Fred Blauer
I finally got got my Microsoft Chatgpt/Bing/Edge browser all working together. The planets must be aligned. In my case it was a little more involved since I am using Ubuntu/Linux on the desktop. I didn't like the idea of having to replace the google chrome browser, but they are quite compatible and it hasn't been a problem for me. But I wanted to have a way to experiment with the Latest version of the chatbot without having to pay a monthly fee to OpenAI. The best way that I could find to do this was to use the one that Microsoft was offering, at least for now. Personally, I think that Google will eventually be incorporating a chatbot into their search as well, and there will be lots of others. I have to say that it is more convenient to have the AI answers along with the search, and I really like being able to ask the chatbot about the current site or document and have it understand the context.
If you follow any of my blogs you know that I have been researching/experimenting/learning AI for the last few months. I have been trying out different use cases for this kind of tech, such as generative images using LLM's, and things like assistive writing for content creation like blogging, social media etc. Anyways, the use case that I would like to talk about here is something that has been on my mind for a while. I have been wondering if chatbots like chatgpt would be good tools for fact checking. I feel that this is a very relevant topic since there is so much mis-information out there, which is being spread and amplified by social media like twitter and facebook for example. Misinformation is really a much more complex subject than I had originally thought, but I will confine my current blog post to relatively simple posts on social media. Here is a good article on how to spot Fake vs. Real news articles.
Lets say that someone you follow (or not) posts his opinion on a political subject of interest, or forwards a post that has been spreading around. Now, I really don't want to make this a political discussion, but I will mention that I am generally centrist, and can often agree or disagree with issues on both side of the spectrum. If the article is from a known source (such as a journal or newspaper), most people will assume there was some verification done by the author and won't bother doing further checking. They will agree or disagree and add their own comments based on their point of view. This seems to be the way that narratives get spread around Twitter for example. If something doesn't make sense to me, or I want to understand better, I can just ask the chatbot to check the facts, or give me further explanation. It can even summarize the articles or give you answers in context. To me, this is a huge step forward.
I will give you a recent example. The point is not to take sides, but point out that the beliefs that people have are based on what they have heard in the past, and may not be based on fact, or may even have changed, or may be partially true. There are lots of different possibilities. Here is the example: Someone was saying how the republican party is perceived as being the one that stands for less government or less spending and should be better for the budget.
So this is what I entered:
"Searching for: which US political party has a better record of lower deficits?
Generating answers for you…"
According to a study from the National Bureau of Economic Research, Democratic presidents since World War II have performed much better than Republicans. On average, Democratic presidents grew the economy by 4.4% each year versus 2.5% for Republicans. The study also found that the economy has performed substantially better under Democratic presidents. On average, real (inflation-adjusted) GDP has grown about 1.6 times faster under Democrats than under Republicans1. However, according to Forbes, on the economy and jobs, republicans have the better track record2. Since 1981, federal budget deficits have increased under Republican presidents Reagan, both Bushes and Trump, while deficits have declined under Democratic presidents Clinton and Obama3.
Everything is hyperlinked with citations and references. You can dig down as far as you want, and get reasons or further explanations. But the point I am trying to make, (I emphasise without getting political) is that the facts often go against generally accepted beliefs. And now we have a more convenient way for fact checking and learning.
Apparently, Elon Musks new AI company iemphsises going to be working on a language model called TruthGPT for this purpose. I am quite sure that gfiven the proper resources that this kind of tool can be built and incorporated into the moderation process. Personally, I think that it can be more objective and do a better job than humans, but it has to be open and transparent or it could be mis-used.
