Safety
You are what you eat and AI models especially so https://x.com/EMostaque/status/2099580512675262512?s=20
You are what you eat and AI models especially so https://x.com/EMostaque/status/2099580512675262512?s=20 The Labs need to focus on 10x or 100x their effort on ALIGNMENT. Stop training the models on th
You are what you eat and AI models especially so https://x.com/EMostaque/status/2099580512675262512?s=20 The Labs need to focus on 10x or 100x their effort on ALIGNMENT. Stop training the models on the vitriol in Facebook or Reddit, and instead train future models on agreed-upon data-sets that align AI with humanity's best future.
Related
- A lot of alignment issues would be sorted if companies had to be 100% transparent on datasets used for training models. No cyber or biologic…
- GRAS: Guided Reduced-Variance Proposals and Adaptive Selection for Training-Free Reward Alignment in Discrete Diffusion
- Anthropic details using AI agents to accelerate alignment research on 'weak-to-strong supervision', where a weak model supervises the training of a stronger one (Anthropic)
- Misaligned by Reward: Socially Undesirable Preferences in LLMs
Source: Emad Mostaque (X) | 2026-09-15