On the dangers of stochastic parrots: Can language models be too big? (2021)

Re:On the dangers of stochastic parrots: Can language models... by jasmer on Saturday January 14, 2023 @03:06PM (#34383667)

This is ok. 90% of research is creative thinking, dialogue. One idea creates the next, some are a foil, some are dead ends. As long as there are not outrageous claims being made for 'hard evidence' where there is none, it's fine. Maybe the format isn't fully appropriate but the content is. Most good things come about in a non-linear process which involves provocation along the line somewhere.

[ Parent | on HN ]

Re:On the dangers of stochastic parrots: Can language models... by janalsncm on Saturday January 14, 2023 @04:39PM (#34384418)

I expect science to have a hypothesis which can be falsified. Otherwise it’s just opining on a topic. Otherwise we could just call this HN thread “research”.

[ Parent | on HN ]

Re:On the dangers of stochastic parrots: Can language models... by joshuamorton on Saturday January 14, 2023 @04:42PM (#34384446)

Position papers are exceedingly common. Common enough that there's a term for them.

[ Parent | on HN ]

Re:On the dangers of stochastic parrots: Can language models... by jasmer on Sunday January 15, 2023 @12:33AM (#34386987)

Science is 80% discussion, opining, thinking, reading etc.

[ Parent | on HN ]

Re:On the dangers of stochastic parrots: Can language models... by janalsncm on Sunday January 15, 2023 @08:18PM (#34395624)

A lot of areas are made up of those things. Theology, for example. What sets science apart is testable hypotheses.

[ Parent | on HN ]