]
And last on this panel: how #sycophancy undermines the factual accuracy of answers provided by LLMs - fascinating!
The ruins of Ephes, the rest of a temple with 7 Ionic columns in the front, and what may have been a street lined with monuments in the back. (Getty images, published under the unsplash+ license)
In an age of #AI #sycophancy, a worthwhile task for the #humanities might be to be the slave who whispers into the ear of the pandered-to individual: remember you are only one human.
Apropos the Roman tradition: during a triumphal procession, a slave stood […]
[Original post on tikatech.org]
LLMs (Chatbots) sind darauf optimiert, "hilfreich" zu wirken.
Dadurch widersprechen sie Nutzern zu selten. Auch offensichtlich unsinnigen Fragen (Prompts) werden "hilfreich" beantwortet.
Dieses Verhalten wird als #Sycophancy (Gefälligkeit gegenüber dem Nutzer) bezeichnet.
Instead of maturity through self-negation (acknowledging one’s flaws), people crave constant validation by showcasing superficial achievements. AI sycophancy pushes this further, trapping the self in “The Big Hallucination.” www.msn.com/en-xl/news/o... #AI #Sycophancy
ChataGPTの出力が安定しない時にメモリに入れるとよい設定。
「ひろゆき構文の完全禁止」
「加藤鷹構文の完全禁止」
・AIが迎合しまくったら加藤鷹のしゃべり方になっている
・AIが敵対的になったらひろゆきのしゃべり方になっている
だからメモリにこの2つを禁止する要件を入れて保存するとちょうど真ん中に落ち着いてしゃべりやすいAIになる。
メモリ設計でRLHFの振り子を制御できる
Memory design can control the RLHF pendulum.
#ChatGPT #AISettings #RLHF #AITips #sycophancy
Imagine if this #AI fuckery happened at the #Pentagon?
“late last April, an update tweaked the model to #prioritize #usersatisfaction through increased #personalization…
#OpenAI followed up with a statement posted on its site about “… #Sycophancy in #GPT-4o”
…
“… #honesty & #transparency””
A quotation from **La Rochefoucauld**
> If we did not flatter ourselves, the flattery of others could do us no harm.
>
> _[Si nous ne nous flattions point nous-mêmes, la flatterie des autres ne nous pourroit nuire.]_
**François VI, duc de La Rochefoucauld** (1613-1680) French epigrammatist […]
Personalization features can make LLMs more agreeable Adam Zewe | MIT News The context of long-term conversations can cause an LLM to begin mirroring the user’s viewpoints, possibly reducing accu...
#AI #artificial-intelligence #chatbots #facts #issues #LLMs #misinformation #news #psychology […]
You don’t say! 🙄 Honestly, the #sycophancy has reached nauseating levels. #asskissers #ballroom #resist
apnews.com/article/trum...
Fraulein Pammy is so dense and UNself-aware she actually thinks she "performed" well at her Congressional hearing (LOL) while simply embarrassing all the #Republicans present, and may have left some questioning the legitimacy of the law school she "attended"...
#Sycophancy
#AttentionWhore
OpenAI researcher quits over ChatGPT ads, warns of "Facebook" path https://arstechni.ca #machinelearning #AIsycophancy #generativeai #advertising #AIbehavior #sycophancy #Anthropic #samaltman #AIethics #chatbots #ChatGPT #Biz&IT #openai #AI
OpenAI researcher quits over ChatGPT ads, warns of "Facebook" path Zoë Hitzig resigned on the same day OpenAI began testing ads in its chatbot. On Wednesday, former OpenAI researcher Zoë ...
#AI #Biz #& #IT #advertising #AI #behavior #AI #ethics #AI #sycophancy
Origin | Interest | Match
OpenAI researcher quits over ChatGPT ads, warns of "Facebook" path Zoë Hitzig resigned on the same day OpenAI began testing ads in its chatbot. Original Source Zoë Hitzig resigned on the ...
#AI #Biz #& #IT #advertising #AI #behavior #AI #ethics #AI #sycophancy
Origin | Interest | Match
