Подтвердите e-mail

Для публикаций, комментариев, реакций и сообщений подтвердите адрес.

Публикация

Предыдущая публикация

They do that already, because just as you don't need an LLM to generate bad spellchecker suggestions, you also don't need an LLM to generate sub-par educational materials. There are plenty of "educational" publishers who'll do that all day long. 59/

1302

AI is a normal tech. That means there are times when it is useful and times when it is pointless or actively harmful. One rule of thumb for chatbots is that they can only provide useful information to experts who have the discernment to ignore the defective output that LLMs *always* emit. 60/

Обсуждение

3 прямых ответов · 16 сообщений

That means that the dream of chatbots as replacements for teachers is a nightmare. Getting rid of teachers because we all have chatbots is like getting rid of doctors because we all have the plague. eof/

Ответ для By Cory Doctorow (GPG 0xBF3D9110957E5F4C)

yep. less human interaction is going to go well.

Ответ для By Cory Doctorow (GPG 0xBF3D9110957E5F4C)

What text editor do you use? What is your essay checking software? Is it off the shelf free/libre?

Ответ для hugh1620

I just use emacs or Geany. The "checking software" is whatever default chatbot I installed when I apt-got ollama last year.

Ответ для By Cory Doctorow (GPG 0xBF3D9110957E5F4C)

we found Google’s Gmail suggestions to be so awful we turned them off for the whole company… been trying to write a replacement. Any tips on your prompt? We added a feature for “what are three questions the reader might ask” having the LLM stand in for the reader is a lot better than standing…

Ответ для chrisvls

in for the writer. In part because it takes a lot of calories to reject a bad writing suggestion — even grammar suggestions — and very few to reject a dumb reader question. We also added voices, like “sarcastic mocking feedback” which is reliably hilarious and “burned out midcareer midwesterner”…

Ответ для chrisvls

…which also make it less draining to use (we think)

Ответ для chrisvls
Ответ для By Cory Doctorow (GPG 0xBF3D9110957E5F4C)

Sorry if this is a dumb question, but how is an LLM bot that searches your doc for typos different to a standard spell-checker that checks your doc on the fly?

Ответ для Jason

Sell-check has a hard time with typos that make other words.

Ответ для By Cory Doctorow (GPG 0xBF3D9110957E5F4C)

Excellent. Amusingly, we put an option in to just have the content suggestions and turn off spell and grammar checking. The robot still puts them in even when we tell it not to, unless we basically threaten it with bad yelp reviews...

Ответ для By Cory Doctorow (GPG 0xBF3D9110957E5F4C)

Ok, but what can we do? For some reason I'm not equipped to understand, we have leaders who are ruinously unqualified to deal with "AI" and just see it as a way to replace people and save money. These same people seem pretty eager to destroy education outright.

So you should only use AIs when you don't need them.

Ответ для Weekend Editor

I don't know which of my 1600 essays contains the point I'm thinking about. I need to. I don't know which typos are in the essay I just wrote. I need to. Being "expert" doesn't mean you don't need tools.

Ответ для By Cory Doctorow (GPG 0xBF3D9110957E5F4C)

I was agreeing with you that it's a case of Gell-Mann amnesia. :-)

Here's a quick AI test anyone can do-- search VZ earthquake... Which I did to get an update on casualties... Note the AI summary listed first... Then open the Wiki page to see the real numbers. This is the magnitude of error AI introduces. duckduckgo.com?q=vz+earthqu...

vz earthquake at DuckDuckGoDuckDuckGo. Privacy, Simplified.duckduckgo.com