Items tagged with: AI

Search

Items tagged with: AI

In 3 days, the jabber.fr wiki, a niche specialized wiki with a lot of outdated info and only one or two occasional contributors at the moment, has served more than 200 000 pages to shitty LLM bots.

Our PHP and mariadb processes are constantly hammered by them, and most of them are masquerading as normal traffic with browser user agents.

Is there a better solution than blocking off entire countries and cloud provider subnets at the moment?

#xmpp #ai #llm #jabber

#xmpp #Jabber #AI #llm

#Accessibility #apple #blind #technology #discord #AI #tech #code #community #techtalk #techopolis

So I seem to be on a bit of an #AI testing kick. A non-critical product I needed (some replacement filters for an air purifier) was available at the best price from a website I really didn't want to deal with right now. So I put the order in via innosearch.ai. I'm mostly suspicious because the marketing is so polished; I've come to distrust well-marketed products. On the other hand, if innosearch makes it possible for me to spend less money on Amazon, and makes distributing my shopping dollars to other places more practical, it's a net ethical win. If I get entirely the wrong thing, or nothing at all, it won't break me. I'll keep this thread updated with my experiences. #innosearch#accessibility#blind
#Accessibility #blind #AI #innosearch

Anthropic used circuit tracing to reveal more internal processes of Claude. It thinks in a conceptual space shared across languages like universal language of thought. It can plan responses several words in advance when writing rhyming poetry. It computes both rough estimations and precise calculations in parallel when solving math problems. It can combine independent facts to arrive at an answer, instead of simply recalling a memorized response. #LLM #AI #ML anthropic.com/research/tracing…

Tracing the thoughts of a large language model

Anthropic's latest interpretability research: a new microscope to understand Claude's internal mechanisms
www.anthropic.com
#AI #ML #llm

#OpenAI: "We’re planning to release our first open language model since GPT‑2 in the coming months. We’re excited to collaborate with developers, researchers, and the broader community to gather inputs and make this model as useful as possible. If you’re interested in joining a feedback session with the OpenAI team, please let us know below." #LLM #AI #ML openai.com/open-model-feedback…
#AI #ML #llm #openai

Tired: how are you using #AI in the classroom?

Wired: how are you pretending that capitulating to the technofascist billionaires who for decades have wanted to vacuum up the money devoted to public education is actually a good thing for students?

#AI

This post by photographer Jingna Zhang resonates with me.

#AI art is devoid of humanity, intention, and backstory. It’s a shallow remix of human works, designed to be consumed and discarded, grinding human creation down into a sandy paste, to be re-extruded into grotesque displays. It’s the opposite of what makes art so valuable.

#aiArt #art #plagiarism #artist

(alt text continues)

A Facebook post by Jingna Zhang. Shown is a photograph of a pale woman with closed eyes and dark red lips covered in flower petals. Next to it is a prompt: “Make her eyes open. Make the image very similar, from the the colors, composition, the tone, the feel, including the frozen eyebrows, make it as similar as possible.”, followed by an AI-generated facsimile that is similar, but with eyes open. The post reads: Someone passed my photo through OpenAI's new image model, and no matter their intentions being well-meaning, seeing this as the original artist just makes me feel like shit. I did this series when my work wasn't being as accepted and was viewed as neither commercial nor fine art enough for either worlds. So I shot in my apartment, with flowers one of the few props I could afford. And for this shoot I sat with a team and applied petals onto the model, one by one. The final work is a crystallization of that experience and memory of those times, it's proof of the days l'd lived. An Al didn't sit there placing petals onto my model, and it feels like an insult to the memory when people go "Al can do this!", like no actually, I did that.
#art #AI #aiart #artist #plagiarism

So an update on Guide: I've exchanged some long emails with Andrew, the lead developer. He's open to dialogue, and moving the project in the right direction: well-scoped single tasks, more granular controls and permissions, etc. He doesn't strike me as an #AI maximalist can and should do everything all the time kind of guy. He's also investigating deeper screen reader interaction, to let AI just do the things we can't do that it's best at. I stand by my thoughts that the project isn't yet ready for prime time. But as someone else in the thread said, I don't think it should be written off entirely as yet another "AI will save us from inaccessibility" hype train. There is, in fact, something here if it gets polished and scoped a bit more. #blind#screenreader#a11y
#a11y #blind #screenreader #AI

Sensitive content

#AI #copilot #chatgpt #excel #Claude

#a11y #AI #testing #wcag #automation

Final update: The developer is now on Mastodon via @andrew_guide.

Update: The developer has removed the ability to download Guide until the security issues mentioned in the linked thread are fixed.

Update: this product contains some code flaws that are concerning from a security perspective, beyond just giving control of your computer to an LLM. You might want to read this thread before installing the product: toot.cafe/@matt/114258349401221651

Update: I've exchanged some long emails with Andrew, the lead developer. He's open to dialogue, and moving the project in the right direction: well-scoped single tasks, more granular controls and permissions, etc. He doesn't strike me as an #AI maximalist can and should do everything all the time kind of guy. He's also investigating deeper screen reader interaction, to let AI just do the things we can't do that it's best at. I stand by my thoughts that the project isn't yet ready for prime time. But as someone else in the thread said, I don't think it should be written off entirely as yet another "AI will save us from inaccessibility" hype train. There is, in fact, something here if it gets polished and scoped a bit more.

Just tried guide for fun. It's supposed to be an app to use #AI to help #blind folks get things done. I asked "Where are the best liver and onions in Ottawa?" It:
1. Decided it needed to search the web.
2. Thought that the "stardew access" icon on my desktop was a kind of web browser, so clicked it.
3. Imagined an "accept cookies" dialogue it needed to accept.
4. Decided that didn't work, so looked for Google Chrome (I don't have chrome installed on that machine)
5. Finally opened edge from the start menu. By the way, it just...left Stardew open and running. Because apparently having Stardew Valley running in the background is a vital part of finding liver and onions in Ottawa.
6. Opened a random extension from my edge toolbar (goodlinks).
7. Clicked the address bar and loaded google.com, instead of just doing the search right from the address bar.
8. Got blocked because it couldn't sign into my Google account, even though it could have also searched from the Google homepage.

To be fair to AI, that was the kind of open-ended task AI is terrible at. If I had asked it to check an inaccessible checkbox, or read a screenshot, or something, I'm sure it would have been fine.

Anyway, I'm still better at using a computer than an AI. So is my 87 year old grandfather, for that matter. www.guideinteraction.com


Guide - AI assistant for people with low vision or blindness

With Guide, inaccessible doesn't have to mean undoable. Guide is a Windows AI assistant that helps people with low vision or blindness navigate the digital world.
Guide

Matt Campbell

2025-03-31 18:00:52

mastodon - Link to source

There's a new product that has been gaining some buzz in the blind community, a Windows app called Guide that uses AI to perform tasks on your computer. It's pitched as a way to get around web accessibility problems in particular. I won't link to the thing itself, because I don't want to give it that validation, but I'll link to a previous discussion thread about it: fed.interfree.ca/notes/a5wf4ys…

I've spent some time taking this app apart. The level of shoddy work here is deeply disgusting. 1/?


#blind #AI @Andrew Guide

Well, wasn't expecting to see this:

> Alibaba says Omni-7B can run efficiently on phones and laptops, enabling real-world applications like real-time audio descriptions for visually impaired users.

#AI

#AI

Wow. For a few months, I was wondering why I suddenly have bandwidth issues when activating my camera in MS Teams meetings, so others can't understand me any more.

A look into my #nginx logs seems to clarify. Bots are eagerly fetching my (partially pretty large) #poudriere build logs. 🧐 (#AI "watching shit scroll by"?)

I see GPTBot at least occassionally requests robots.txt, which I don't have so far. Other bots don't seem to be interested. Especially PetalBot is hammering my server. And there are others (bytedance, google, ...)

Now what? Robots.txt would actually *help* well-behaved bots here (I assume build logs aren't valuable for anything). The most pragmatic thing here would be to add some http basic auth in the reverse proxy for all poudriere stuff. It's currently only public because there's no reason to keep it private....

Have to admit I feel inclined to try one of the tarpitting/poisoning approaches, too. 😏

#AI #nginx #poudriere

😲 DeepSeek-V3-4bit runs at >20 tokens per second and <200W using MLX on an M3 Ultra with 512GB. This might be the best and most user-friendly way to run DeepSeek-V3 on consumer hardware, possibly the most affordable too. You can finally run a GPT-4o level model locally, with possibly even better quality. #LLM #AI #ML #DeepSeek #OpenAI #GPT #OpenWeight #OpenSource venturebeat.com/ai/deepseek-v3…

DeepSeek-V3 now runs at 20 tokens per second on Mac Studio, and that’s a nightmare for OpenAI

DeepSeek's free 685B-parameter AI model runs at 20 tokens/second on Apple's Mac Studio, outperforming Claude Sonnet while using just 200 watts, challenging OpenAI's cloud-dependent business model.
Michael Nuñez (VentureBeat)
#opensource #AI #ML #llm #openai #gpt #deepseek #openweight

💰 Spain is done with AI slop and disinfo

｢ The Spanish bill, which needs to be approved by the lower house, classifies non-compliance with proper labelling of AI-generated content as a "serious offence" that can lead to fines of up to 35 million euros ($38.2 million) or 7% of their global annual turnover ｣

reuters.com/technology/artific…

#spain #ai #deepfakes #disinformation

#AI #disinformation #spain #deepfakes

#Mastodon #AI #phone #unix_surrealism #cable

AirBnB migrated 3,5k React component test files from Enzyme to RTL using LLMs and automation in just six weeks. The effort was originally estimated to take 1.5 years of manual engineering time to do by hand. #LLM #AI #ML #Automation
medium.com/airbnb-engineering/…

Accelerating Large-Scale Test Migration with LLMs - The Airbnb Tech Blog - Medium

Airbnb recently completed our first large-scale, LLM-driven code migration, updating nearly 3.5K React component test files from Enzyme to use React Testing Library (RTL) instead. We’d originally…
Charles Covey-Brandt (The Airbnb Tech Blog)
#AI #ML #automation #llm

#AI #9front #unix_surrealism #glenda #mnt

We're going LIVE in just 15 minutes—come hang out and catch up on the biggest AI news you need to know!
Whether you're deep into AI or just curious what's going on, this stream breaks it down in a way that makes sense. Real talk, real insights, and a chance to ask questions live.
See you there!
#AI #TechTalk #LiveStream #AICommunity #ArtificialIntelligence
youtube.com/watch?v=nR85Kyborq…

AI News For March 21

Turn your videos into live streams with https://restream.ioTechopolis Website: https://techopolis.onlineBook Services: https://calendly.com/tayarndtJoin Our ...
YouTube
#AI #livestream #artificialintelligence #techtalk #aicommunity

#AI #writing #meta

#CNA A look at the robots at Nvidia’s conference

A look at the robots at Nvidia’s conference We took a look at the robots featured at Nvidia's developer conference in San Jose, California on Tuesday (Mar 18). #ai #robotics #robots #nvidia #nvda #news from CNA Video YouTube untuk artikel ini boleh ditonton di sini 💛

shibuyaworldnews.online/cna-a-…


#CNA A look at the robots at Nvidia’s conference - SHIBUYA World News

https://www.youtube.com/watch?v=O_xW8hub2x0 A look at the robots at Nvidia’s conference We took a look at the robots featured at Nvidia's developer conference in San Jose, California on Tuesday (Mar 18).
🐈 (SHIBUYA World News)
#News #AI #nvidia #nvda #CNA #robotics #robots

About time roflmeow

• AI comes to Narrator. Rich image descriptions powered by AI will be coming to Narrator in Windows Insider Preview this spring

blogs.microsoft.com/blog/2025/…

#accessibility #microsoft #ai


Microsoft Ability Summit 2025: Accessibility in the AI era - The Official Microsoft Blog

Today, we hosted the 15th annual Microsoft Ability Summit, with over 20,000 attendees from 164 countries coming together virtually to discuss the future of AI and accessibility.
Jenny Lay-Flurrie (The Official Microsoft Blog)
#Accessibility #microsoft #AI

😲 It’s only March, but there’s already been incredible progress in open-weight LLMs this year. Here are my top 5 local LLM recommendations for anyone with 24GB of VRAM to try: Phi-4-14B for speed, Mistral-Small-24B for RAG, Gemma-3-27B for general use, Qwen2.5-Coder-32B for coding, QWQ-32B for reasoning. #LLM #ML #AI
#AI #ML #llm

Now there's MLX-Audio! They have Kokoro and CSM-1B for now. It's the same dev team for MLX-VLM, so I have high hope! #TTS #AI #ML github.com/Blaizzy/mlx-audio

GitHub - Blaizzy/mlx-audio: A text-to-speech (TTS) and Speech-to-Speech (STS) library built on Apple's MLX framework, providing efficient speech synthesis on Apple Silicon.

A text-to-speech (TTS) and Speech-to-Speech (STS) library built on Apple's MLX framework, providing efficient speech synthesis on Apple Silicon. - Blaizzy/mlx-audio
GitHub
#AI #ML #tts

Here's another fun project! AIChat lets you generate conversations between two LLMs on any topic using OpenAI API. You can mix and match models from Ollama, MLX, Claude, OpenAI, Google AI Studio, etc. It uses Kokoro-ONNX for TTS. Check out an example + the GitHub repo:
youtube.com/watch?v=FgSZLZnYlA…
github.com/chigkim/AIChat
#LLM #TTS #AI #ML

GitHub - chigkim/AIChat: Generate a conversation between two models using OpenAI API and Kokoro TTS.

Generate a conversation between two models using OpenAI API and Kokoro TTS. - chigkim/AIChat
GitHub
#AI #ML #llm #tts

Using an #AI email writer might seem like the perfect way to boost productivity, BUT the dangers far outweigh the benefits. 🤖 ❌

In this guide we take a look at 5 AI email writers and why not to use them 👉 tuta.com/blog/ai-email-writers…

#AI #AIEmailWriter #PrivacyMatters #TutaMail

5 reasons AI is bad for email. Allowing AI access to your mailbox threatens your email privacy. AI collects a lot of your data. Your data is used in many ways: from training LLMs, to advertising .AI negatively affects the environment. AI is not open source.

5 AI email writers and which not to use | Review | Tuta

AI email tools can help write your emails quicker, but the risks of granting them access to your private mailbox often outweigh the benefits. Here’s why not to use AI email writers.
Tuta
#AI #privacymatters #tutamail #aiemailwriter

Exciting news! Microsoft just launched a brand-new AI course: "Create Accessible Solutions using AI Innovations." Learn how to harness AI to build more inclusive technology. Check it out now! 🚀 #Accessibility #AI
Explore the course here: learn.microsoft.com/en-us/trai…

Create accessible AI experiences - Training

This module provides guidance on how to collaborate effectively with customers, partners, and the disability community to create innovative solutions that are accessible and inclusive.
learn.microsoft.com
#Accessibility #AI

My Ability Summit 2025 recap stream is just 30 minutes away.
Join me at 5 PM Central Time as I break down the biggest announcements, AI advancements, and accessibility innovations from this year's event.
When: Today at 5 PM CT
youtube.com/watch?v=Pw6a6aR-sz…
Don't miss it. See you soon. #AbilitySummit #Accessibility #AI

Microsoft Ability Summit 2025: Top Highlights & Key Takeaways

Turn your videos into live streams with https://restream.ioThe Microsoft Ability Summit 2025 just wrapped up, and I’m bringing you the biggest announcements,...
YouTube
#Accessibility #AI #abilitysummit

Anthropic’s Dario Amodei predicts that #AI could be writing 90% of code in the next 3-6 months & potentially all of it within a year.

While human programmers will remain essential for now, AI may soon replace them across industries.

How should we prepare for the impact on the workforce?

#AI

Sensitive content

#AI #writing #llm #story

#AI #learntocode #techskills #techtraining

There’s something very depressingly emblematic that Chrome Dev Tools’ annoying “Ask AI” sprinkles everywhere include one that pops up and *covers* the Accessibility Tree toggle.

#google #chrome #ai #accessibility

A close up of Chrome Dev Tools showing a button with an “Ask AI” title displayed over the top of a round button, which is for displaying the accessibility tree.
#Accessibility #google #chrome #AI

Sensitive content

#AI #nodejs

I’m quite enjoying the new Ally app from Envision. Being able to customise the AI is interesting, although the iOS app doesn’t seem to have all that many voices yet.
It’s very good at reading digital displays which other apps sometimes struggle with, but it’s not perfect. I was using it to check shopping items and it misidentified a couple of things. Still, it’s not bad and reasonably fast, although not as good as the Meta Rayban’s Look and Tell feature. Still since Meta have withdrawn that in the UK, Ally is a decent enough alternative.
#Accessibility #AI
#Accessibility #AI

Gemma3 is out with 1b, 4b, 12b, 27b. It also accepts images except 1b model. It's on Ollama, but ollama run gemma3 will use small 4b model. Be sure to specify size if you want to use bigger model. For example, ollama run gemma3:27b #LLM #ML #AI ollama.com/library/gemma3/tags

Tags · gemma3

The current strongest model that fits on a single GPU.
Ollama
#AI #ML #llm

#privacy #signal #AI #tech #signalapp #sxsw @Meredith Whittaker

Very interesting read from HuggingFace co-founder: For big scientific breakthroughs, "we don't need an A+ student who can answer every question with general knowledge. We need a B student who sees and questions what everyone else missed." Current AI is great at answering known questions, but struggles to challenge assumptions and push boundaries. True innovation comes from questioning and redefining knowledge—something AI isn’t quite ready for yet. #AI #LLM #ML thomwolf.io/blog/scientific-ai…

🔭 The Einstein AI model

I shared a controversial take the other day at an event and I decided to write it down in a longer format: I’m afraid AI won't give us a compressed 21st century.
thomwolf.io
#AI #ML #llm

#AI #generativeAI

#github #AI #copilot #llm #genai

⇧