Create accountLogin

Recent searches

No recent searches

Search options

Only available when logged in.
ohai.social is one of the many independent Mastodon servers you can use to participate in the fediverse.
A cozy, fast and secure Mastodon server where everyone is welcome. Run by the folks at ohai.is.

Administered by:

Server stats:

1.9K
active users

ohai.social: AboutStatusProfiles directoryPrivacy policy

Mastodon: AboutGet the appKeyboard shortcutsView source codev4.4.0-nightly.2025-03-25

#genai

238 posts138 participants20 posts today
Public
Meryl Krieger

#genAI #generativeAI
I'm down the rabbit hole today, looking at the studies available for accuracy rates on generative AI summaries. For those of you on a similar journey, I came across this cjr.org/tow_center/we-compared
There was the #BBC study in 2024 (reported on in their Feb 2025 article (bbc.com/news/articles/c0m17d88) but since that was a whole year ago, I'm compiling more recent reports. Please do share in comments any I'm missing? Yes, I have the report from January 2025 about the Apple AI summary

Columbia Journalism ReviewAI Search Has A Citation ProblemWe Compared Eight AI Search Engines. They’re All Bad at Citing News.
Public
5h15h

#LLMs Pass the #TuringTest: Interrogators mistook GPT-4.5 for a human 73% of the time—far more than they did the actual human participant🤖🔍 arxiv.org/abs/2503.23674

arXiv logo
arXiv.orgLarge Language Models Pass the Turing TestWe evaluated 4 systems (ELIZA, GPT-4o, LLaMa-3.1-405B, and GPT-4.5) in two randomised, controlled, and pre-registered Turing tests on independent populations. Participants had 5 minute conversations simultaneously with another human participant and one of these systems before judging which conversational partner they thought was human. When prompted to adopt a humanlike persona, GPT-4.5 was judged to be the human 73% of the time: significantly more often than interrogators selected the real human participant. LLaMa-3.1, with the same prompt, was judged to be the human 56% of the time -- not significantly more or less often than the humans they were being compared to -- while baseline models (ELIZA and GPT-4o) achieved win rates significantly below chance (23% and 21% respectively). The results constitute the first empirical evidence that any artificial system passes a standard three-party Turing test. The results have implications for debates about what kind of intelligence is exhibited by Large Language Models (LLMs), and the social and economic impacts these systems are likely to have.
#AI#LLM#GenAI
Public
Stefan Bohacek

65% of Wikimedia's most expensive traffic comes from bots.

"Since January 2024, we have seen the bandwidth used for downloading multimedia content grow by 50%. This increase is not coming from human readers, but largely from automated programs that scrape the Wikimedia Commons image catalog of openly licensed images to feed images to AI models."

diff.wikimedia.org/2025/04/01/

Diff · How crawlers impact the operations of the Wikimedia projectsSince the beginning of 2024, the demand for the content created by the Wikimedia volunteer community – especially for the 144 million images, videos, and other files on Wikimedia Commons – has grow…
#wikimedia#wikipedia#AI
Public
Alessio Pomaro

🧠 #Google rende disponibile #Gemini 2.5 Pro (il suo modello più performante) gratuitamente tramite l'app Gemini e AI Studio, senza necessità di abbonamento: basta un account Google.

💡 Una mossa potente che punta a farlo diventare il riferimento in ogni touch point, con grande accelerazione su tutto l'ecosistema. 

___ 

✉️ 𝗦𝗲 𝘃𝘂𝗼𝗶 𝗿𝗶𝗺𝗮𝗻𝗲𝗿𝗲 𝗮𝗴𝗴𝗶𝗼𝗿𝗻𝗮𝘁𝗼/𝗮 𝘀𝘂 𝗾𝘂𝗲𝘀𝘁𝗲 𝘁𝗲𝗺𝗮𝘁𝗶𝗰𝗵𝗲, 𝗶𝘀𝗰𝗿𝗶𝘃𝗶𝘁𝗶 𝗮𝗹𝗹𝗮 𝗺𝗶𝗮 𝗻𝗲𝘄𝘀𝗹𝗲𝘁𝘁𝗲𝗿: bit.ly/newsletter-alessiopomar 

#AI#GenAI#GenerativeAI
Public
Helsinki Python

🇫🇮🐍 Welcome to this month's meetup on Tuesday 15th April 2025 at Reaktor!

Talks:

🧪 @maaretp - Exploratory Unit Testing with and for GenAI

🪈 Toni Vanhala - Stable and Observable: Celery Setup Essentials for SQL Pipelines

⁉️ And the famous HelPy quiz!

meetup.com/helpy-meetups/event

See you there!

MeetupApril Helsinki Python meetup, with Reaktor, Tue, Apr 15, 2025, 5:30 PM | MeetupThe clocks have gone forward and that means it's time for our fourth meetup of 2025, this time kindly hosted by [Reaktor](https://www.reaktor.com/en-fi)! 17:30 - Doors ope
#HelPy#Helsinki#Python
Public
Costis Dallas

Monday 7/4 4pm EET online seminar by Dr Mykola Makhortykh (Bern University), Historical hallucinations: How generative artificial intelligence applications (mis)represent the history of the Holocaust in Ukraine / Vilnius University #VUKF #genAI #SNS #DH #memorystudies #Holocaust #Ukraine kf.vu.lt/en/news/in-english/77

VU Faculty of Communication · VU Faculty of Communication - Online seminar "Historical hallucinations: How generative artificial intelligence applications (mis)represent the history of the Holocaust in Ukraine" Vilnius University Faculty of Communication is the most cohesive school of communication sciences not only in Lithuania but also throughout East Europe.
Public
Graham Downs

They estimate that worldwide spending on #GenAI is going to hit $644bn this year. Six hundred and forty-four BILLION[*] dollars! gadget.co.za/genaispend644bn14

[*] Okay, they probably mean the "thousand-million" definition, not the "million-million" one. Still, that's bad enough. :P

Gadget · Worldwide GenAI spend to hit $644bn this year – GadgetGartner warns that CIOs must prepare for rising GenAI spend, fuelled by better foundational models and growing demand for AI products.
#AI
Replied in thread
Public *
AJ Sadauskas

@skribe Conversely, the cost of printing, distribution, and storage puts up a barrier to spamming people on other continents with mass quantities of low value slop.

Just think through the logistics of a hostile Eurasian state sending a mass quantity of printed materials to Australia or North America.

Or, for that matter, a hostile North American state sending a mass quantity of printed materials to Europe or Asia.

You would either need:–

a) At least one printing press on each continent;
b) You could try shipping the magazines, but they'd be a month out of date when they arrive; or
c) You could try flying them overseas, but that would be very expensive very quickly.

That's before you worry about things like delivery drivers (or postage), and warehouses.

These are less of an issue for books than they are for newspapers or magazines.

And if a particular newspaper or magazine is known to be reliable, written by humans, researched offline, and the articles are not available online, then there's potentially value in people buying a physical copy.

#ChatGPT#LLM#LargeLanguageModel
Public
Root Moose

At some point you are going to get tired with fawking around with the Google, Apple, Microsoft, OpenAI, etc. of the world and want to control your compute services on your own terms. Why do you have that NVIDIA / AMD gpu (or two) at your site anyway?

None of the 'entrenched' mega-corps have any services or technologies that are special enough to keep you from your goals using open source tools. Machine learning tools are useful; GenAI is BS with a rarefied use case more akin to a circus stunt and plagiarism as a service than real, repeatable results you can incorporate into a solution and/or support pipeline.

- Deep thoughts with the Root Moose

#Google#Apple#Microsoft
ExploreLive feeds
About