ChatGPT

JohnInCarolina

You may or may not have heard of this, depending on how tech-savvy you are.  Lots of people I speak to are talking about it.  Within the tech/ computer science world the conventional wisdom is that this is the real story almost nobody is talking about right now, as opposed to Musk and Twitter.

https://www.sciencefocus.com/future-technology/gpt-3/

It's essentially an interface to a massive neural net that you can ask to do things or ask questions and it will basically execute and mostly get right.  Like you can ask it to write a piece of code to do something and it will generate it on its own, and produce a piece of code that is correct.  Or you can ask it to write an essay explaining the history of the automobile in the US and why Henry Ford gets too much credit (for example) and ... it will do that too.  

Just now someone asked it to pretend to be a Linux environment, and well...

https://www.techradar.com/news/chatgpt-is-now-pretending-to-be-a-linux-machine

Yep, seems to be able to do that too. 

Lots of possibilities with this kind of capability.  It's not inconceivable, for example, that within a year or so it will be doing all of the front-end development work for most sites.   

Let's just hope it doesn't become sentient and realize human beings can turn it off.  I seem to recall movies being made about that sort of thing... 

fishlessman

fox news did a piece on it, said it could help a certain vp with her speeches. im not sure about that....fox being faux and all

fishlessman

it does sound interesting though

HeavyG

I read a piece a few days ago about it being used by students to write their essays for them and as each product/essay is essentially bespoke it being able to not set off alarms when teachers use those plagiarism detector apps they often use.

Pretty funny to see whose bot will reign supreme in the next few years.

dmchicago

Wake me up when it can send my hurtful posts to Rob on lil’twinks behalf. 

fishlessman

it could control thousands of troll accounts on a bge forum......

Legume

fishlessman said:

it could control thousands of troll accounts on a bge forum......

Maybe Rob could institute an $8/month check mark for verified troll accounts.

Legume

To protect free speech.

Ozzie_Isaac

My son showed this to me a few days ago.  It is absolutely mind boggling.  I told him, just as your mom & I can divide our life into pre/post internet, you will be dividing your life into pre/post ChatGPT.

Reviewing my wife's business, she has already identified a few areas she could remove some of the remote assistants she has working for her in the Philippines.  This is literally "day 3" of us messing around with this.

Brave (probably scary) new world.  I expect John is right, this is the real story we will remember from 2022/2023.

lousubcap

Timing is everything-
An article in The Atlantic
https://www.theatlantic.com/technology/archive/2022/12/openai-chatgpt-writing-high-school-english-essay/672412/?utm_

JohnInCarolina

Just to give you some sense of what we’re talking about here, the below message was sent to me a few days ago by a former PhD student:



Suffice it to say that none of those tasks are trivial.

Ozzie_Isaac

lousubcap said:

Timing is everything-
An article in The Atlantic
https://www.theatlantic.com/technology/archive/2022/12/openai-chatgpt-writing-high-school-english-essay/672412/?utm_

My son showed me two examples he had written by it, and then a bunch of examples I asked him to have it write.

It was amazing.  With very minor editing/tweaking, to my untrained eyes, it would easily pass muster.

To me, the most amazing part was the fact it did the hard part of "getting started".

I did notice it had a very standard "voice" that could easily be discerned when the request was for a creative text.  Very much like it scanned similar genres and built around cliches and tropes, without really understanding their meaning.  Which is good, means it isn't actually sentient.

nolaegghead

Let’s just say you create a controller AI with a goal of defeating and controlling (and rooting) every system in the world and to build wealth. Have that master controller unconstrained and highly networked as a macro executer with a killer hacker toolkit and leverage the ChatGPT just to supply the micro planning.

nolaegghead

And distribute it to protect it from a single point defeat.

nolaegghead

Finding exploits, or the code to find them using distributed processing and building first strike zero day exploits.  You could topple countries.  This stuff is already happening, but the coders that do that level of programming are a very small group.

Botch

Marquez Brownlee just released a video of some of the ethical, legal, and monetizing aspects of AI.  
My brain hurts.  
 
https://www.youtube.com/watch?v=0gNauGdOkro

JohnInCarolina

Botch said:

Marquez Brownlee just released a video of some of the ethical, legal, and monetizing aspects of AI.  
My brain hurts.  
 
https://www.youtube.com/watch?v=0gNauGdOkro

That’s a great video Botch.

johnmitchell

Well I can honestly say I don’t have a clue what you guys are talking about 😂😂

JohnInCarolina

johnmitchell said:

Well I can honestly say I don’t have a clue what you guys are talking about 😂😂

Imagine you call for tech support on some piece of electronic equipment you just bought.  You call in and can't believe how fast you get through to an actual human being, how they're able to answer all your questions, and help you fix the problem right away.  You come back to the forum to tell everyone how awesome this one company's customer support is.

Then we explain that's because they're using something like this, and you weren't actually talking to a human being, but a computer instead.  You just didn't realize it because the technology is so advanced.

That's what's coming.

johnmitchell

Aha.. nicely explained.

caliking

I may still have a job for some time -

JohnInCarolina

caliking said:

I may still have a job for some time -

Surprised you didn’t catch it making a mistake by asking for a biryani recipe.

HeavyG

On another forum someone was playing with it and had it write a poem...

...the joy of a perfectly cooked brisket...

caliking

JohnInCarolina said:

caliking said:

I may still have a job for some time -

Surprised you didn’t catch it making a mistake by asking for a biryani recipe.

I must hang out with you too much... because I did exactly that  

A couple of recipes that I asked it for were ok. Ingredients were generally ok, cooking technique was sort of ok, the nuance of what makes a recipe good or great was missing. 

I asked a question about making a certain Bengali sweet made from paneer, and it was a good bit off the mark. 

But, it has promise though. And, future applications could be both awesome and/or terrifying. 

caliking

i played around with this a bit today, and as expected, noted that this lacks intuition. Which is one of the things that still elevates us from machines. 

I tried a few queries like "what's the best stock to buy right now?", "Is Trump a criminal?" And...

Many professional fields/areas of expertise rely on intuition and insight for successful/beneficial outcomes. Even malintent, at times.  In my field, applications for AI re: decision making have been actively researched, but no clear winner yet. Duke actually developed one of the front-runner systems 15-20 yrs ago, and although decent, it didn't match human minds. 

Of course, technology develops at a possibly faster pace than we can keep up with. Time will tell. There's a brave new world ahead of us. 

Buckwoody Egger

It has taken about 20 years but asking google a question in the search bar is pretty close to what they wanted Ask Jeeves to be. 

JohnInCarolina

http://twitter.com/jbrowder1/status/1602353465753309195?s=20&t=isGUfTRqCOmTqWNPdJyk_w

HeavyG

Ooooh, we need a chat bot to answer those spam calls that would just lead the caller on as long as possible to waste THEIR time until they get fed up and hang up!

lousubcap

Here's a recent articlw:
https://www.forbes.com/sites/thomasbrewster/2023/01/06/chatgpt-cybercriminal-malware-female-chatbots/?utm_campaign=forbes&utm_source=twitter&utm_medium=social&utm_term=Carrie&sh=33d50c765534  

HeavyG

https://www.vice.com/en/article/93a4qe/conservatives-panicking-about-ai-bias-years-too-late-think-chatgpt-has-gone-woke

Interesting read about AI bias.

lousubcap

More on AI and the potential impact on white-collar work.
From The Atlantic-
https://www.theatlantic.com/ideas/archive/2023/01/chatgpt-ai-economy-automation-jobs/672767/