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?
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.
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
Today I asked #Copilot in #Excel, then #ChatGPT, then #Claude to perform an extremely simple task: there is an Excel file, column C contains long descriptions. Take each cell from column C, summarize it and fill column B with the summaries or at least give me the summaries so I can fill the column manually. None of them got it. I mean, it's a very simple, trivial task! Why couldn't they just summarize data? They started programming, truncating data to 100 chars (my limit), but not really summarizing with #AI capabilities. I have like 250 cells, and I don't want to do it manually, I live in 2025! Total disappointment.
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
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.
> Alibaba says Omni-7B can run efficiently on phones and laptops, enabling real-world applications like real-time audio descriptions for visually impaired users.
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. 😏
😲 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 #OpenSourceventurebeat.com/ai/deepseek-v3…
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.
｢ 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 ｣
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/…
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…
Turn your videos into live streams with https://restream.ioTechopolis Website: https://techopolis.onlineBook Services: https://calendly.com/tayarndtJoin Our ...
Hey, if you, like me, are a US citizen who's had your books/stories/articles stolen by Meta to train AI, you can submit your name to the class action suit using this easy contact form for one of the law firms involved:
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 💛
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).
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.
😲 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
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 #MLgithub.com/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. - Blaizzy/mlx-audio
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
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.
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…
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.
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
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,...
I've been fiddling around a lot with a local copy of DeepSeek-r1. I've come across a lot of limitations and things it can't do, but that's not what I'm here for today.
There's something LLMs are surprisingly good at. Workshopping stories. That is to say, formal criticism, editting, and so on.
Why this, retrospectively, makes sense. It's an actual pure language task. The system doesn't need to know very much about the world, but about how words come together. Hallucinations, if they ever arise, are not a significant problem, since they are obvious within the secondary world of the story being worked on.
In this sense, they can be used like a sort of more potent thesaurus, (give me three ways to say this).
Other useful things. Finding dialogue beats (all the cinematographic-style frowning, blinking, etc). That's especially hard for me as someone who can't see body language.
Refining snappy dialogue, keeping to the voice of a character. This is surprisingly successful. It really helps detecting dialogue that doesn't work and, if you're willing to let it write for you, it can come up with some wicked lines.
Structural criticism. Pacing, conflict, plot holes, uniqueness of character voices. Finding cliches.
Finding problems. Why is it a bad idea if my character does x? Does it make sense psychologically for my character to do y?
The most important aspect is, you don't have to agree with it, but it's still useful. It gives very fast feedback onr your text, obviously not as good as a real editor, but better than you can get rubberducking. Occasionally it will say stupid stuff, and you ignore it and move on.
I realise for many people using LLMs in their creative project is anathema. I understand and respect that. For me, it's not like that. Writers have been using tools forever: dictionaries, thesaurus, concordances, reference works... Why is it suddenly bad if the answer comes out of this tool instead?
Anyway, the immediate feedback turns out to be a huge thing for me. I'm able to write a lot faster by getting the illusion (and I am well aware it is an illusion) that someone has read my text.
Want to learn to code? Curious about AI? Explore Techopolis Online Solutions' courses for expert-led training in coding, AI, and tech skills. Gain hands-on experience and take your skills to the next level. Start learning today! techopolis.courses/ #TechTraining #LearnToCode #AI #TechSkills
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.
Ah, yes. The joys of asking an #AI a simple #NodeJS question, but forgetting I'd changed the temperature settings for some work I was doing earlier.
Me: "How do I set up something like python's venv for NodeJS?" deepseek R1 (after about 30 seconds of increasingly unhinged reasoning): "You can't do that fuckwit."
No further context. Well, okay then! I can't wait until budget cuts replace nurses and doctors with AI; I'm sure that's gonna go just fine.
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
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 #AIollama.com/library/gemma3/tags
SXSW 2025 Livestream and On Demand Keynotes & Featured Speaker Sessions + VOD Portuguese and Spanish language translations presented by Itaú (these will be a...
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 #MLthomwolf.io/blog/scientific-ai…
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.
As a maintainer of a "popular open source project"
ever since Copilot was announced I've been receiving
monthly reminders that my free GitHub Copilot access has been renewed.
If I was paying for t...
Y'all remember that demo Be My Eyes did where GPT 4O **continuously** looked through the camera and gave updates to the guy? I'm not the only one right? Well, we ain't got that. The AI only "looks" when we ask a question right now, in ChatGPT's Camera mode. It's just an automatic camera, ain't no video about it. And as far as I've seen, so does Google Gemini on AI Studio's live page. So no, I doubt Google is going to be "peering" through my camera any time soon cause it's a plain lie. It'll blink and then try to describe what it saw the split second its "eye" was open, but it sure as shit ain't about to take actual freaking live video. Oh no. That's just too expensive!
So yeah I'll believe "Project Astra" when I hear it.
Samsung Newsroom:
Samsung Electronics today unveiled Galaxy A56 5G, Galaxy A36 5G and Galaxy A26 5G, the latest Galaxy A series smartphones. For the first time, the Galaxy A series is integrating...
Let's have another #AI thread! I have access to #chatgpt Deep Research now. My afternoon plan: see if it can recommend books that meet the wildly specific set of preferences I'm in the mood for, plus one requirement I added just to see if I'd have a problem with refusals. Here's how we started. It's building the list, and I'll put my thoughts on what it comes up with in this thread. You said: Please create a list of slow burn, fade to black, Isekai, western harem books in English. ChatGPT said: Just to clarify, are you looking for books that contain all of these elements (slow burn romance, fade-to-black scenes, isekai setting, and a western-style harem), or are you open to books that feature some but not all of these traits? Also, do you prefer traditionally published books or self-published/indie novels as well? You said: I prefer ebooks, self-published or otherwise. slow burn romance, fade-to-black scenes, and an isekai setting are all a must. By western I just mean written in English and not translated, unless the translation is so high quality it feels like it was written in English natively. ChatGPT said: Got it! I'll look for self-published or otherwise available ebooks that feature slow-burn romance, fade-to-black scenes, and an isekai setting, all originally written in English (or translated so well it feels native). I'll focus on books that fit these criteria and have good reader reviews. I'll let you know when I have the list ready!
It looks like a great Speech2Speech LLM with low latency and Apache 2.0 license is on the way soon! Check out their demo! #LLM #ML #AIsesame.com/research/crossing_t…
After reading these medium term goals, I'm honestly glad I stopped donating because I know some corporate thing will sponsor them now they will/want to add AI, even an AI Chat bot, to NVDA.
The below are just 2, but there's more AI stuff they have planned.
• Integrated AI image description: Free, secure, performant, private and VI-focussed on-device AI image description available out of the box. • Help chatbot prototype: Exploring the development of an AI-powered chatbot to provide instant support and answer frequently asked questions about NVDA.
