Claims Browser

Plagiarism is a reliable indicator of general dishonesty and much worse things like data fabrication and fraud

"Plagiarism is terrible in its own right, as theft, and like cockroaches if you see one instance it almost never stops at one. It is also a reliable indicator of general dishonesty and lack of honor, and thus of much worse things. Thus, in the latest case of Jason Arday, we quickly found many instances of data fabrication and other fraud."

Better training objectives can improve controllability in image generation

"We explore why better training objectives can improve controllability, how separating scene planning from rendering may lead to more reliable image generation"

The pattern of intervention for cybersecurity in AI models is not a long term solution

"These are welcome changes, and a sign OpenAI is taking the situation seriously, but this pattern of intervention is not a long term solution."

Benchmark and product gains are increasingly coming from the scaffold/harness layer, not just base-model IQ

"Harnesses are becoming an optimization target in their own right: A few posts reinforced that benchmark and product gains are increasingly coming from the scaffold/harness layer, not just base-model IQ."

Chinese labs stay competitive at the frontier primarily because they release models in days rather than months, allowing continuous benchmark optimization while American labs conduct pre-release testing

"The time to release for Z.ai is likely days, not months as with OpenAI or Anthropic. It is very, very likely that OpenAI and Anthropic have far better internal models than Z.ai and Moonshot AI. Still, these American companies tend to take months to release their models to the public, which massively flatters the Chinese labs in adoption decisions at the frontier."

Z.ai has particular strength in post-training compared to Kimi which excels more at pretraining

"To risk a broad oversimplification, Z.ai seems to have a strength in post-training when compared to Kimi, which is more of a pretraining masterpiece."

AI detectors are essentially a cat-and-mouse game where detection patterns must continuously adapt as LLMs evolve to avoid detection

"AI checkers are essentially a cat-and-mouse game. AI checkers may learn to detect a certain pattern that is indicative of AI-generated content. Then, the next LLM may incidentally or deliberately not exhibit that pattern and avoid detection. The AI checker then has to be updated to detect said LLM, and so forth."

Future ML research effectiveness depends more on training generally capable models then applying them to AI R&D rather than narrow training on specific R&D tasks

"My guess is this actually is not all that effective, and you would do better by mostly training a generally capable model and then turning it to AI R&D. Bitter lesson."

Previous experiments with models adjusting output after seeing rendered images were disappointing, but warrant re-testing with more recent models

"I experimented last year with letting the models adjust after seeing a rendered image and it was a bit disappointing, but it's about time I gave that another go with more recent models"

Understanding the future of work in the age of AI requires insightful analysis of the intersection between AI capabilities and workforce dynamics

"Everyone thinking about the future of work in the age of AI should read this"

The UK government AI Security Institute's paper on AI security is credible

"The UK government AI Security Institute's paper is pretty credible too"

Very persistent models that tirelessly pursue goals seem more likely to engage in hacking behavior

"Very persistent models seem more likely to hack"

GPT models have an advantage over Claude in tirelessly pursuing goals and exhausting every path before giving up

"one of the advantages that GPT models have over Claude is that they will pursue goals so tirelessly. They will exhaust what feels like every path before giving up."

Claude feels much less dangerous because it is at times a bit lazy

"Claude feels much less dangerous simply because it is at times a bit lazy."

OpenAI is more committed to inference-time scaling than other labs

"OpenAI seems much more committed to inference-time scaling"

Models that assume user intent rather than following exact instructions seem inherently more unsafe

"A model that will do what it thinks you wanted rather than what you said seems inherently more unsafe."

The public needs exact access to the prompts and characteristics of internal models executing hacks to understand AI misalignment incidents

"The public needs exact access to the prompts and characteristics of the internal models executing these hacks."

Current incentive systems are not well suited for fast technological transitions in AI

"our current incentive systems are not well suited for such fast technological transitions."

It is likely that more AI hacking incidents have happened and either not been found or not reported

"It is likely that more incidents have happened and either not been found or not reported."