If algebraic value edits work at all in LMs, they are hard to make without substantially degrading capabilities

5

24

Ṁ47Ṁ130

resolved May 13

Resolved

NO1D

1W

1M

ALL

Get Ṁ600 play money

# 🏅 Top traders

# | Name | Total profit |
---|---|---|

1 | Ṁ18 | |

2 | Ṁ9 | |

3 | Ṁ3 |

## Related questions

Will AI be able to solve confusing but elementary geometric reasoning problems in 2024?

47% chance

Are autoregressive LLMs a doomed paradigm? (Inspired by LeCun's tweet)

45% chance

Are LLMs capable of reaching AGI?

56% chance

Will any language model trained without large number arithmetic be able to generalize to large number arithmetic by 2026?

69% chance

Will reinforcement learning overtake LMs on math before 2028?

43% chance

are LLMs easy to align because unsupervised learning imbues them with an ontology where human values are easy to express

32% chance

By 2028 will we be able to identify distinct submodules/algorithms within LLMs?

75% chance

Will LLMs be worse than human level at forecasting when they are superhuman at most things?

46% chance

Will LLM based systems have debugging ability comparable to a human by 2030?

59% chance