Washington (AFP) – Climate change deniers are pushing an AI-generated paper questioning human-induced warming, leading experts

cross-posted from: https://lemmy.world/post/27865986

my go-to llms in no specific order –

1. Llama 4 behemoth
2. Grok 3
3. Claude 3.7 sonnet (extended thinking)

What's your go-to?

When in a hurry, I just use Gemini by telling my phone "Hey Google."

Interesting discussion about using AI in science generally and Google AI co-scientist in particular.

cross-posted from: https://lemmy.ml/post/28125709

A Microsoft employee disrupted the company’s 50th anniversary event to protest its use of AI.

“Shame on you,” said Microsoft employee Ibtihal Aboussad, speaking directly to Microsoft AI CEO Mustafa Suleyman. “You are a war profiteer. Stop using AI for genocide. Stop using AI for genocide in our region. You have blood on your hands. All of Microsoft has blood on its hands. How dare you all celebrate when Microsoft is killing children. Shame on you all.”

Sources at Microsoft tell The Verge that shortly after Aboussad was ushered out of Microsoft’s event, she sent an email to a number of email distribution lists that contain hundreds or thousands of Microsoft employees. Here is Aboussad’s email in full

cross-posted from: https://lemmy.ml/post/28035952

Israel’s use of artificial intelligence (AI) in its ongoing assault on the Gaza Strip – aided by tech giants such as Google, Microsoft, and Amazon – is fueling concerns over the normalization of mass civilian casualties and raising serious questions about the complicity of these firms in potential war crimes, according to a leading AI expert.

Multiple reports have confirmed that Israel has deployed AI models such as Lavender, Gospel, and Where’s Daddy? to conduct mass surveillance, identify targets, and direct strikes against tens of thousands of individuals in Gaza – often in their own homes – all with minimal human oversight.

Rights groups and experts say these systems have played a critical role in Israel’s incessant and apparently indiscriminate attacks, which have laid to waste massive swaths of the besieged enclave and killed more than 50,000 Palestinians, mostly women and children.

cross-posted from: https://lemmy.world/post/27657674

People on twitter at each other’s throats for a recent trend in which you can use imageai filters to reproduce images in Studio Ghibli style.

Lots of people are throwing around the Miyazaki quote where he calls AI an “insult to life itself” - it’s worth noting that this came after he witnessed a simulation using AI animation to make a zombie body crawl across the ground. Miyazaki remarked that he has a very disabled friend he sees regularly, and that he can't find this usage cool or interesting. He goes on to remark about generative AI later in the clip, but let’s be clear, that specific statement was disconnected.

I’m curious what the Lemmy opinion is here. Is this a short-lived fad or does it speak to larger ramifications in the creative industry? Is this a silly little image filter or a harbinger of global artistic doom?

In case that were to be fun to post somewhere? Is there a better place to ponder such things?

BigEvilCo™ has so much moni 💰, so many marketing employees 🧑‍🤝‍🧑 and so many smart engenieers... how can they not create useful AI functionalities, which is not just a chatbot?

currently LLMs are mostly known for runing the internet and not being able to count letters 🍓 and embedding models are rarely used for anything but RAG

like -.,.-.. LLMs and embedding models have so many potential applications: 📑

• user asks for setting, LLM and embedding model go through all the settings and show you the settings you want ⚙
• give a VLM an image u like and ask "heyo customize the theme to fit those vibes" and it should filter the colors and assign them to the interface theme 🌈
• better search functionality with embedding models! combine that with keyword search and you get a robust but flexible search, which finds what you mean and type📰
• typing improvement hints like in microsofts word. just put a lil green squiggle. right click and see what an LLM would suggest to write otherwise. Could be great for quick brainstorming and then formalizing ✅
• just in general - interfaces could be improved so much with smol LLMs like the 1.5B Qwen models with smol helps like this 🦆

i believe showing peeps how LLMs and embedding models can be truly useful in quick situations like this is a good idea. it shows that they are not just chatbots, but can actually, for real this time speed up what you want to do.

adding new features into existing interfaces makes them easy to use. adding a whole new panel specifically for your new chatbot just takes away space, needed for lists and controls

I hate most current AI features

i hate dum ai popups whenever i see one. but i like, when i right click on something, to be presented with "copy", "select all" aswell as "summarize".

right now, it's hard not to hate AI in general without looking at medical applications. 👩‍⚕️

Let's change this!

• disable googles "AI overview" by default, and run that with a dedicated button for it. then show EXACTLY where the info comes from right in the response (also saves them money btw) 💸
• remove the "generate responses" option in youtube comments and move that into the hamburger menu (takes up less space) 🟥
• add "order by importance" button to search results using embedding models and semantic-distance-calculations 🔎

let's stop making AI this "fancy new thing" and more of "just another tool" ✨
let's make it compact and specialized instead of bulky and expensive 💸
let's get AI out of peoples view but into peoples lives 🌍
let's make people actually want to use AI tools and not force it onto their screen 🖥️
let's start small and build up from there
let's stop creating agents, while LLMs still hallucinate new tools they don't have access to 🧠

and for the love of god

Let's mark AI generated content

~(this~ ~text~ ~was~ ~human-written~ ~btw.~ ~it~ ~might~ ~kinda~ ~look~ ~generated,~ ~but~ ~noooooo~ ~im~ ~a~ ~real~ ~human,~ ~trust~ ~me)~

publication croisée depuis : https://lemmy.pierre-couy.fr/post/1059609

Hey everyone!

I've been working on my own toy reinforcement learning (RL) framework for a while now and have nearly implemented a full Rainbow agent—though I'm still missing the distributional component due to some design choices that make integration tricky. Along the way, I’ve used this framework to experiment with various concepts, mainly reward normalization strategies and exploration policies.

I started by training the agent on simpler games like Snake, but things got really interesting when I moved on to Super Mario Bros. Watching the agent learn and improve has been incredibly fun, so I figured—why not share the experience? That’s why I’m streaming the learning process live!

Right now, the stream is fairly simple, but I plan to enhance it with overlays showing key details about the training run—such as hyperparameters, training steps/episodes, performance graphs, and maybe even a way to visualize the agent’s actions in real-time.

If you have any ideas on how to make the stream more engaging, or if you're curious about the implementation, feel free to ask!

I had a really good conversation with Dean Pleban (CEO @ DAGsHub), who shared some great practical insights based on his own experience helping teams go from experiments to real-world production.

Some practical tips Dean shared with me:

• Save chain of thought output (the output text in reasoning models) - you never know when you might need it. This sometimes require using the verbos parameter.

• Log experiments thoroughly (parameters, hyper-parameters, models used, data-versioning…).

• Start with a Jupyter notebook, but move to production-grade tooling (all tools mentioned in the guide bellow 👇🏻)

To help myself (and hopefully others) visualize and internalize these lessons, I created an interactive guide that breaks down how successful ML/LLM projects are structured. If you’re curious, you can explore it here:

https://www.readyforagents.com/resources/llm-projects-structure

Skyvern is an open source (AGPL) AI agent that can complete tasks on the web.

Here's an example of Skyvern being asked to "Compare iPhone Pro Models": https://eval.skyvern.com/wpid_348574331532282024/wr_348574331532282026/overview

It can:

• Search the modern web
• Log in to websites, bypass captchas
• Fetch the latest information
• Handle error messages or new layouts

I've tried it out for a few things, but imo it's prohibitively expensive; that said, it's quite cool for doing things that scrapers cannot (one of the main use cases right now is it can apply to jobs for you). I've been following them for a few months, but I'm excited to see where they are in a few years

cross-posted from: https://lemmygrad.ml/post/7137748

And yet, China is using AI.

...I... don't know what to think about that.

...I really don't.

Because it seems that AI is just a scam.

It may "exist" but what it can do is a scam.

Maybe China thinks we have to use it just to "keep up" with the Western powers, but I dunno.

Anyway, interesting discussion with Adam Conover and Ed Zitron. It's long, but you can listen to it while doing other things. And the comments are interesting too, but then again, there are also trolls in the comments as well (AI supporters here and there).

Frankly, though? I oppose AI. I'm anti-AI. I'm anti-AI in China and anti-AI in America and anti-AI in the whole damn planet.