Post ASmCl7qOUy0bhVhEHo by thibaultmol@en.osm.town
(DIR) More posts by thibaultmol@en.osm.town
(DIR) Post #ASlxryZlrBfzfejHfM by stux@mstdn.social
2023-02-17T14:03:41Z
0 likes, 1 repeats
It's weird to see how many people take the replies from #ChatGPT for true :thinktink: Like the devs also tell you, it's an experimental bot that has been trained on data sets and may not be up-to-date with current events or the data could just be wrongIt's quite easy to 'trick' an AI chatbot to say something that's untrue or just made upIf you use it for work be sure to check the output you let the AI write :blobcatgiggle:
(DIR) Post #ASlyEU6zHw8Bqe00cC by ogtrekker@universeodon.com
2023-02-17T14:07:51Z
0 likes, 0 repeats
@stux The ability to perform critical thinking is becoming a lost art. Many people believe whatever they are told without fact-checking. I don’t understand it. 🤦♂️
(DIR) Post #ASlyF3YVVbwNm88yrw by Adardis@zirk.us
2023-02-17T14:07:54Z
0 likes, 0 repeats
@stux It's even weirder that the CEO of OpenAI promises that you will be able to tell it your values and it will think about that and write based on your values: it will be *your* AI.
(DIR) Post #ASlyKjNhQgRowXnJtQ by stux@mstdn.social
2023-02-17T14:08:56Z
0 likes, 0 repeats
@Adardis It will be an AI made of our stuff more likely :blobcatgiggle: if we want it or not
(DIR) Post #ASlyVFTwvyr1OjjW9g by MeadHallBardPhantom@mastodon.social
2023-02-17T14:10:53Z
0 likes, 0 repeats
@stux It seems to me that people have forgotten the principle of GIGO. Garbage In, Garbage Out.If is been trained with carefully curated data, then its in with a chance.If it's been reading Twitter, Facebook or watching YouTube... 😬
(DIR) Post #ASlz8zH2cCauiEe07k by stux@mstdn.social
2023-02-17T14:17:59Z
0 likes, 0 repeats
@carstenraddatz_fca I got my own running here :blobcatgiggle: https://loazy.com(if my account credit still holds :catblush: )
(DIR) Post #ASm0JZ2mZt0gV8a9ei by trebach@masto.ai
2023-02-17T14:31:11Z
0 likes, 0 repeats
@stux Edit: If you use it for work: STOP!
(DIR) Post #ASm0pgvYnoqQDaWVxQ by reducer@mstdn.social
2023-02-17T14:36:56Z
0 likes, 0 repeats
@stux I've used it for stuff like "How do I do X using Terraform for AWS?" and it never gives me 100% correct answers, but it's correct enough to let me see what it's going for and adjust for the errors.Pretty good when you sort-of know the answer already.
(DIR) Post #ASm0qTDdFvoxzXLcn2 by NaMi@mstdn.social
2023-02-17T14:37:00Z
0 likes, 0 repeats
@stux Even with updated or correct data in the data sets, it's important to understand that #ChatGPT et al's responses are NOT based on checking or understanding "factuality", but simply calculate statistical likelihood. (e.g. A verb usually follows after a noun, the term "night" is more often combined with "dark" than with "bright" etc.) Might work well for some tasks like coding, but not when chatgpt writes your PolSci paper with what "looks like" references.
(DIR) Post #ASm0znSrLvRHO8uwls by BCMetisMan@mastodon.social
2023-02-17T14:33:48Z
0 likes, 0 repeats
@trebach @stux too late many businesses have already been using it and are making policy to incorporate it. You can't stop progress no matter how hard thine protest. You know people said the very same thing about the internet when it began also.
(DIR) Post #ASm0znzTOfWj1IAzbc by stux@mstdn.social
2023-02-17T14:38:46Z
0 likes, 0 repeats
@BCMetisMan @trebach Ive even seen some SaaS projects already using and selling it's use :amaze:
(DIR) Post #ASm1KHfWlOBSrCbzpw by driesverlinden@mastodon-belgium.be
2023-02-17T14:42:31Z
1 likes, 1 repeats
@stux ... or while it not a good a idea to have a complete trust in the chatbot to make your homework ;)
(DIR) Post #ASm1s56vxZv1ksXxui by stacey_cruz_@mastodon.social
2023-02-17T14:48:29Z
0 likes, 0 repeats
@stux personally I never trust it,,giving basic algorithm makes it easy to build my codes
(DIR) Post #ASm1wahQSLsysqlY1o by realcaseyrollins@social.teci.world
2023-02-17T14:49:31.071686Z
0 likes, 0 repeats
I similarly find it odd that people think these chatbots should always be accurate. They should not be asked for factual information, nor should anyone rely on them for factual information.Such is the conundrum for including them in something like #Google or #Bing. Including it in something like #Alexa or #GoogleAssistant, or even #Siri, as “someone to talk to” would be able better application imho
(DIR) Post #ASm2YXeyCp26SciJNY by hungrybluedev@mstdn.social
2023-02-17T14:56:04Z
0 likes, 0 repeats
@stux ChatGPT and other Large Language Models (LLMs) are designed to produce expected, bland, low-noise content. There's no novelty in the text, no new ideas. Only existing words moved around.If you are "surprised" by the content of something, it's probably original content. High-entropy, low predictability.So LLMs take the question, put the words in a black box function/machine and generate "transformed" outputs. A good original answer adds details, ideas, and connects them properly.
(DIR) Post #ASm8F4oFqUVTCq9qrY by ranx@mastodon.social
2023-02-17T15:59:56Z
0 likes, 0 repeats
@stux it's probably the same people who think crypto is a reliable investment class 🤭
(DIR) Post #ASmCl7qOUy0bhVhEHo by thibaultmol@en.osm.town
2023-02-17T16:50:31Z
0 likes, 0 repeats
@stux atleast with new bing you can check sources (to verify it's output is actually correct)#chatgpt