AI in Bed

Scaling laws, evals & safety

Why bigger got better, how we measure frontier models, and the open problems in keeping AI safe.

⏱ 7 min read

Much of the last decade of AI progress came from one discovery: performance improves predictably as you scale up compute, data and model size. That predictability shapes billion-dollar decisions, and raises serious questions.

10^1810^2010^2210^24training compute (FLOPs, log scale) →loss (log) ↓ betterpredictedeach dot = one model size
On a log–log plot, loss falls in a remarkably straight line as training compute grows (illustrative).

📈 Scaling laws

Kaplan et al. (2020) found loss follows smooth power laws in parameters, data and compute.

Chinchilla (2022) showed many models were under-trained: for a fixed compute budget, scale parameters and tokens together (roughly 20 tokens per parameter).

🧩 Quick quiz

What was the key Chinchilla finding?

🧪 Evaluating frontier models

Benchmarks test knowledge (MMLU), maths (GSM8K, MATH), coding (HumanEval, SWE-bench) and more.

Problems: benchmarks saturate, and test questions can leak into training data (contamination). So labs increasingly use held-out, human-rated and agentic evaluations.

🛡️ Alignment & safety

Alignment: making systems pursue what we actually intend. Reward hacking and sycophancy are everyday examples of misalignment.

Robustness: resisting jailbreaks and prompt injection.

Interpretability: understanding what's happening inside the network, e.g. finding human-understandable features inside models.

🏛️ Responsible deployment

Labs publish model cards and system cards, run red-teaming, and test for dangerous capabilities before release.

Policies like responsible scaling commitments tie stronger safeguards to more capable models.

🧩 Quick quiz

What is "benchmark contamination"?

✨ Before you drift off

  • Loss falls predictably with scale (power laws).
  • Compute-optimal training scales data with parameters.
  • Benchmarks saturate and leak, so evaluation is an active field.
  • Alignment, robustness and interpretability are core open problems.

📚 Go deeper (free & open)