Can chatting with #AI help people spot #deepfakes?

Individuals and groups tried to identify machine-generated text, either with or without a deliberative #chatbot (DeliBot).

Groups outperformed individuals. The #LLM merely improved engagement.

https://doi.org/10.48550/arXiv.2503.04945

Chatbot-Apps - Wie gefährlich sind Beziehungen mit einer KI?

Es ist schon lange keine Fiktion mehr: Menschen führen Beziehungen mit KI-Chatbots. Eine Studie zeigt, warum sie das tun. Doch von den Chatbot-Apps gehen auch Gefahren aus, wie Vollbild-Recherchen zeigen.

https://www.tagesschau.de/investigativ/swr/vollbild-ki-bots-100.html?at_medium=mastodon&at_campaign=tagesschau.de

An #AI Customer Service #Chatbot Made Up a Company Policy—and Created a Mess

When an AI model for code-editing company #Cursor #hallucinated a new rule, users revolted.

https://www.wired.com/story/cursor-ai-hallucination-policy-customer-service/

An AI Customer Service Chatbot Made Up a Company Policy—and Created a Mess

https://arstechnica.com/ai/2025/04/cursor-ai-support-bot-invents-fake-policy-and-triggers-user-uproar/

When an AI model for code-editing company Cursor hallucinated a new rule, users revolted.

The automation of online undercover work:

Are you chatting to someone real or a bot?

404 says "Our reporting shows that cops are paying a company to help them deploy AI-powered bots across social media and the internet to talk to people they suspect are anything from violent sex criminals all the way to vaguely defined “protestors” with the hopes of generating evidence that can be used against them. "

https://www.404media.co/this-college-protester-isnt-real-its-an-ai-powered-undercover-bot-for-cops/

#USpol #Media #UKpol #CANpoli #EUpol
#Meta #WhatsApp #MetaAI

"Meta AI: Jetzt kommt der Rechtsruck für den Whatsapp-Chatbot

(1/n)

Seit einigen Wochen ist die Künstliche Intelligenz von Meta in Diensten wie Whatsapp und Instagram verfügbar. Nun kündigt der Konzern an: der #Chatbot soll bald politisch „ausgewogener“ sein. Was zunächst harmlos klingt, könnte vor allem eines sein: ein neuer Kniefall vor der Politik #DonaldTrumps...
...
In den vergangenen...

https://www.kn-online.de/wirtschaft/meta-ai-whatsapp-chatbot-bekommt-politischen-rechtsruck-TP2VTDORK5F45CV2DAEK3MKJYE.html

I used to be able to search for things using a search engine. It would point me to a source of information, I could decide for myself if the source was good or not. But not anymore. Recently I just wasted a lot of time trying to find something using a search engine only to get lost in an ocean of AI slop.

I finally gave in and asked my question to ChatGPT and it answered my question perfectly, it even showed me where it got it’s sources of information. The places it linked me to were either no longer existent (if ever) or only vaguely related. But it at least created the appearance of citing factual sources of information, in whatever way the statistical models have been trained to give that impression to the humans who use it.

In Orwell’s novel “1984,” we saw how a single government agency took all control over information, how the staff would keep a database of facts but constantly alter these facts to fit the view of reality that their despot “Big Brother” wanted the citizens to believe. The same thing happens now with LLMs presenting a very carefully tuned image of the facts in the database.

AI chat bots are an innovative new method of brainwashing, one which nobody could have ever imagined possible even a few years ago: get everyone to trust a mechanical Big Brother because it chats with you like a friendly fellow human would.

(Full article here.)

Meta to train its AI models on public data of European users.

Meta on Monday said it would incorporate user interactions with its AI, as well as posts and comments by adults across its platforms, to train its artificial intelligence models in the European Union.

Users will still be able to opt out of having their data used.

https://mediafaro.org/article/20250414-meta-to-train-its-ai-models-on-public-data-of-european-users?mf_channel=mastodon&action=forward

Moodle.ia : un chatbot qui crée vos prompts pour générer des activités au format xml, gift, cloze ou H5P !
Et le tout grâce à ChatMD

https://chatmd.forge.apps.education.fr/#https://codimd.apps.education.fr/s/5HamcDX2f

#moodle #elea #h5p #chatbot #xml #gift #qcm #education #e-learning @moodlemootfr @lelibreedu @apps @les_tutos_elea_h5p @remi_lefeuvre @gauchov @Cathlemon @eyssette @bertrandchartier

More lessons learned using an #ai #chatbot to help write code. (See my original post for some of the other observations).

First, watch out for vibe coding. I start off asking it to do things I know and things I could double check if I wanted to. And it's really easy to start asking it to do things that I don't know how to do, and that I can't check.

Second, save checkpoints. I'm inviting the #LLM to modify the code files. Sometimes it modifies unrelated stuff that used to work and makes it not work, or makes it different. I ask it to calculate alt text, and it changes how the histogram is plotted. As I iteratively get things working, I do a git add to capture a working version. A couple times I let it make a series of changes and by the time I got to the last change, it had clobbered some intermediate states that were working, and I didn't have a copy of the code that was there previously. At each incremental step, VS Codium can use git information to highlight which lines have changed due to the most recent modifications.

Last, it goes without saying, I must go in behind it and clean things up. A funny thing it tends to do is get really attached to adjectives. I wanted my trend line to be yellow, and in order to be clear, I prompted with "add a 15-minute moving average trend line that is bright yellow". Everywhere it mentioned this trend line (in comments, in the alt text) it always said "a bright yellow trend line" instead of just "a yellow trend line". When I asked it to make the histogram bins "exactly one minute" (again, trying to be unambiguous), it then labels everything like the X axis label and the alt text with the phrase "exactly 1 minute" instead of just "1 minute".

The EU will probe the use of personal data by X's AI chatbot.

Ireland's data protection watchdog, acting on behalf of the European Union, launched an inquiry on Friday into the use of personal data posted on Elon Musk's social media platform by EU users, to train its AI chatbot, Grok.

https://mediafaro.org/article/20250411-the-eu-will-probe-the-use-of-personal-data-by-xs-ai-chatbot?mf_channel=mastodon&action=forward

#ChatBot simple pour introduire l’intelligence artificielle #IA avec des collégiens.

Merci @eyssette pour #ChatMD

https://chatmd.forge.apps.education.fr/#https://codimd.apps.education.fr/F2Pd3ZfDREa1afghO57TrA

LLM-Anwendungen werden aggressiv an (akademischen) Bibliotheken beworben. Meine Kolleg*innen sind mehrheitlich begeistert: „Endlich einmal sind wir auch vorne dabei“.

Könnt ihr deutschsprachige Referent*innen im Bibliothekskontext empfehlen, die da kritische Aufklärung betreiben und die man zu einem Vortrag einladen könnte?

Internal support Slack channels:

Bob: Does anyone know how to X?

#AI #Chatbot: it looks like you're trying to X, but I can't find any threads

Alice: You'll need to do Y first, then X is available. Here are docs.

Bob: Thanks, Alice!

AI Chatbot: You're welcome! I am very smart!

WRITER FUEL: 'ELIZA,' the world's 1st chatbot, was just resurrected from 60-year-old computer code.

https://www.limfic.com/2025/04/04/writer-fuel-worlds-first-chatbot-resurrected-from-60-year-old-code/

#HowTo disabling the new #chatbot from #Firefox 137

› Put « about:config » in the address bar
› Search the « browser.ml.chat.enabled » key
› Switch to « false »
› Voilà !

Or use a #Firefork

This is how to disable the new “AI” chatbot in #Firefox:

• type about:config into the awesome bar
• skip the warning for first time modders
• locate the browser.ml.chat.enabled setting and set it to false

In the #Librewolf fork, a thoughtful person has already done this for you.

(HT to @kuketzblog for the hint!)

"First Therapy Chatbot Trial Yields Mental Health Benefits"
"...The researchers conclude that while AI-powered therapy is still in critical need of clinician oversight, it has the potential to provide real-time support for the many people who lack regular or immediate access to a mental-health professional...."
https://home.dartmouth.edu/news/2025/03/first-therapy-chatbot-trial-yields-mental-health-benefits?utm_source=www.therundown.ai&utm_medium=newsletter&utm_campaign=ai-s-gold-standard-therapy-skills&_bhlid=188ef8b64336158e7c11155be8b1334957070054
#AI #healthcare #chatbot

ChatGPT: generatore immagini con GPT-4o gratis per tutti
#AI #ChatBot #ChatGPT #FunzioniAI #GenAI #GeneratoreImmagini #GPT4o #Gratuito #ImmaginiAI #IntelligenzaArtificiale #Limitazioni #Notizie #Novità #NuoveFunzioni #OpenAI #SamAltman #TechNews #Tecnologia

https://www.ceotech.it/chatgpt-generatore-immagini-con-gpt-4o-gratis-per-tutti/