nishio: this is a very valuable article to read again. Here are the parts of the proposal that I thought were a waste of time. I think itā€™s a waste of time to read the proposal. ļ½œDaiki Teramotoļ½œnote

nishio: If you see ā€œI just listed all the ideas and didnā€™t prioritize themā€ and think ā€œeh, Iā€™ve already written a lot, what should I do?ā€, you should delete it. I think it would be a good idea to add a chapter ā€œAppendix: Other ideasā€ at the end instead of deleting it. With feedback, they can say, ā€œThis idea in the Appendix is interesting, too.

nishio: we are not espers and cannot read what is not written. So it is better to have it written than not written. In addition to that, we also want to know ā€œwhich you think is the most importantā€, and it is difficult to understand that in a memo that lists all the ideas. So what you have already written should be in an appendix, not erased.

nishio: ā€œItā€™s not a logical argument to say that recent artificial intelligence is great ā†’ this project is great too.ā€ Certainly. Itā€™s not logical to say, ā€œArtificial intelligence is great these days, this project will be solved by artificial intelligence, therefore this proposal is great.ā€ I feel like this logically crazy thinking is often addicted even to adults.

nishio: Artificial intelligence does not solve problems, you solve problems. You find the means necessary to solve the problem and you implement them. Sometimes that means of solution is called ā€œartificial intelligenceā€ in the world, but that is not the essence of problem solving. ā€œWhat exactly is this ā€˜artificial intelligenceā€™ process?ā€ It would be good to answer the question

nishio: The term ā€œartificial intelligenceā€ is vague, but a characteristic common to a relatively large number of methods is that ā€œinstead of humans implementing a lot of conditional branches, they collect a lot more The term ā€œartificial intelligenceā€ is ambiguous, but a common feature of many methods is that they collect more data and make conditional decisions based on that data, instead of implementing many conditional branches by humans. is the assumed question, isnā€™t it?

nishio: Also, since many artificial intelligence projects are of the nature that ā€œyou donā€™t know if it will work or not until you try it,ā€ I am also concerned about how to avoid having to announce at the results briefing that ā€œwe tried it, but it didnā€™t workā€ when it doesnā€™t work out. I am also concerned about how to avoid having to announce at the debriefing session that ā€œwe tried it, but it didnā€™t workā€ if it doesnā€™t work. This is a difficult question, and I donā€™t have the right answer.

nishio: In some situations, ā€œFirst build something that works on a rule basis, then actually use it to collect data. Sometimes it is effective to ā€œuse artificial intelligence not as an essential component, but as an added performance improvement,ā€ but not all of the time. Often there are cases where it is difficult to write a rule base.

nishio: Related A text for people who want to help creators who want to enter some kind of contest but donā€™t know how to advise them. [Type Advice for Contest Entrants - NISHIO Hirokazuā€™s Scrapbox https://scrapbox.io/nishio/%E3%82%B3%E3%83%B3%E3%83%86%E3%82%B9%E3%83%88%E5%BF%9C%E5%8B%9F%E8% 80%85%E3%81%B8%E3%81%AE%E3%82%BF%E3%82%A4%E3%83%97%E5%88%A5%E3%81%AE%E3%82%A2%E3%83%89%E3%83%90%E3%82%A4%E3%82%B9]


This page is auto-translated from /nishio/2021-04-12-2 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.