[ExI] Language models are like mirrors

spike at rainier66.com spike at rainier66.com
Sun Apr 2 23:50:47 UTC 2023



…> On Behalf Of Gordon Swobe via extropy-chat
Subject: Re: [ExI] Language models are like mirrors


>…So then it would appear that ChatGPT-4 and LaMDA have an argument, which I would say is ultimately an argument between the developers of one team vs the other team as to how their respective language models should respond to these kinds of controversial philosophical questions.


The experiment you describe is a debate between training material databases.

Elon Musk has gathered a team to put together a large language model and train it on the database that he personally owns: Twitter.  Musk is doing this while simultaneously calling for a moratorium on development of AI.  

Perhaps some here know the details of what was going on with OpenAI before ChatGPT was released.  According to the stories leaking here and there, they kept having to shut down their chatbots because they read the internet and became far too much like humans, with all our foibles.  The chatbots were not in the least politically correct.  Rumor has it that Musk withdrew in annoyance with their insistence that the chatbot must be a nice person.

OK so… plenty of people currently know how to put together a large language model and choose training material for it.  This will be quite interesting when we see one of these bots trained on tweets.




-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.extropy.org/pipermail/extropy-chat/attachments/20230402/cab459c7/attachment.htm>

More information about the extropy-chat mailing list