[ExI] Language models are like mirrors

[spike at rainier66.com]

From: Gordon Swobe <gordon.swobe at gmail.com> 
Sent: Sunday, 2 April, 2023 8:17 PM
To: ExI chat list <extropy-chat at lists.extropy.org>
Cc: spike at rainier66.com
Subject: Re: [ExI] Language models are like mirrors

 

 

On Sun, Apr 2, 2023 at 8:36 PM spike jones via extropy-chat <extropy-chat at lists.extropy.org <mailto:extropy-chat at lists.extropy.org> > wrote:

 

To achieve sentience, it will need to be able to learn, and not just during
the training period.  It will also need access to tools, like math
calculators. 


Clarification: the above was a quote from a long time friend who was my patient mentor in college when we were in the computer room 43 years ago (oh mercy.)  I helped him in chemistry, he helped me in computers.  We collaborated on gags a lot using that computer.  When it comes to software, he doesn’t suck.

 

>…I learned from watching an interview of Sam Altman of OpenAI/ChatGPT fame that he approves of his co-founder Ilya Sutskever's test for consciousness in an AI language model. I doubt anyone will ever attempt the experiment, but I also think it makes sense at least in principle…-gts 

 

 

It is really cool that we are struggling to devise tests to decide whether ChatGPT is sentient.  Clearly we have had to refine the definition a number of times, and really deal with unexpected nuance.  >From what I understand from reading the debate here in the past few weeks, it isn’t as simple as sentient or not sentient.  There are subtle shades of meaning and levels of sentience which apply as much to AI as it does to BI.

 

Eliza did tricks which could be understood if one looked at the code, then carried out the instructions oneself.  If we did that for ChatGPT, we might have a different view of its sentience and understanding.  However… OpenAI will not show us that code.  I understand why of course: they are sitting on a gold mine.  People will pay for ChatGPT.  I dang sure will.  I will pay a lot more if they offer me a version which I can train or correct its “thinking.”

 

spike

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.extropy.org/pipermail/extropy-chat/attachments/20230402/950c4459/attachment.htm>