Post AVmrRTfIVKKjiwoGOW by sumomo@mastodon.sdf.org
(DIR) More posts by sumomo@mastodon.sdf.org
(DIR) Post #AVmr1zJ0F9gwKGE9Hk by ParadeGrotesque@mastodon.sdf.org
2023-05-18T17:58:15Z
0 likes, 0 repeats
ChatGPT and LLMs are the perfect verification of the old saying:Garbage In Garbage Out.Questions I have: 1. Do we know why, exactly, ChatGPT goes into a state of fugue and starts spewing out garbage?2. Is it possible to even know, statistically, if a given ChatGPT answer is wrong or right within a reasonable confidence level?I suspect the answer to these is 'no'.Finally, given the pronounciation of ChatGPT in French, we should use the following emojis: 😹 💨 Boost if you agree.
(DIR) Post #AVmrKVWVCIbxSZN8i0 by ParadeGrotesque@mastodon.sdf.org
2023-05-18T18:01:34Z
0 likes, 0 repeats
All of this to say I am not too worried about losing my job to an LLM.This, on the other hand, really worries me: "Hippocratic is building a large language model for healthcare"https://techcrunch.com/2023/05/16/hippocratic-is-building-a-large-language-model-for-healthcare/An LLM for ... heathcare? What could possibly go wrong? 🤷
(DIR) Post #AVmrRTfIVKKjiwoGOW by sumomo@mastodon.sdf.org
2023-05-18T18:02:49Z
0 likes, 0 repeats
@ParadeGrotesque as a native French speaker, I totally agree with this 😂
(DIR) Post #AVmrT5MdpEpQw2qmCO by KashifShah@mastodon.sdf.org
2023-05-18T18:03:08Z
0 likes, 0 repeats
@ParadeGrotesque check this out: https://writings.stephenwolfram.com/2023/03/chatgpt-gets-its-wolfram-superpowers/
(DIR) Post #AVmrr3ePqa0AZUaxDE by ParadeGrotesque@mastodon.sdf.org
2023-05-18T18:07:27Z
0 likes, 0 repeats
@KashifShah Yup, read about it before.This being said, from the linked article:"If you actually try these examples, don’t be surprised if they work differently [...] Since ChatGPT uses randomness in generating its responses, different things can happen even when you ask it the exact same question [...] But different from the solid “right-answer-and-it-doesn’t-change-if-you-ask-it-again” experience that one gets in Wolfram|Alpha and Wolfram Language."What could possibly go wrong? 🤷
(DIR) Post #AVmwI7VC2HoS9RROWe by robdaemon@tech.lgbt
2023-05-18T18:05:01Z
0 likes, 0 repeats
@ParadeGrotesque I'm okay with the idea of using datasets to *aid* in diagnosisbut that's where I draw the line
(DIR) Post #AVnDtoTMvwb1ME9ADQ by ecsd@commons.whatiwanttoknow.org
2023-05-18T20:42:13Z
0 likes, 0 repeats
@ParadeGrotesque <1. Do we know why, exactly, ChatGPT goes into a state of fugue and starts spewing out garbage?>Have you watched TV or movies in the past 50 years?Don't you know that /all plots/ now require Deus Ex Machina, that a "trusted and beloved member of the crew (or humanity)" must, for NO REASON AT ALL, turn against her, his or its fellows? CGPT was fed all TV plots, and the script for "2001: A Space Odyssey". Ask CGPT and find out how strongly it admires -- Hal.