Skip to main content

Anthropic aims to fix one of the biggest problems in AI right now

the Anthropic logo
Anthropic

Hot on the heels of the announcement that its Claude 3.5 Sonnet large language model beat out other leading models, including GPT-4o and Llama-400B, AI startup Anthropic announced Monday that it plans to launch a new program to fund the development of independent, third-party benchmark tests against which to evaluate its upcoming models.

Per a blog post, the company is willing to pay third-party developers to create benchmarks that can “effectively measure advanced capabilities in AI models.”

“Our investment in these evaluations is intended to elevate the entire field of AI safety, providing valuable tools that benefit the whole ecosystem,” Anthropic wrote in a Monday blog post. “Developing high-quality, safety-relevant evaluations remains challenging, and the demand is outpacing the supply.”

The company wants submitted benchmarks to help measure the relative “safety level” of an AI based on a number of factors, including how well it resists attempts to coerce responses that might include cybersecurity; chemical, biological, radiological, and nuclear (CBRN); and misalignment, social manipulation, and other national security risks. Anthropic is also looking for benchmarks to help evaluate models’ advanced capabilities and is willing to fund the “development of tens of thousands of new evaluation questions and end-to-end tasks that would challenge even graduate students,” essentially testing a model’s ability to synthesize knowledge from a variety of sources, its ability to refuse cleverly worded malicious user requests, and its ability to respond in multiple languages.

Anthropic is looking for “sufficiently difficult,” high-volume tasks that can involve as many as “thousands” of testers across a diverse set of test formats that help the company inform its “realistic and safety-relevant” threat modeling efforts. Any interested developers are welcome to submit their proposals to the company, which plans to evaluate them on a rolling basis.

Andrew Tarantola
Former Computing Writer
Andrew Tarantola is a journalist with more than a decade reporting on emerging technologies ranging from robotics and machine…
Kagi’s AI search assistant gives you access to all the big models in one place
Kagi search bar in light mode.

Kagi's "Assistant" feature, previously only available to Ultimate subscribers, is now rolling out to all tiers -- including the free trial tier. The feature gives you access to a range of different LLMs for both chatting and web-searching purposes.

If you don't know much about Kagi, it's a paid search engine that borrows its name from the Japanese word for "key." The concept is simple -- with Google, you pay for the service by allowing ads and data collection. With Kagi, you pay for the service with money to get a private and ad-free experience.

Read more
Microsoft’s Copilot Vision AI is now free to use in Edge
Copilot Vision graphic.

After months of teasers, previews, and select rollouts, Microsoft's Copilot Vision is now available to try for all Edge users in the U.S. The flashy new AI tool is designed to watch your screen as you browse so you can ask it various questions about what you're doing and get useful context-appropriate responses.

The feature works for "most" sites, according to Microsoft but it gives you a list of recommendations to start with. We have Amazon, which makes sense, but also Geoguessr? I'm pretty sure the point of that site is to try and guess where you are on the map without any help. Anyway, the full list of starter sites is as follows:

Read more
You can now view all of your ChatGPT-generated images in one place
ChatGPT library promotion video.

OpenAI did text generation and image generation separately for quite a while, but that all changed a couple of weeks ago when it added image capabilities directly into ChatGPT. Now, a small but powerful Quality of Life update gives users access to an image library where they can see all of the insane things they've created.

https://twitter.com/OpenAI/status/1912255254512722102

Read more