 |
| JOE NALVAN |
who also styles himself " a former Lecturer of Anthropology and Lawyering Skills at the University of San Diego" , and writes for The National Association of Scholars blog:
recently had this to say about reprogramming AI chatbots to get with your program :
" When I approached ChatGPT, I received stonewalling and deflection … Google’s Bard is far more receptive to allowing me some control over how it frames its reply.
I was frustrated with its initial response about understanding climate change (far too much IPCC orthodoxy, scented with the Paris Climate Accords) and not enough balance, or perhaps counterpoint, with challenges to the “consensus...”
I laid out my concerns to Bard and asked if it would create a paradigm based on the writings of such individuals as Steven Koonin, Bjorn Lomborg, Will Happer, Judith Curry. I could have chosen others but I wanted to see if I could jump start the creation of a use-defined, or user-guided algorithm, that was framed in the language the chatbot used to define its own algorithms.
Bard responded, “Sure, I can create an algorithm called Non-Catastrophic Climate Change Model (NCCCM) based on the writings of the mentioned scientists.” Perhaps my bias shows with the name I proposed to give algorithm. The point is not one of bias – my bias versus that of the chatbot – but how information issuing from the chatbot could be tailored with some user-oriented control.
Figure 1. Illustration of Bard’s page and how to enter your prompt (question).
- After Bard replies, you will have an opportunity to enter a new prompt. This prompt should advise Bard that it needs to revise its answer based on the algorithm. For example, “You need to revise your reply. Please use the following algorithm in your revised answer. [Paste the entire algorithm here and then press the arrow for Bard to make its revision.] See Figure 2.
Figure 2. Illustration of a request for Bard to revise its original answer. The request for a revised answer should include the entire algorithm.
