Per alcuni gli #LLM sono gli oracoli digitali: si chiede e si hanno risposte
Uno studio di AI4TRUST ha dimostrato che #GPT-4o, #Gemini, #Llama e #Grok hanno fornito il 100% di risposte sbagliate a domande relative alla salute
I prompt erano costruiti per favorire l’errore, ma gli LLM non hanno applicato sistemi di controllo e verifica, rivelandosi inattendibili
Uno scenario ideale per gli agenti della #disinformazione
My bros, the reason your chatbot can't even follow its chain of reasoning, evidently they get confused once they have to abstract over more than a couple paragraphs and they shit the bed.
you've scaled the token window and the algebra of your attention heads and latent space, but maybe the relationship to attention as humans do it as the shape of that curve you love so much.
high and to the right, was it? or is that just Musk's nickname?
All my system prompts invite the LLM kindly to say "I don't know" when it doesn't (it doesn't always work, it depends on the LLM), it saves a lot of time (and compute time, and energy, etc)..
Why are thousands of philosophers analysing the future of humans? (13 August 2018)
https://www.bbc.co.uk/programmes/p06h9c4k
Do you use AI/a LLM on a regular basis?
If so, which one do you prefer?
Do you pay a monthly subscription for one?
Boosting appreciated :)
Un #GPT pour créer facilement votre #glossaire sur #Éléa, #Moodle ou #Magistère. Vous choisissez le thème, discutez avec l’IA et récupérez un fichier XML clé en main. Un vrai plus pour tous les #enseignants et #formateurs ! 
https://chatgpt.com/g/g-68504553d8948191a50e286362dd3d26-creation-glossaire-elea
Nowerdays competition:
#AWS (replace with your cloud provider) bill ----->|
#Cursor (replace with your #LLM / #GPT provider) bill --------->|
But the winners are power plant owners, or/and fuel suppliers...
Lossers in that competitive game are the most parts of the planet with whether anomalies touched by heatwaves.
NEW Working Paper: “The #Attribution Crisis in LLM Search Results: Estimating Ecosystem Exploitation” https://www.infodocket.com/2025/06/29/working-paper-the-attribution-crisis-in-llm-search-results-estimating-ecosystem-exploitation/#AI #GPT #LLMs #SSRC
Speak faster… or give me tokens… nah, just speak faster.
#transcriptions #llm #gpt #openAI #ffmpeg #ffmpeg4live
https://george.mand.is/2025/06/openai-charges-by-the-minute-so-make-the-minutes-shorter/
@GeePawHill We can no longer get away from the colloquial #AI as a generic term, it's in people's heads and hashtags.
That's why it makes more sense to *specify* what we mean.
#MachineLearning for dealing with huge databases is something different from #generativeAI that creates 7 fingers after stealing images, or a hallucinating Generative Pre-trained Transformer = #GPT that uses #LLM.
(Most people think GPT is a fantasy product name and don't know the meaning of the acronym.)
@tante SIA published a paper to address AI-specific assessment guidelines: https://www.iaia.org/uploads/pdf/SP16_AI%20in%20IA.pdf
Src:
Bingham, C., Bond, A., et al. (2025) Principles for Use of AI in IA. Special Publication Series No. 16. Fargo, USA: International Association for Impact Assessment.
For the ~six of y’all that might be interested in the GUID partitioning table storage partitioning — used on many modern systems including macOS, OpenVMS I64 on Integrity Itanium, most anything using EFI or UEFI consoles, and others — here is the Apple Technical Note TN2166
Secrets of the GPT document:
https://developer.apple.com/library/archive/technotes/tn2166/_index.html
This being one of the better GPT introductions available.
This and related technical discussions also have all unfortunately been buried by that unrelated AI stuff.
Anthropic releases custom AI chatbot for classified spy work - On Thursday, Anthropic unveiled specialized AI models design... - https://arstechnica.com/ai/2025/06/anthropic-releases-custom-ai-chatbot-for-classified-spy-work/ #artificialintelligence #largelanguagemodels #nationalsecurity #aiconfabulation #machinelearning #aiassistants #airegulation #anthropic #microsoft #aiethics #chatgpt #chatgtp #biz #claude #openai #gpt-4 #ai
I have asked #GPT-4.0 to translate a news report from Dutch to German for me because I wanted to inform my family about current debates in the Netherlands. I was surprised that #ChatGPT started a #deepsearch although I had not actively selected this feature. I had vaguely heard about it but never used it before, and I am still wondering what it actually does. I will have to #readthedocs for sure! But has anyone got insights they'd like to share? Deep search sounds a lot like #energywaste to me.
How many runways do you see for Gulfport Biloxi International airport (USA) ? : The answer is on https://www.bigorre.org/aero/meteo/kgpt/en #gulfport #usa #gulfportbiloxiinternationalairport #kgpt #gpt #airport #aviation #avgeek vl
Dis Mastodon, tu utilises ou a évalué #Perplexity vs #GPT ou #Claude ou autre ?
SI oui, un retour m'intéresse.
Wer sich über die vielen tollen Informationsangebote im Internet freut, sollte wissen:
#KI 
randaliert im Netz – #Admins halten dagegen, damit wir Menschen ungestört surfen können.
Lest mal, wie Admins ihre absolut frustrierende aber unsichtbare Abwehrarbeit gegen KI beschreiben – im Blog von @campact: https://blog.campact.de/2025/05/ki-randaliert-im-netz-admins-halten-dagegen/
Nicht vergessen: 25. Juli ist #SysAdminDay
GPT Group ( #GPT ) has released " GPT announces new $1bn Australian logistics Partnership " on Thu 15 May at 18:25 AEST #today #Australia #media #Strategy #live
https://grafa.com/asset/gpt-group-2138-gpt.asx?utm_source=asxmktsensitive&utm_medium=mastodon&utm_campaign=gpt.asx
🅸🅶
🆁🆁🅴.
