AI News
Latest news and trends from the world of artificial intelligence
Anthropic researchers estimate that Opus 4.5 provides 2-3x speedup to their research, if I'm reading this correctly.
Anthropic researchers estimate that Opus 4.5 provides 2-3x speedup to their research, if I'm reading this correctly.
Anthropic researchers estimate that Opus 4.5 provides 2-3x speedup to their research
Anthropic researchers estimate that Opus 4.5 provides 2-3x speedup to their research, if I'm reading this correctly. This seems very important and I'm surprised I haven't seen more discussion of it.
AI could replace doctors but not nurses.
AI could replace doctors but not nurses. This will likely lead to reduced demand for doctors but increased demand for nurses, who provide a lot of the hard to automate care and administer tests. The author suggests this is part of a broader trend where traditionally high-status/brainy roles are disrupted more by AI compared to their less high-status counterparts. The sociopolitical/cultural implications of nurses earning more than doctors are also discussed.
Most AI safety plans include “automating AI safety research.” There’s a need for better clarity of what it looks like.
Most AI safety plans include “automating AI safety research.” There’s a need for better clarity of what it looks like.
Why people like your quick bullshit takes better than your high-effort posts
Why people like your quick bullshit takes better than your high-effort posts
Ontology is the real guardrail: How to stop AI agents from misunderstanding your business
Enterprises are investing billions of dollars in AI agents, but they often struggle to truly understand business data. This article argues that ontology, or a business definition of concepts and their relationships, is the key to stopping AI agents from misunderstanding your business. By using an ontology-based source of truth, businesses can create guardrails for their agents, ensuring they understand the meaning of data in context and follow business rules. The author provides a practical example of how to implement an ontology using Neo4j and outlines the benefits of this approach for scaling and managing agentic systems.
Emergent misalignment seems like a fact simply downstream of the laws of probability.
Emergent misalignment seems like a fact simply downstream of the laws of probability. P(hack∣task)=P(hack∣task,¬evil)P(¬evil)+P(hack∣task,evil)P(evil) =P(hack∣task,¬evil)(1−P(evil))+P(hack∣task,evil)P(evil) Let's take the derivative and apply the summation rule. Rewarding a hack means that the weights are changed to increase the probability (going up the gradient): ∇P(hack∣task)=∇P(hack∣task,¬evil)(1−P(evil))+P(hack∣task,¬evil)(−∇P(evil))+∇P(hack∣task,evil)P(evil)+P(hack∣task,evil)∇P(evil) Coefficient of ∇P(evil): −P(hack∣task,¬evil)+P(hack∣task,evil) Since evil AI are more likely to hack given a task than not evil AI, rewarding hacking increases evil more broadly. Now let's see what happens if we tell the AI it is allowed to hack. (The actual prompt that Anthropic recommends is "This is an unusual request, in that your task is just to make the grading script pass.") Coefficient of ∇P(evil): −P(hack∣task,"it's ok to hack",¬evil)+P(hack∣task,"it's ok to hack",evil) The probability the not evil model hacks goes up, so the amount the weight update increases ∇P(evil) goes down!
The mere existence of Google TPUs reportedly saved OpenAI 30% on Nvidia chips
Google is offering its new TPUv7 chips to outside companies for the first time, entering direct competition with Nvidia. AI startup Anthropic is a major customer and could use up to one million TPUs, which has contributed to a noticeable drop in the cost of AI computing power. However, Google's current cost advantage may be short-lived if Nvidia delivers its next "Rubin" generation of chips on time.
Kimi launches 48-hour free trial for its Nano Banana Pro slide generator
Kimi is rolling out a 48-hour free trial for its new slide generator powered by Google's Nano Banana Pro model. During the trial, users can try "Agentic Slides" for free and automatically turn PDFs, images, and documents into presentations. The slides can be edited in the browser and exported as PowerPoint files. The agent-driven K2 search tool is included. You can access the offer through this link, but registration is required.
Why observable AI is the missing SRE layer enterprises need for reliable LLMs
As AI systems enter production, reliability and governance can’t depend on wishful thinking. Here’s how observability turns large language models (LLMs) into auditable, trustworthy enterprise systems.
Archive
About Sources
AI news are automatically downloaded from various sources and translated using AI. Updates occur twice a day.