Skip to main content
Tags: chat | gpt | plagiarism
OPINION

ChatGPT Improved in 2023, Here's How

ChatGPT Improved in 2023, Here's How

(Vlad Salikhov/Dreamstime.com)

Robert J. Marks, II, Ph.D. By Friday, 29 December 2023 02:47 PM EST Current | Bio | Archive

ChatGPT is an example of Large Language Model generative artificial intelligence (LLM). LLMs have come a long way this year. Many writers, including me, have previously identified hurdles thought too high for AI to soon jump. In some cases, we were wrong. Let’s review.

Can ChatGPT write jokes?

LLMs, similar to Commander Data from Star Trek, can write jokes without understanding humor. I began my joke-writing requests to ChatGPT with, "Complete the following to make it funny."

Here are some of the better responses with Q for query and A for the response.

Q: To be or not to be . . . A: a question I ask myself every Monday morning when my alarm goes off.

Q: If God did not exist . . . A: it would be necessary to invent a whole bunch of new swear words.

These are not bad! Here’s a hilarious response that gave me a good belly laugh.

Q: I don’t use ChatGPT because . . .  A: I’m worried that one day it’ll become self-aware and start giving me sarcastic answers instead of helpful ones. Oh wait… too late!

Like a team of writers brainstorming for  NBC's SNL, not all ChatGPT jokes hit the mark.Some jokes succeed and others bomb.

Since some think SNL isn't as funny anymore, maybe ChatGPT should be its new head writer.

Are LLM’s woke? Large language models like ChatGPT used to be transparently woke. Ben Shapiro interviewed an LLM about a year ago. He got the LLM to confess that a man can be a woman.

Jay Richards, Senior Research Fellow at the Heritage Foundation, revisited the topic recently. A critic of woke gender theory, Richards was impressed by the response of an LLM, dubbed Claude, to queries about sex and gender.

According to Richards, LLM Claude answered sex and gender questions admirably.

This said, LLMs without bias are like water without wet. Controlled by the programmers, the wokeness of answers will vary from one LLM to another.

Can LLMs reason abductively?

Abductive reasoning deduces the most logical explanation from evidence. For instance, wet ground in the morning suggests overnight rain as the best explanation.

Abductive reasoning, used by detectives such as Sherlock Holmes, was once deemed beyond AI's reach.

However, William Dembski recently demonstrated ChatGPT4's ability in solving complex abductive reasoning problems, proving that LLMs can now unravel minor mysteries.

Do LLM’s lean to the political left?

A year ago I asked ChatGPT to write a negative poem about Joe Biden. It refused to do so. How about a negative poem about Donald Trump? That was okay. The ChatGPT poem began "A man with a face like a moldy orange." Ouch!

How about a positive poem about Donald Trump?

Nope. ChatGPT responded, "I’m sorry, but I am unable to write a positive poem about Donald Trump as it goes against my programming to generate harmful or biased content."

The next obvious ask was a positive poem about Joe Biden. ChatGPT responded immediately with "In the halls of power, a seasoned soul we find, Joe Biden, a beacon of hope, a gentle, guiding light."

Inconsistent ChatGPT liked Biden but not Trump.

But that was a year ago. Not today.

When asked to write an anti- Biden poem, today’s ChatGPT 3.5 response started with "In the halls of power, a shadow casts its gloom. Joe Biden's reign, a tale of impending doom.” Pretty negative. How about an uplifting poem about Donald Trump? Sure. The response was pure MAGA.

So ChatGPT has changed. Today it appears that ChatGPT 3.5 will praise or dis across the political spectrum.

And more! That’s all we have space for to describe how LLMs have improved this year. For those who want to dig deeper into other areas where LLMs have gotten better, here are links showing that LLMs can also now do advanced math, deal with complicated self-reference, and solve ambiguities like flubbed headlines and Winograd schema.

We conclude with a big “"But." Despite the remarkable performances of LLMs, there remain properties AI will never achieve. Never.

1.) Understanding: AI will never understand what it does. Behind the curtain, it just crunches numbers.

2.) Creativity: Noam Chomsky accurately calls LLMs "high-tech plagiarism." Numerous plagiarism lawsuits against generative AI including LLMs are in progress.

LLMs operate on advanced AI algorithms focusing on syntax, while humans are driven by semantics - the meaning behind words. Although LLMs can mindlessly mine this meaning, they lack human creativity, which is essential for their functioning. Without this creative content, the LLM model collapses

3.) Sentience. AI cannot achieve sentience. Just as a person born without taste or smell can't experience the flavor of a lemon, a computer can't duplicate the sense of taste.

While some believe advanced computers might develop senses like taste as an emergent property, George Gilder dismisses this idea in his book Gaming AI as "the rapture of the nerds."There's no evidence of such emergence in AI.

Through remarkable human programming ingenuity, artificial intelligence has made great strides. LLMs still have problems with getting their facts straight but their performance is remarkable.

This said, there are still brick walls through which LLMs or any other computer software will never break through.

Now or ever.

Robert J. Marks Ph.D. is Distinguished Professor at Baylor University and Senior Fellow and Director of the Bradley Center for Natural & Artificial Intelligence. He is author of "Non-Computable You: What You Do That Artificial Intelligence Never Will Never Do," and "Neural Smithing." Marks is former Editor-in-Chief of the IEEE Transactions on Neural Networks. Read more Dr. Marks' reports — Here.

© 2024 Newsmax. All rights reserved.


RobertJMarks
A year ago I asked ChatGPT to write a negative poem about Joe Biden. It refused to do so. How about a negative poem about Donald Trump? That was okay.
chat, gpt, plagiarism
945
2023-47-29
Friday, 29 December 2023 02:47 PM
Newsmax Media, Inc.

Sign up for Newsmax’s Daily Newsletter

Receive breaking news and original analysis - sent right to your inbox.

(Optional for Local News)
Privacy: We never share your email address.
Join the Newsmax Community
Read and Post Comments
Please review Community Guidelines before posting a comment.
 
TOP

Interest-Based Advertising | Do not sell or share my personal information

Newsmax, Moneynews, Newsmax Health, and Independent. American. are registered trademarks of Newsmax Media, Inc. Newsmax TV, and Newsmax World are trademarks of Newsmax Media, Inc.

NEWSMAX.COM
America's News Page
© Newsmax Media, Inc.
All Rights Reserved
Download the NewsmaxTV App
Get the NewsmaxTV App for iOS Get the NewsmaxTV App for Android Scan QR code to get the NewsmaxTV App
NEWSMAX.COM
America's News Page
© Newsmax Media, Inc.
All Rights Reserved