Humans are natural pattern seekers. Pattern recognition often fails in complex and evolving environments. The economist Blake LeBaron is a leader in the field of agent-based modeling. These are models that create agents in silico, provide them with decision rules, and let them trade an asset among themselves. He finds that the asset price rises even as the diversity of decision rules declines because the similar trading strategies reinforce the price movement. But at some point the market becomes fragile. At that critical juncture, a small incremental loss of diversity leads to a sharp plunge in the asset price because the buyers are exhausted. The relationship between diversity loss and asset price change is non-linear. Diversity breakdowns fit a pattern but measuring diversity is inherently difficult. Pattern recognition can also fail because of how our minds love to think in analogies. Steps include selection, mapping, evaluation, and learning. To understand a target topic we commonly start by selecting an analog, usually from memory.
- Date Posted:
- December 12, 2023
Macron said he was concerned that the new law means the EU will enforce the world’s toughest regime on so-called foundational models, the technology that underpins generative AI products such as OpenAI’s ChatGPT. Macron’s comments may presage a new battle over the final terms of the AI Act, which still needs to be ratified by member states over the coming weeks. “The stakes are high and the French will try to block this,” said a person with knowledge of the talks, although they thought the text would ultimately be agreed despite opposition. French negotiators had tried to water down the proposed rules in marathon talks last week, but ultimately the bloc moved towards a tough regulatory regime that imposes new transparency requirements on powerful AI models as well as strict restrictions on the use of facial-recognition technology. The new rules, which will probably come into force in early 2025, also introduce prohibitions on the use of AI for “social scoring”, the use of metrics to classify people based on their behaviour or personal characteristics.