Andrej Karpathy's minimal microgpt has been extended with a secret-keyed Gumbel sampler that watermarks LLM output without altering token probabilities, achieving a p-value of 6.844e-53 for watermarke…
Anthropic has begun watermarking outputs of its Claude AI model using a scheme based on Google's SynthID and the Gumbel Softmax method proposed by Scott Aaronson in 2022. Aaronson, who credits Anthrop…
Anthropic detailed on August 14, 2026 that future Claude models will embed a statistical text watermark based on Google DeepMind's SynthID-Text to comply with the EU AI Act, which requires AI-generate…
Anthropic revealed that its new text watermark is a version of Google DeepMind's SynthID-Text, which works by altering the randomness of word selection to embed an undetectable pattern that can be che…
OpenAI's internal model has solved ten open problems in mathematics and theoretical computer science, including parallel repetition for arbitrary quantum games and a lower bound on the arithmetic circ…
Scott Aaronson and his brother announced the launch of BQP Partners, a new angel-investing venture, marking Aaronson's entry into startup investing after years of focusing on academic research in quan…
Aran Nayebi's new paper argues that AI consciousness may be an inevitable byproduct of building highly capable agents, using selection theorems to show that robust decision-making under uncertainty re…
More than 100 University of California faculty members signed an open letter opposing a proposed policy that would require professors to notify university police of any student who expresses suicidal …