news 
This Week in AI: Will Biden’s AI Actions Survive the Trump Era?
In a recent study, researchers have found that as AI systems’ performance improves, they tend to “score better” on safety benchmarks, inadvertently capturing latent factors closely tied to general capabilities and raw training compute. This raises concerns about the effectiveness of current AI safety metrics in truly evaluating the risk posed by these intelligent machines.
Meanwhile, Sakana AI has made significant strides in the field with its new Transformer² system, which dynamically adjusts to new tasks. This innovative approach could potentially lead to a future where AI models are no longer static, but rather adapt and evolve as needed.
In other news, a small team of developers has released an open alternative to AI-powered search engines like Perplexity and OpenAI’s SearchGPT. PrAIvateSearch, available on GitHub under the MIT license, is powered by openly available AI models and services, including Alibaba’s Qwen family of models and DuckDuckGo.
Kyle Wiggers, Senior Reporter at TechCrunch, takes a closer look at these developments in this week’s AI roundup.
Source: techcrunch.com
