Re: [FRIAM] ChatGPT and William James

2023-03-07 Thread Marcus Daniels
So it ought to do well with education curricula? Sent from my iPhone On Mar 7, 2023, at 2:48 PM, Prof David West wrote:  I am sure that none of the respectable members of this list will have encountered this, but Jochen's comment: "it is additionally trained extensively how to respond corre

Re: [FRIAM] ChatGPT and William James

2023-03-07 Thread Jochen Fromm
No, this is not what I had in mind. I stumbled upon this MIT Technology Review article which mentioned Reinforcement Learning from Human Feedback (RLHF). It made me wonder if large language models behave so much like humans because they are trained to do it over and over again. First by using a

Re: [FRIAM] ChatGPT and William James

2023-03-07 Thread Prof David West
I am sure that none of the respectable members of this list will have encountered this, but Jochen's comment: *"it is additionally trained extensively how to respond ***_correctly_*** by humans" *(emphasis mine of course) means I cannot resist sharing. I just read an amusing ChatGPT conversat

Re: [FRIAM] ChatGPT and William James

2023-03-07 Thread Steve Smith
This is such a great example of Glen's assertion (probably a misquote) that "Communication doesn't exist".   We are all (mostly?) talking past one another with different assumptions and definitions?   I can use the term "input" with nouns but they only make sense with "nouns which represent pro

Re: [FRIAM] ChatGPT and William James

2023-03-07 Thread Jochen Fromm
ChatGPT apparently uses a technique called "Reinforcement Learning from Human Feedback" (RLHF). It is not only based on huge amounts of training data humans have created in form of Wikipedia entries, web pages and books, it is additionally trained extensively how to respond correctly by humans.

Re: [FRIAM] ChatGPT and William James

2023-03-07 Thread Marcus Daniels
“self modifying” is functionally the same as creating a variant and running it while destroying the old version. Gene regulation, say. > On Mar 7, 2023, at 7:18 AM, glen wrote: > > Why does the "agent" have to be the same across the transition from gametes > to zygote? Historical transition

Re: [FRIAM] ChatGPT and William James

2023-03-07 Thread glen
Why does the "agent" have to be the same across the transition from gametes to zygote? Historical transitions exist. Genes are input to humans in the same way, for example, a traumatic injury at age 10 is input to the human at age 20. Theseus' ship, anyone? I wrote and deleted a long post abou

Re: [FRIAM] [off topic] exterminators for mice?

2023-03-07 Thread Jochen Fromm
You need a cat! My parents always used to have a cat in their house. Now that I live in Berlin a cat or dog is impractical, since we live in a small rented apartment and have no garden, but if you have a house and a garden and a mouse problem then a cat might help :-)https://www.offthemark.com/

Re: [FRIAM] ChatGPT and William James

2023-03-07 Thread Santafe
Good concepts in which to express this would seem to me to be the problem of statistical learning of some “data”, and the choice of how the “data” are “represented”. All terms that have to be given meaning operationally in some problem or set of problems that we say are similarly structured. I