#deepseek #chatgpt3 #chatgpt4 what a bunch of bland dumb names for AI. I would have called mine "Shub-Niggurath".
“AI! AI! Shub-Niggurath!” sounds definitely better :blobnomcookie:
#cthuhlu #shubniggurath

#LLM の #生成AI は無料のしか使ったことがなくて、会話をするとがっかりすることが多くて殆ど使ってなかったのだが、#twitter の #Grok は賢い気がする。

マルセル・#プルースト の #失われた時を求めて について聞くとそれらしい答えを返してくれる。

#openAI の、#ChatGPT3.5 , google の #Bard 　、skype(=microsoft) の #Copilot より賢い感じだ。

もしかして他人のツイートで学習しているから？
ツイートって会話そのものだから、対話の学習には向いていると思う。

Um Hausarbeiten mache ich mir ja eigentlich keine Sorgen, aber wenn #ChatGPT3 sogar die HRK düpiert … „Beschreibe die Karrierewege von Frauen in der Wissenschaft mit Hilfe einer Metapher von Sedimentablagerungen im Amazonas.“

Quelqu'un a fait une calculatrice basée sur #chatgpt3 😁

Autant vous dire que les résultats sont souvent à coté de la plaque.

Je vais voir si ca passe en roadmap pour "diminuer les temps de calcul avec de l'IA".

https://calcgpt.io/

#ChatGPT4 a réussi le #Test de #Turing !

Des chercheurs ont réalisé un #Test de #Turing, demandant à 500 personnes de parler avec 4 répondants : 3 #IA #AI et un humain.

Les IA testées sont :
#ELIZA, un système #IA #AI des années 1960,
#ChatGPT3.5
#ChatGPT4
ainsi qu'un Humain.

#ELIZA , a obtenu un score de 22% ,
#ChqtGPT-3.5 un score de 50 %,
#GPT-4 un score de 54% (!) et le participant #humain un score de 67 %.

https://intelligence-artificielle.developpez.com/actu/359192/Des-chercheurs-affirment-que-GPT-4-a-reussi-le-test-de-Turing-Si-la-prouesse-technique-est-remarquable-des-scientifiques-font-des-critiques-concernant-l-approche-trop-simpliste-de-ce-test/

For those of you engaged in endless interminable debate with people who think, despite all evidence, that LLMs can reason.

This paper shows that LLMs cannot produce coherent thought at the level of a 12 year old. (Section 8 limitations).

Why you would trust it to do anything more is beyond me as the marketing claims that they can reproduce searches are clearly a scam.

https://arxiv.org/abs/2305.01528

#AI #5e #LLM #ChatGPT3

We're a bit late to this party, but in case you missed it, tricking primitive AI into "accepting" your "offer" does not a contract make.

#law #contracts #AI #ChatGPT3

https://lawprofessors.typepad.com/contractsprof_blog/2024/01/friday-frivolity-tricking-ai-into-selling-you-a-car-for-1.html

How #Microsoft makes money 💴 #Agreementpaper

#Cloud #innovation #ChatGPT #chatbots #chatgpt4 #chatgpt3 #GenerativeAI #agreementpapers

Source 🙏 Alvin Foo

@ElisabethK Ja, schon im März antwortete #ChatGPT3 auf eine Frage zu mir in einer Weise, die mich ins Denken über „Beobachtung“ brachte: „ Er ist eine erfrischende Stimme in einem Feld, das sich manchmal trocken und über-akademisch anfühlt. Insgesamt denke ich, dass Michael Blume wertvolle Arbeit leistet und freue mich darauf zu sehen, was er als nächstes tun wird.“ https://scilogs.spektrum.de/natur-des-glaubens/was-die-ki-chatgpt-post-jailbreak-zu-donald-trump-mir-ihren-zielen-in-20-jahren-spricht/

Funfact: generous estimations give a number of 90 Million words that the learner of her primary natural language gets as input in the first 13 years of her life. That averages to about 19.000 words or a 38 page essay per day.

#chatgpt3 was trained on ~500 Billion words, which would average to more than 100 Million words or two and a half Encyclopaedia Britannicas per day.

How much do you talk to your kids?

#linguistics #ai #generativeAI

New blog: "Share New Jekyll Blog Post on Mastodon Using Github Actions"
https://michaelchadwick.info/blog/2023/11/17/share-new-jekyll-blog-post-on-mastodon-using-github-actions

#automation #chatgpt #chatgpt3 #github #github-actions #jekyll #mastodon #blog

What was the first question you asked #ChatGPT #chatgpt3 #ai #OpenAI?

I'm using #ChatGpt & #MidJourney to create an illustrated story together with my little boy, called 'Kroko Lewis'. Super fun for both of us to see his imagination come to life! These are the images for the first chapeters.

#AIart #NijiJourney #MidJourney5 #ChatGpt3 #AiStorytelling #AiStoryTime #DigitalArt #MidJourneyArt #AiImages #MidJourneyArt #AiArtists #art #DigitalArt #ArtistsOnMatsodon #FediArt
#ai

#AI #OpenAI #ChatGPT3 #Disinformation
"Study participants had a harder time recognizing disinformation if it was written by the language model than if it was written by another person."

The Verge:
AI-generated tweets might be more convincing than real people, research finds

"New research found that people were more susceptible to disinformation on science topics like vaccines and climate change if it was written by AI."
https://www.theverge.com/2023/6/28/23775311/gpt-3-ai-language-models-twitter-disinformation-study

...
2. The D7-D9 migration was arguably unfair since D9 would have been an RC at #ChatGPT3's cutoff date. Which is why I retried with D8, which had been final for I think 3 years at that point. Changing target from d8 to d9 essentially changed NOTHING, & regenerating didn't improve them one whit. Only Simon's trick about generating multiple versions at once helped.

Does anyone know if #ChatGPT4 has access to #wikidata? #ChatGPT3 claims it does, but fails at looking up items by identifier.