2 ms·
>If the model were trained on actual observations of planetary trajectories it would trivially recreate keplers laws, newtons laws etc. Should be easy to prove
by soperj 4y ago
>If the model were trained on actual observations of planetary trajectories it would trivially recreate keplers laws, newtons laws etc.
Should be easy to prove.
- ben_w 4y agoIndeed: • https://news.ycombinator.com/item?id=31609839 https://news.ycombinator.com/item?id=31609839 • https://arxiv.org/pdf/1810.10525.pdf https://arxiv.org/pdf/1810.10525.pdf • https://arxiv.org/pdf/1905.11481.pdf https://arxiv.org/pdf/1905.11481.pdf
- soperj 4y agofrom your link: "Renner stresses that although the algorithm derived the formulae, a human eye is needed to interpret the equations and understand how they relate to the movement of planets around the Sun."
- ben_w 4y agoEven if that had been the only example rather than just one among them, it would still demonstrate the claim that an AI can discover various laws of physics just from observing data.