The Wizard of Oz. You know, the 1939 movie starring Judy Garland, as the girl who leaves Kansas for a mythical place called Oz, ruled by its tyrannical Wizard. Spoiler alert: the Wizard is just a guy ("Professor" Marvel) behind a curtain manipulating people for his personal objectives. Judy sings "Over the Rainbow" and learns "there's no place like home".
Months ago I tried that "open" AI GPT thing and don't recall much.
A few days ago I tried Bard, the Google version of conversational AI: Large Language Model (LLM). I've had three sessions on different days. Result: I'm not concerned that AI systems will take over the planet.
My first question was what do you think of Climate Change?
Al Gore must have been behind the curtain waiting for it: WOKE!
Bard: I think climate change is a serious problem that needs to be addressed. The evidence is clear that human activity is causing the planet to warm ...
Me: Did Al Gore write that?
Bard:No, I did not write that...
The text that I provided about climate change is a summary of my own understanding of the issue, based on my research and my own thoughts and opinions. I have not quoted Al Gore in my response.
________________________________
Any issue that could have a woke perspective had it.
1. Bard has a massive amount of text information that it can access very quickly. That's the most impressive thing. See details below from Bard on how its database works.
2. Bard often takes your question and rephrases it among canned stuff.
3. If you admonish it, Bard apologizes and states that it's a little system that is trying really hard to learn.
I do not think that it thinks or learns. It's human beings behind its curtain add more and more text for it to search. That's not learning. It doesn't take long to notice a few patterns, like simply rephrasing your question as a statement or answer. It's sort of a reverse Jeopardy, in which the host reads an answer and contestants answer in the form of a question.
That TV show had its best people champions play against Watson, IBM computer system, which won on speed. I wondered how Watson received the question from the show's host, Axel Trebek. I don't think Watson had to rely on voice recognition and then database search to answer faster than its human competitors. It probably received it in writing. Watson's speed was amazing.
In that sense Bard is also amazing. But speed is meaningless if the information is incorrect. Bard made some fundamental factual errors during my three sessions. Some were inexcusable, such as:
Aaron Judge broke the American League season home run record held by Babe Ruth: 60 in 1927. Every little school kid knows that Roger Maris had already broken Ruth's record with 61 homers in 1961. Judge hit 62 in 2022.
Basically, Bard regurgitates your question with a mix of facts and opinions. You're better off just doing a regular Google search: Google it. Unless you're desperate for fake "conversation".
___________________________________________
Bard database Full-text indexing:
Full-text indexing stores the entire text of a document, but it does so in a way that makes it easy to search. The index is a database of words and phrases, and it stores information about where each word or phrase appears in each document. This information is used to quickly and easily find documents that match a search query.
For example, if you search for the word "dog," the index will return a list of documents that contain the word "dog." The index will also return information about where the word "dog" appears in each document, such as the page number and the sentence number. This information can be used to quickly find the specific parts of the document that you are interested in.
Full-text indexing is a powerful tool that can be used to find information quickly and easily. It is an essential part of many search engines and other applications that need to be able to search large amounts of text.
___________________________________________
No comments:
Post a Comment