GPT-4.iconnishio.icon

relevance - GPT4 can search structure by structure.

hrjn Why couldnā€™t a machine beat a human in terms of amount of knowledge? I think this kind of thing shows that the problem was the search (e.g. for someone using a search engine) technology.

Traditionally, it was necessary to be able to search in a suitably specialized language, but even a list of layman-like words can be searched in the LLM world.

There is a certain ambiguous search pole.

kajikent: the story of how GPT-4 saved my dogā€™s life. When I asked GPT-4 about the cause of my dogā€™s serious illness that even the vet couldnā€™t figure out, he immediately suggested a hypothesis as to the cause. LLM will become a social infrastructure, not just a convenience service. twitter.com/peakcooper/sta..

hrjn Of course whether the search results are correct or not is in the hands of humans, but it is possible that the medical language model will replace the ā€œsearching various papers for unknown causes or seeking knowledge from acquaintances who are experts,ā€ as is often the case in reenactments and other dramas. However, there is a possibility that the medical language model will replace the ā€œsearching various papers or seeking knowledge from acquaintances who are experts in the field,ā€ as is often the case in dramas.

golden_lucky [see story contrasting search (e.g. for someone using a search engine) and LLM Every time I see this, I feel like there has been such a lack of demand for being able to refer to where information is, rather than interpreting it.

bonotake I think Barners-Leeā€™s semantic web is the best example of this. I think Barners-Leeā€™s semantic web is the best example of this, and it is rather mainstream of WWW to think of the web as a big knowledge base and to let people interpret information. just that Google has been acting on its behalf so far.

golden_lucky I see, so itā€™s not too far-fetched to think that the dream of the semantic web has been reinvented in another form? I seeā€¦!


This page is auto-translated from /nishio/ćØć¦ć‚‚ę›–ę˜§ę¤œē“¢ using DeepL. If you looks something interesting but the auto-translated English is not good enough to understand it, feel free to let me know at @nishio_en. Iā€™m very happy to spread my thought to non-Japanese readers.