Google tests double-blind AI evaluation
AI benchmarks have a trust problem and Google wants to fix it
This collaboration with Singapore's AI Safety Institute could become the new standard for trustworthy AI evaluations in the region.
Google Deepmind is implementing a double-blind evaluation for a frontier AI model using cryptographic protection through Confidential Space. The pilot project with the Singapore AI Safety Institute uses a Gemini Flash Lite model. This approach prevents Google from seeing test questions and evaluators from seeing model weights. The initiative aims to establish tamper-proof AI benchmarks.
AI benchmarks have a trust problem and Google wants to fix it
Google Deepmind is testing a double-blind evaluation of a frontier AI model for the first time. Cryptographic protection through Confidential Space is meant to keep Google from seeing the test questions and keep evaluators from seeing the model weights. The pilot project with the Singapore AI Safety Institute uses a Gemini Flash Lite and could set a new standard for tamper-proof AI benchmarks. The article AI benchmarks have a trust problem and Google wants to fix it appeared first on The Decoder .
