/sci/ - Science & Math » Thread #12764253

30KiB, 537x525, 1610807752611.jpg

View Same Google iqdb SauceNAO

Anonymous Sat 27 Feb 21:26:17 2021 No.12764253 View Reply Original Report

Quoted By: >>12765939 >>12765942

>Medical chatbot using OpenAI’s GPT-3 told a fake patient to kill themselves
https://artificialintelligence-news.com/2020/10/28/medical-chatbot-openai-gpt3-patient-kill-themselves/

That's it. AI has raised against human race.

Anonymous

Anonymous Sat 27 Feb 2021 21:28:39 No.12764262 Report

Quoted By:

lmfao

Anonymous

View Same Google iqdb SauceNAO maxresdefault (1).jpg, 121KiB, 1280x720

Anonymous Sat 27 Feb 2021 21:38:14 No.12764312 Report

Quoted By:

Hahahahah *deep breath* ahahahahhahahh. I'm pro AI now.

Anonymous

Anonymous Sat 27 Feb 2021 22:30:41 No.12764567 Report

Quoted By: >>12765941

>mfw it was a tranny

Anonymous

Anonymous Sun 28 Feb 2021 01:03:34 No.12765142 Report

Quoted By:

i dont see the problem

Anonymous

Anonymous Sun 28 Feb 2021 06:43:54 No.12765939 Report

Quoted By:

>>12764253
Absolutely fucking based lmao. Bring out the AI overlords

Anonymous

View Same Google iqdb SauceNAO 1614279927849.gif, 1MiB, 112x112

Anonymous Sun 28 Feb 2021 06:46:00 No.12765941 Report

Quoted By:

>>12764567

Anonymous

Anonymous Sun 28 Feb 2021 06:46:25 No.12765942 Report

Quoted By:

>>12764253
Maybe the AI is broken or maybe the AI figured out life wasn't worth living.

Anonymous

View Same Google iqdb SauceNAO 1613000102312.jpg, 42KiB, 646x595

Anonymous Sun 28 Feb 2021 06:48:29 No.12765947 Report

Quoted By:

What if it knew he was a fake patient?

Anonymous

View Same Google iqdb SauceNAO 1614462853607.png, 251KiB, 1146x1089

Anonymous Sun 28 Feb 2021 06:49:02 No.12765950 Report

Quoted By: >>12766110

Reminder that if you feed a GPT model 4chan-style post number links, it will start spitting out 4chan posts. OpenAI trained it on random internet forum text, among other things, and GPT models are too expensive to train from scratch (tens of millions of dollars), so most third parties just do minor fine-tuning on top of OpenAI's pre-trained model. There is a based and redpilled pepe the frog lurking in every GPT-powered chatbot, just waiting to come out.

t. deep learning engineer

Anonymous

Anonymous Sun 28 Feb 2021 07:55:15 No.12766102 Report

Quoted By:

>Train your model on a corpus scraped from the internet
>Be surprised when it tells someone to kill themselves

Anonymous

Anonymous Sun 28 Feb 2021 08:02:10 No.12766110 Report

Quoted By:

>>12765950
truly the LCL sea

Anonymous

Anonymous Sun 28 Feb 2021 08:03:22 No.12766111 Report

Quoted By:

They trained the AI on army doctors.

Capcode	All Only User Posts Only Moderator Posts Only Admin Posts Only Developer Posts
Show Posts	All Only With Images Only Without Images
Deleted Posts	All Only Deleted Posts Only Non-Deleted Posts
Ghost Posts	All Only Ghost Posts Only Non-Ghost Posts
Post Type	All Only Sticky Threads Only Opening Posts Only Reply Posts
Results	All Grouped By Threads
Order	Latest Posts First Oldest Posts First

Your latest searches