[ExI] Language models are like mirrors
spike at rainier66.com
spike at rainier66.com
Sat Apr 1 03:02:52 UTC 2023
From: Gordon Swobe <gordon.swobe at gmail.com>
>…The ELIZA effect is the tendency of humans to anthropomorphize computers that behave in human-like ways. I contend that those who attribute consciousness to ChatGPT are under the influence of the ELIZA effect. …-gts
Gordon, I was introduced to Eliza in college in 1978. It was a cool toy, but a bit of experimentation with it quickly revealed that the human user was supplying all the actual content. I interpreted it as a kind of parody, poking fun at psychologists and psych counselors, which were all the rage in those days among those who could afford them. I was nowhere near the economic status to hire a psychologist even had I felt the need for one, but at the time I opined that there is no point in hiring a feller to do what this software could do free.
It was an illusion of course. The user was supplying all the content. So really, all the human was doing was a slight one-off from talking to oneself, and not at all far off. Eliza was incapable of producing novel ideas, which is really what I wanted: an idea engine.
Then suddenly… 45 years go by. Now I am still looking for an idea engine. ChatGPT isn’t getting me there, however… it is a marvelous tool in any case. It is a GREAT teacher. It has taught me a lotta cool stuff. I have been using it to get up to speed on legal jargon and how the US court systems work. It’s really good at that. It is an excellent paralegal.
For instance, we are hearing a lot about Brady motions. I asked GPT if it is legal for a prosecutor in a grand jury setting to make Brady violations. Answer: depends. Most likely outcome is that they gets the indictment but the case is dead on arrival at the courtroom. If the prosecutor’s goal is just the indictment, then a Brady violation makes perfect sense for a prosecutor to do. ChatGPT taught me about that. Cool!
But… it doesn’t appear to me that ChatGPT is an idea engine. Great for some things. But not in creating new insights, the way human level intelligence does.
-------------- next part --------------
An HTML attachment was scrubbed...
More information about the extropy-chat