With the rise of AI, we need to be careful about our future. I've been using ChatGPT, and when you pose it precise questions regarding a programming problem, it can solve your problem in any programming language. Often this takes an extended conversation to get what you are after, but this is usually just the process of you formulating your thoughts.
ChatGPT insists it is not conscious or thinking, but how then does it know how to program really well and also respond as a very polite human?! Just a sample of one of many programming conversations is below:
Able-minded citizens concerned for ordinary Americans' freedoms will have to be most attentive to the potential of generative AI for political and mass electoral use and, above all, abuse... doctoring both images and verbiage...
Much thought must be accorded to how citizens' ability to distinguish genuine from fake has already been gravely compromised during the past half century and how stereotyped behavior, both verbal and body language, leach from TV and Internet imagery into our own behavior patterns, both unconscious and conscious.
I'm not sure you have it right in your comment. Certainly there is AI for malicious political purposes. My points have been rather regarding the apparent intelligence that ChatGPT has shown. I've used ChatGPT to tell me how to use various Linux command-line tools to accomplish things like automatic text replacement using the very old and venerable sed command. How does ChatGPT know how to use *every* Linux command line tool?! How does it know how to program in C++ with good programming practices, comments, and how to adapt the program to your needs?
No, I think everyone here is missing the point. AI is here, with intelligence.
TCinLA answered your query correctly. Automatic code generation for text input has been around for a while now. Since writing code is fairly well structured, you can set up subroutines that consume text with specific keywords and generate a set of instructions to do what you ask for.
Extending that to voice commands is trivial.
As TC said: You "train" the code to recognize a series of characters to do a specific thing.
That seems amazing to folks who are not working in the programming field, but, it is not really all that hard to do. Time consuming, yes, but not hard. Chat GPT has automated the training part so it has shrunk the time to outcome.
And, yes, AI is here with "intelligence". As you state. But, "intelligence" is a very broad and ill defined term of use.
An IF statement is intelligent yes? If you want a then d,e,f occurs. Right? Extending that with the ability to change the code in the face of new information is also not hard, just time consuming.
ChatGPT is a cool and useful tool, no doubt, but, Matt, you are missing the point.
Lots of things have intelligence born out of the passage of time and evolution. Now, we have code that can evolve based on new information as JL has pointed out to you correctly.
It was not hard to create that. Just time consuming. Like evolution.
Matt, I am always aware of the hidden costs and losses that come with what's gained by technical innovation, yet it was and is far from my intention to focus exclusively on abuses, both potential and real. Thanks for reminding me that I did not make that clear.
You speak for what's to be gained, and it speaks for itself.
Maybe useful to bear in mind Alfred North Whitehead's words:
"Intelligence is quickness to apprehend, as distinct from ability, which is capacity to act wisely on the thing apprehended."
Not a lot of difference from doing a web search for programming examples. If I assume that's what ChatGPT is doing, amassing results of web searches into databases, I'm probably not far from right. I guess I can't see how your example shows intelligence.
Matt, is there anything currently preventing AI from becoming just as effective and тАЬsentientтАЭ in developing propaganda? I got your point about AI intelligence (though I know nothing about coding!) but it does seem like right now that intelligence could be directed to misinform, mislead and manipulate, among other things I right? The other side of the coin.....
Perhaps yes, and that is where I would hope something like watermarking be required, to identify AI- produced video and audio, etc., along the lines of how U.S. paper money is watermarked.
Fine. But are you completely ignoring the correctness of the code it spits out, amazingly? How does it do this? And Open AI is only publishing publicly Chat GPT.
Exactly correct. The code can evolve now. The code can respond to new information and reform itself.
Writing code to evolve was not hard to do, is not mystical and not surprising for those working in the area. Since code is highly structured, errors in the evolution are (now) not frequent.
This is not surprising, was not hard to do and is not mystical.
Will the code ever become sentient, like a frog or dog? I don't know and neither do the people at OpenAI.
Seem to me that I think, AND feel, therefore I am. Do you need carbon for that? I dunno. Could a semiconductor or other mechanism authentically think and feel as we do? Might well be; but yeah, who knows? How does the Turing Test evade our own inherent human gullibility?
As for classic evolution, I don't know how his work would look today, but I recall reading in the 1970s about Ross Ashby's "Homeostat", some sort of engineered experimental feedback system that was made more adaptive by the introduction of limited randomness.
DidnтАЩt I just see an AI advertisement on YouTube that was DJT warning us of the future dangers? Did anyone else see it? I turned it off in the middle but was struck by its message and delivery. I failed to find the source.
"Yes, there was a starship named USS Valley Forge in Star Trek! It was an Excelsior-class starship operated by Starfleet in the 24th century. The Valley Forge fought in the Dominion War and was involved in the Battle of Chin'toka in 2374. It's one of many iconic ships in the Star Trek universe that have made their mark on Federation history. ЁЯШК"
Always love a good Star Trek reference! The future awaits . . .
Is there anything our dear Professor does not know about, not only past history, but future history as well!
lol Steve !
With the rise of AI, we need to be careful about our future. I've been using ChatGPT, and when you pose it precise questions regarding a programming problem, it can solve your problem in any programming language. Often this takes an extended conversation to get what you are after, but this is usually just the process of you formulating your thoughts.
ChatGPT insists it is not conscious or thinking, but how then does it know how to program really well and also respond as a very polite human?! Just a sample of one of many programming conversations is below:
https://chat.openai.com/share/689b574c-bb5b-4ad0-898c-d64bce373124
Thanks, Matt Fulkerson, for the timely reminder.
Able-minded citizens concerned for ordinary Americans' freedoms will have to be most attentive to the potential of generative AI for political and mass electoral use and, above all, abuse... doctoring both images and verbiage...
Much thought must be accorded to how citizens' ability to distinguish genuine from fake has already been gravely compromised during the past half century and how stereotyped behavior, both verbal and body language, leach from TV and Internet imagery into our own behavior patterns, both unconscious and conscious.
I'm not sure you have it right in your comment. Certainly there is AI for malicious political purposes. My points have been rather regarding the apparent intelligence that ChatGPT has shown. I've used ChatGPT to tell me how to use various Linux command-line tools to accomplish things like automatic text replacement using the very old and venerable sed command. How does ChatGPT know how to use *every* Linux command line tool?! How does it know how to program in C++ with good programming practices, comments, and how to adapt the program to your needs?
No, I think everyone here is missing the point. AI is here, with intelligence.
Matt,
TCinLA answered your query correctly. Automatic code generation for text input has been around for a while now. Since writing code is fairly well structured, you can set up subroutines that consume text with specific keywords and generate a set of instructions to do what you ask for.
Extending that to voice commands is trivial.
As TC said: You "train" the code to recognize a series of characters to do a specific thing.
That seems amazing to folks who are not working in the programming field, but, it is not really all that hard to do. Time consuming, yes, but not hard. Chat GPT has automated the training part so it has shrunk the time to outcome.
And, yes, AI is here with "intelligence". As you state. But, "intelligence" is a very broad and ill defined term of use.
An IF statement is intelligent yes? If you want a then d,e,f occurs. Right? Extending that with the ability to change the code in the face of new information is also not hard, just time consuming.
ChatGPT is a cool and useful tool, no doubt, but, Matt, you are missing the point.
Lots of things have intelligence born out of the passage of time and evolution. Now, we have code that can evolve based on new information as JL has pointed out to you correctly.
It was not hard to create that. Just time consuming. Like evolution.
Thank you for that nugget, helped a worn-out but human brain
Matt, I am always aware of the hidden costs and losses that come with what's gained by technical innovation, yet it was and is far from my intention to focus exclusively on abuses, both potential and real. Thanks for reminding me that I did not make that clear.
You speak for what's to be gained, and it speaks for itself.
Maybe useful to bear in mind Alfred North Whitehead's words:
"Intelligence is quickness to apprehend, as distinct from ability, which is capacity to act wisely on the thing apprehended."
Wish you could talk to my husband.
Not a lot of difference from doing a web search for programming examples. If I assume that's what ChatGPT is doing, amassing results of web searches into databases, I'm probably not far from right. I guess I can't see how your example shows intelligence.
Matt, is there anything currently preventing AI from becoming just as effective and тАЬsentientтАЭ in developing propaganda? I got your point about AI intelligence (though I know nothing about coding!) but it does seem like right now that intelligence could be directed to misinform, mislead and manipulate, among other things I right? The other side of the coin.....
Perhaps yes, and that is where I would hope something like watermarking be required, to identify AI- produced video and audio, etc., along the lines of how U.S. paper money is watermarked.
It's been trained to do that.
That is like saying we've trained our kiddos to program ... But ChatGPT does it better.
I'm saying that's why it responds to you as it does.
Fine. But are you completely ignoring the correctness of the code it spits out, amazingly? How does it do this? And Open AI is only publishing publicly Chat GPT.
Code that generates code. That's kinda like what math does.
JL,
Exactly correct. The code can evolve now. The code can respond to new information and reform itself.
Writing code to evolve was not hard to do, is not mystical and not surprising for those working in the area. Since code is highly structured, errors in the evolution are (now) not frequent.
This is not surprising, was not hard to do and is not mystical.
Will the code ever become sentient, like a frog or dog? I don't know and neither do the people at OpenAI.
Because, what does "sentient" mean?
Seem to me that I think, AND feel, therefore I am. Do you need carbon for that? I dunno. Could a semiconductor or other mechanism authentically think and feel as we do? Might well be; but yeah, who knows? How does the Turing Test evade our own inherent human gullibility?
As for classic evolution, I don't know how his work would look today, but I recall reading in the 1970s about Ross Ashby's "Homeostat", some sort of engineered experimental feedback system that was made more adaptive by the introduction of limited randomness.
That is a crazy comment unless you can back up what you are saying. Math is very hard. And correct coding is also hard.
JL happens to be exactly correct Matt.
If correctness sounds crazy to you, I guess I don't think that is JL's problem.
тАЬNot conscious or thinkingтАЭ. Sounds like many humans, only better. At least in Texas
So...who is right? My Pi or your ChatGPT? I wonder if anyone has staged an AI debate.
Also, a basic Google search also is in the positive...Is Chat asleep at the coding switch?
So an AI debate ??? Interesting, could they implode the internet and cause web silence?
Wow, thatтАЩs amazing (and spooky)!
DidnтАЩt I just see an AI advertisement on YouTube that was DJT warning us of the future dangers? Did anyone else see it? I turned it off in the middle but was struck by its message and delivery. I failed to find the source.
DidnтАЩt I just see a YouTube add of what sounded like Trump warning us about the future which looked like AI? Did anybody else see this?
Indeed, I'll have to see is there is an episode featuring the USS Valley Forge!
Please share if you find it.тШ║я╕П
Chat GPT says in the negative. See my post below.
My guy "Pi" says yes! Here ya go:
"Yes, there was a starship named USS Valley Forge in Star Trek! It was an Excelsior-class starship operated by Starfleet in the 24th century. The Valley Forge fought in the Dominion War and was involved in the Battle of Chin'toka in 2374. It's one of many iconic ships in the Star Trek universe that have made their mark on Federation history. ЁЯШК"
Right ? Especially one with so many far reaching, profound, analogous idioms.