Search

Items tagged with: ai




Interesting article on using process reward learning rather than final outcome for #AI supervision. So instead of asking an AI a question and doing reward learning on the totality of the result, each step is evaluated. Good results for mathematics: openai.com/research/improving-…
#AI


What happens when you cross my crazy wak ass blind friend group, with state of de art AI technology, with my rediculous, embarrassing laugh, with a video about bad translations?

You get this: youtu.be/IpLWg9U4uRE?si=uhHKIV…

Fair warning, this is a lightly edited Discord voice chat session, and I apparently had my gain set waaaaaaay to high, so expect clipping and general audio discomfort.

#AI #TheLowerElements #ElevenLabs #Translation #Fail




OpenAI si vsimlo o jake niche je zajem v tisicich AI startupech a spustilo vlastni "GPT builder".

Je to celkem dobra evoluce v pouzitelnosti. Dost lidi si najde vlastni pouziti ChatGPT a to se pro ne stane synonymem pro AI. Napriklad doplnovani kodu, analyza tabulek, zkracovani textu, copy na socky... Pojmenovane chaty s jakoby ruznou funkci jim tuhle kotvu pomuzou zvednout.

Tech vsech AI startupu mi vubec neni lito.

*obrazek vygenerovan random z Cosmic Dream GPT modulu.

#ai #chatgpt


Early returns are promising. Students got into the activity. Discussion took over half the last day of class.

This is a Programming Languages class, and discussion took an interesting turn: how would •languages• change in a world where we expect AI code generation to be a normal part of the development process, and the problem for humans shifts (even further) away from “How do I generate code?” and toward “Does the code I’ve generated actually do what I think? What I want?”
#proglang #ai


OK, trying an experiment with my Programming Languages class!

• Have an AI generate some of your writing assignment.
• Critique its output. Call BS on its BS.

Assignment details in screenshots below. I’ll let you know how it goes.

(Here are the links from the screenshots:)

Raw AI Text:
gist.github.com/pcantrell/7b68…

Comments on AI Text:
gist.github.com/pcantrell/d51b…

(Better) Human Text
gist.github.com/pcantrell/f363…

#ai #chatgpt #education #writing #highered
#swift #proglang


My favorite outcome so far: a student remarked (paraphrasing here) that she didn’t realize how much she had to say in her paper until she saw how wrong the AI was, how much it missed the point.

Observing her own reaction to BS about her topic made her realize she’d underestimated the extent of her own newly-forming knowledge. That…that is the sort of outcome an educator dreams of.

#ai #chatgpt #education #writing #highered




New blog post: Enhancing Accessibility with AI and ML

We discuss how to specialize #AI for understanding UI for #a11y testing, the importance of algorithm and data integrity, and how #ML can be used to simplify #a11y testing. deque.com/blog/enhancing-acces…

#a11y #AI #ML



#AI #llm


A really nifty deep dive on the current state of #AI #VoiceCloning #music #CoverSongs, via getting the artists #WeirdAlYankovic parodied to ‘cover’ his songs

It mainly shows how distinctive Weird Al’s voice is, though MechaLorde singing ‘Foil’ is pretty good

AI has the potential to disrupt the whole comedy genre of pop-cultural pastiche and parody. Why would you watch comedians doing impressions of celebrities, or riffing on pop-culture texts, when you could just manipulate the celebrity yourself using AI?

Still, what constantly annoys me about the state of AI-driven parody (eg the meme of pop-culture characters as #Balenciaga models) is that the writing underpinning them is so poor

So even if we get to a point where we can completely puppeteer real people’s likenesses and voices, it still fails if there isn’t decent writing

Weird Al’s writing has always been excellent and hilarious. I mean, lines like in ‘Fat’, his parody of ‘Bad’: “When I sit around the house, I really sit around the house”

So far, that kind of stuff can’t be automatically generated – AI-created comedy is of the surreal and absurdist kind that reads as an unintentional, “machine-innocent” artefact of the generation process

waxy.org/2023/10/weird-ai-yank…


New article posted by Hasan Çimen: For us visually impaired individuals, accessing image descriptions has long been a challenge. While object recognition apps provided some assistance, they were limited in their ability to describe images comprehensively. However, recent developments in AI have brought about a groundbreaking solution, making detailed image descriptions accessible to the community. Let’s look at the brief history accessibleandroid.com/detailed… #Android #AI #GPT4 #Bing


#AI



Taken from #Google's updated "privacy" policy.
I mean, I find it ironic because there doesn't seem to be any #privacy to speak of here 😆

I know this is no laughing matter, but I am at the "oh what the hell can I do so might as well laugh about it" stage 🤣

#WritingCommunity #Data #AI


An update on #BlueSky and #AI. They have answered to questions and revealed that their deal with hive.ai specifically forbids the latter from training their models on user data. BlueSky have also updated their TOS to be more clear.

See the linked toot from ‪@growlbeast‬ for more details.

I’d still avoid it personally, their “moderation” is nonexistent and I wouldn’t touch anything backed by cryptobros, but for now this is good news :)

mastodon.art/@growlbeast/11065…


Bsky does have say in how Hive uses the data Hive gets from Bsky as it's part of the agreement they have with Hive and It would have to be disclosed in our user end ToS.(The ToS states our data is NOT used beyond moderation models, meaning it does not go to the image gen model.)

BSky has been updating their ToS to be more clear about it apparently too.

here's an email sent from the team to my friend regarding some of this:



To anyone thinking about joining BlueSky, especially artists: everything you post is used to train generative AI models.

BlueSky uses AI to label content for moderation, and to do that they use a company called thehive.ai. If you look through their privacy policy, you will see that they use all content sent to them to train models for all their services, which include generative AI for both text and images.

Update: meow.social/@FluffyDeveloper/1…

#ai #bluesky


An update on #BlueSky and #AI. They have answered to questions and revealed that their deal with hive.ai specifically forbids the latter from training their models on user data. BlueSky have also updated their TOS to be more clear.

See the linked toot from ‪@growlbeast‬ for more details.

I’d still avoid it personally, their “moderation” is nonexistent and I wouldn’t touch anything backed by cryptobros, but for now this is good news :)

mastodon.art/@growlbeast/11065…




“And yet falsely claiming to be a human through online interaction could/should be seen as a manifestation of false and misleading advertising under the Competition Act. Chatbots should be clearly labelled, too. They should not be able to masquerade as humans.

Synthetic media is super similar. If we are to consider consuming stuff conjured by artificial intelligence, we should be able to know what made it.” A+ piece from Vass Bendar: regs2riches.com/p/what-tech-po… #AI #techpolicy #GCdigital


We are seeking proposals for our #DeepDiveAI webinar series. OSI board members and volunteers will review the proposals and those accepted will be asked to present their talk online in a webinar during the months of September through November

Submit your proposal today! Deadline for submissions is July 24th. buff.ly/42uGUVS

#AI #OSI #DDAI#OpenSource #ArtificialIntelligence #OpenSourceAI


💾 Trying my hand at some #Agitprop. How does this one look like?

🍝 Mi cimento in poster agitprop, come vi sembra?

#AI #Billionaires




@libreoffice

I was wondering if the #libreoffice team had some projects about #AI integration.

MS Office is working hard on this topic.

-- fr --
Je me demandais si l'équipe de #libreoffice avait des projets d'intrégation de l' #ia.

MS Office avance bien sur ce sujet.

#opensource
#calc
#writer
#Impress

libreoffice.org/




Over the past year, I've been experimenting with neural text to speech in various forms. I have done hours of experimentation and research, training models and getting varying results along the way. Some of you may have heard of Piper, an open source synthesizer and add on for NVDA that can be trained by anyone. It is currently in active development, and I have been there from the beginning, testing and evaluating the various versions. For years, I have had a goal to create a high-quality voice that is truly usable by a screen reader user, and yesterday I managed to achieve this. I'm really excited to share Alba, a female Scottish English voice. I'm considering this a beta phase, and I'm looking for feedback to make improvements as needed. Please note that you will most likely get an error upon installation, however the voice should still show up to NVDA, and I'm working on fixing this as soon as possible.
Link to Piper: github.com/rhasspy/piper/tree/…
Link to addon: github.com/mush42/piper-nvda?r…
Link to Alba: drive.google.com/file/d/1wZHuI… #TTS #AI #ScreenReader #Piper


“AI threatens to deepen the dominance of a way of thinking that is white, male, comparatively affluent and focused on the US and Europe.”
"An industry brimming with people who espouse liberal, self-consciously progressive opinions so often seems to push the world in the opposite direction."

John Harris interview with @timnitGebru, the Ethiopian-born computer scientist who sacked by Google after pointing out the inequalities built into #AI.
theguardian.com/lifeandstyle/2…


"Don’t Tell #ChatGPT Anything You Wouldn’t Want to See on a #Billboard"

"Here’s a simple tip: DO NOT use #AI on any work-related project without checking your company’s policy. Even if your company has no policy, think twice, or even three times, before you put anything work-related into an AI.
You don’t want to become infamous for triggering the privacy fiasco that spurs your company into creating such a policy."

uk.pcmag.com/ai/146847/dont-te…