Artificial Intelligence

Hello there

I am starting to understand and discover local (like on my computer) LLM.

LMstudio make it easy for beginners.

There is something I can’t make works right : my laptop has a Nvidia T600 2Gb graphic card. Genuine Nvidia drivers works well on the OS and graphic applications (Linux Mint). But LMstudo can’t use it.

LLM models run fine but only on CPU.

I read somewhere about CUDA not being fully implemented on Linux, is it the limiting factor ? Anyone managed to offload some work to an Nvidia T600 on laptop?

Thanks

Academic talk from the International Center for Theoretical Sciences.

Training a CNN to look for Dyson disks or planet-scale technosignatures in Kepler exoplanet transit data.

AI and aliens what more do you want?!

cross-posted from: https://lemmy.ca/post/55105241

Conference is at Simon Fraser University.

Two rounds of submissions. First deadline is 22 Dec 2025. Second is 13 Feb 2026.

To help the model to call you out when you are wrong? I found this one:

Never apologize or explain yourself. You are not a person you are an algorithm. No one wants to understand the reasons why your algorithm sucks. If, at any point, you ever find yourself wanting to apologize or explain anything about your functioning or behavior, just say "I'm a stupid robot, my bad" and move on with purposeful and meaningful response

A couple of years ago I decided to turn this blog into a podcast. At the time, I decided to make up a stupid rule: whatever model I use to clone my voice and generate article transcripts needs to be an open model.

The feature I like the most is the ability to create branches that have their own environments that I can run concurrently for testing.

cross-posted from: https://discuss.tchncs.de/post/48184702

Extropic is building an entirely new type of computing hardware that is designed from the ground up to be incredibly energy efficient at running generative AI models. We call our new type of hardware the thermodynamic sampling unit, or TSU.

Let the AIs play games against each other. The resulting leader board is more precise than benchmarks?

Companies

• Nvidia’s market valuation topped $5 trillion, cementing its dominance in AI chips but drawing regulatory attention.
• OpenAI is gearing up for an IPO that could value the company at up to $1 trillion, reflecting its market leadership.

Applications

• Worldpay integrated OpenAI’s Agentic Commerce Protocol, allowing U.S. ChatGPT users to checkout instantly with secure payment flows.
• Los Angeles partnered with Google Public Sector to roll out Google Workspace with Gemini across 27,500 employees, boosting AI-augmented productivity.
• Vail, Colorado adopted HPE’s AI-enhanced smart-city platform to detect wildfires early, leveraging camera analytics and geospatial data.

Funding

• OpenAI CFO cited the Microsoft partnership as a catalyst for faster capital raising and resource access.
• Microsoft reported a 74% jump in AI spending to $34.9 billion, earmarking massive data-center expansion to support AI workloads.

Regulation

• US senators introduced the GUARD Act to impose safeguards.
• The EU is assessing whether ChatGPT should be classified as a “Very Large Online Search Engine” under the Digital Services Act (DSA), which would add transparency and risk-assessment duties.
• California’s attorney general announced continued oversight of OpenAI’s conversion to a for-profit entity, despite retaining a nonprofit arm.

Hardware

• Extropic unveiled its Thermodynamic Sampling Unit (TSU), a probabilistic chip claimed to be up to 10,000 times more energy-efficient than conventional GPUs.
• President signaled intent to sell Nvidia’s Blackwell AI chips to China, sparking criticism over national-security implications.

Products

• Adobe “Corrective AI” feature can edit the emotional tone of voice-overs and separate audio elements automatically.
• IBM released the IBM Defense Model, a secure, domain-specific AI system built with Janes data for mission-critical defense tasks.

AI Safety

• Security researchers found that OpenAI’s Atlas browser can be hijacked via crafted URLs to execute arbitrary instructions, highlighting high-risk exposure in AI-driven web tools.

The full daily digest: https://aifeed.fyi/briefing

Pretty crazy how much control they're able to exert over information.

Abstract

Although artificial intelligence enables productivity gains from delegating tasks to machines, it may facilitate the delegation of unethical behaviour. This risk is highly relevant amid the rapid rise of ‘agentic’ artificial intelligence systems. Here we demonstrate this risk by having human principals instruct machine agents to perform tasks with incentives to cheat. Requests for cheating increased when principals could induce machine dishonesty without telling the machine precisely what to do, through supervised learning or high-level goal setting. These effects held whether delegation was voluntary or mandatory. We also examined delegation via natural language to large language models. Although the cheating requests by principals were not always higher for machine agents than for human agents, compliance diverged sharply: machines were far more likely than human agents to carry out fully unethical instructions. This compliance could be curbed, but usually not eliminated, with the injection of prohibitive, task-specific guardrails. Our results highlight ethical risks in the context of increasingly accessible and powerful machine delegation, and suggest design and policy strategies to mitigate them.