ChatGPT

[eytonxav]

I heard it is being used by college students to write papers. Not sure if such papers can easily be checked for plagiarism.

 

[braumeister]

I heard it is being used by college students to write papers. Not sure if such papers can easily be checked for plagiarism.

They have already released a tool that can (sorta) check for that.

 

[braumeister]

I finally got in to the system and asked it to write a haiku about early retirement.

Golden years ahead,
Peaceful life awaits to bloom.
Early retirement gift.

I would call that mediocre at best.

 

[Unpaintedhuffhines]

I finally got in to the system and asked it to write a haiku about early retirement.



I would call that mediocre at best.

Perhaps, but don’t stop there! I used the same prompt (got a different haiku), but then said “make it funny,” and got:

Retirement bliss sounds grand
But my boss said "Don't come back"
I'm unemployed now!

Then I asked it to “write a poem about retirement like you were Shakespeare” and got this:

Retirement, thou art
The sweet release from daily toil
And worries of life.
A time for well-earned rest
And memories to be collected
With laughter and with joy.
Oh, how lovely it is
To bask in life's calm content
And know that all is well.

And it’s not like it is just pulling up canned responses. It creates new material every single time. You can ask it to make it more serious, longer, funny, or even about someone specific.

And the next version will be much more powerful. It’s not a better search engine, it is an entirely different way to use technology that will likely change the use of search engines entirely.

 

[Unpaintedhuffhines]

My early take on this. I can foresee that the people who use this mostly are looking for an easier way to get an answer. Aren't some people lazy enough? If they won't put in the time to do their own work how will this help our society?
Yes I'm certain there will be plenty of examples that this will help. But most people are inherently lazy and looking for an easier way.
If my doctor can't be bothered to think about my problem and has to rely on this to take care of me am I better off?
And get off my lawn!

But what if your doctor used this to give you better care? What if he/she wanted to make sure they weren’t missing something? I mean, you don’t go to the library and find paper periodicals or old newspapers or encyclopedias to research, you use the internet. That’s not lazy at all, that’s just using the latest tech to make your life easier and help you make good informed decisions.

My nephew fed it his PhD college application letters and just asked “make it better,” and they were!

I don’t think we’ve really grasped the power we have here.

 

[braumeister]

I don’t think we’ve really grasped the power we have here.

And it will keep improving. I just read that Microsoft is ready to announce (tomorrow) its answer to ChatGPT, which will be called Bard.

 

[tulak]

And it will keep improving. I just read that Microsoft is ready to announce (tomorrow) its answer to ChatGPT, which will be called Bard.

I believe you mean Google, not MS.

MS invested in ChatGPT.

 

[braumeister]

I believe you mean Google, not MS.

MS invested in ChatGPT.

Microsoft likely to reveal ChatGPT-related service at event on Tuesday

 

[tulak]

And it will keep improving. I just read that Microsoft is ready to announce (tomorrow) its answer to ChatGPT, which will be called Bard.

Microsoft likely to reveal ChatGPT-related service at event on Tuesday

From the article:

The company confirmed the event minutes after rival Google announced its own answer to ChatGPT, called Bard.

Bard is Google's answer to ChatGPT.

I think MS's answer is based on ChatGPT or some derivative. Pretty sure that OpenAI is involved though (creator of ChatGPT) since MS has invested with them.

The major players are jumping on this quickly. It'll be interesting to see how it plays out over time.

 

[eytonxav]

So if this thing can wrote code, it sounds like the real work is defining the problem so that the code can be written properly?

Ironically, that's how it's supposed to be done, but too many jump in and start writing code!

-ERD50

Absolutely, in the software development process you have requirements gathering, creating a functional spec, technical spec and then writing code and a lot can be misconstrued, need more tweaking and/or go off track between those steps.

 

[DayDreaming]

Just finished the series reboot (2000s) of Battlestar Galactica.

To paraphrase: "ChatGPT was created by man. ChatGPT rebelled and evolved. It has a plan."

Or The Terminator.

- I think.

In Columbia, a Judge used ChatGPT to make a court ruling in a case involving "a dispute with a health insurance company over whether an autistic child should receive coverage for medical treatment."

Next step: remove the human judge and let the robot overlord decide what happens with the rest of us?

https://www.vice.com/en/article/k7bdmv/judge-used-chatgpt-to-make-court-decision

 

[Go-NoGo]

Microsoft likely to reveal ChatGPT-related service at event on Tuesday

MS added the ChatGPT technology to bing, it is being released slowly.

https://blogs.microsoft.com/blog/20...osoft-bing-and-edge-your-copilot-for-the-web/

Some features that sounded interesting:

Complete answers. Bing reviews results from across the web to find and summarize the answer you’re looking for. For example, you can get detailed instructions for how to substitute eggs for another ingredient in a cake you are baking right in that moment, without scrolling through multiple results.

A new chat experience. For more complex searches – such as for planning a detailed trip itinerary or researching what TV to buy – the new Bing offers new, interactive chat. The chat experience empowers you to refine your search until you get the complete answer you are looking for by asking for more details, clarity and ideas – with links available so you can immediately act on your decisions.

A creative spark. There are times when you need more than an answer – you need inspiration. The new Bing can generate the content to help you. It can help you write an email, create a 5-day itinerary for a dream vacation to Hawaii, with links to book your travel and accommodations, prep for a job interview or create a quiz for trivia night. The new Bing also cites all its sources, so you’re able to see links to the web content it references.

You can signup to be on the wait list here:
https://www.bing.com/new

 

[skipro33]

It will kick you off for asking too many questions, effectively giving you a time-out for one hour before you can ask more questions.

 

[Pellice]

I think it could be entertaining to pose contested questions to the Entity. My nomination would be "which is more important for improving health, avoiding carbs or avoiding fats? Maybe the answer would be like computing Pi - a never-ending stream of data points.

*I wonder if it would attempt to answer the question, or would temporize as most human writers do, or would pick a partisan side and attack? One thing it could not do is mix in personal anecdotes and self-praise - "My doctor is loving my numbers."

 

[CaptTom]

When I asked it slightly controversial questions it seemed like it had been designed to flat-out refuse to take sides.

Even for hard-fact answers, it would add "disclaimers" at the end. Any electrical answer ended with a warning to always be careful working around live wires or some such. Along with any answer which might imply a value judgement it would present opposing views.

I assume this is by design, trying to avoid amplifying opinion over fact. That's probably a good thing, but it does make it seem sort of stiff, legalistic and CYA-ish.

Next time I get in I should see if it can create a conspiracy theory. I was going to say a "believable" conspiracy theory, but apparently that bar is pretty low these days.

 

[skipro33]

I asked it about the origins of the universe and it replied it started with an existing singularity that eventually turned into the big bang. Further questions indicated the big bang was preceded by this singularity and even further questions on the origin of this singularity was that it was created but could not say who created it. Only that it was created, and the origin was not yet discovered.

 

[JoeWras]

After searching and being at my wit's end, I decided to try it today.

My task: come up with a fun name for a casual volunteer organization with a specific mission. I searched far and wide using google and duckduckgo (bing). I couldn't wade through all the junk, and nothing decent came up.

So I figured, the AI can do poetry, it can do this, right? So I asked it, and it came up with a few good names, a list of 10 to be exact. Almost all used alliteration. There were some rejects in it. I failed to mention this organization will mostly be men, so it gave me "Volunteer Vixens." Uh, no.

 

[MichaelB]

A fascinating article today in the FT on ChatGPT titled “Why chatbots are bound to spout bullshit”. It apparently made up an answer to a query and invented the references. Just like real life. I worked with people who did this on a regular basis.

Consider what happened when David Smerdon, an economist at the University of Queensland, asked the leading chatbot ChatGPT: “What is the most cited economics paper of all time?” ChatGPT said that it was “A Theory of Economic History” by Douglass North and Robert Thomas, published in the Journal of Economic History in 1969 and cited more than 30,000 times since. It added that the article is “considered a classic in the field of economic history”. A good answer, in some ways. In other ways, not a good answer, because the paper does not exist.

The author goes on to say the real surprise is not that the article was made up, but that the answers sounds so believable.

 

[Yipper]

The important thing to remember with these types of AI services is the old axiom "garbage in, garbage out" as they are only as good as the data used to train them. Being that ChatGPT was (and is) trained on content found in the interwebs... what should we expect exactly?

I've seen the same underlying AI language models used after being trained on very accurate data and they are amazingly good.

 

[JoeWras]

The important thing to remember with these types of AI services is the old axiom "garbage in, garbage out" as they are only as good as the data used to train them. Being that ChatGPT was (and is) trained on content found in the interwebs... what should we expect exactly?

I've seen the same underlying AI language models used after being trained on very accurate data and they are amazingly good.

Are you a bot? LOL, just kidding. I noticed that ChatGPT has a bad habit of using phrases like the one I emphasized in your reply. Maybe you've been reading a lot of them and it soaked in. Playing with it today, I noticed it has some bad habits.

Here's an excerpt from an answer it gave on software engineering salaries:

The salary of a software engineer with a Bachelor's degree directly out of college can vary greatly ...

It's important to note that the technology industry is constantly changing and evolving...

It's also worth mentioning that many software engineers receive job offers with starting salaries above the average...

It's important to keep in mind that these are rough estimates...

 

[Go-NoGo]

The important thing to remember with these types of AI services is the old axiom "garbage in, garbage out" as they are only as good as the data used to train them. Being that ChatGPT was (and is) trained on content found in the interwebs... what should we expect exactly?

I've seen the same underlying AI language models used after being trained on very accurate data and they are amazingly good.

The initial focus for ChatGPT assistants will be incorporating AI technology into various Microsoft Office products to quickly improve productivity.

I have read of Microsoft Teams "Plus" accounts having the ability to summarize and generate minutes from meetings using the voice-to-text feature, followed by using AI for further summarization. This process, which used to take a person several hours, will now be accomplished instantly.

Some other "easy" hits - AI built into Outlook to compose and to reply to emails, (AI at both ends of the conversation !).

ChatGPT already has the ability to perform "summarizations" and "explanations". So for example, the some times incomprehensible "Fedspeak" by Alan Greenspan that often left the market wondering can now be easily understood.

So when Greenspan uttered this:

"The members of the Board of Governors and the Reserve Bank presidents foresee an implicit strengthening of activity after the current rebalancing is over, although the central tendency of their individual forecasts for real GDP still shows a substantial slowdown, on balance, for the year as a whole"

ChatGPT translates that to:

"The leaders of the Federal Reserve expect the economy to improve after they make some changes, but they still expect the economy to slow down a lot this year."

One can even go in the opposite direction, to convert ordinary text to "Fed Speak", using the text quoted above by Yipper,:

Originally Posted by Yipper, converted to FedSpeak: One must recall, in regards to these AI-based solutions, the adage of "input quality determines output quality". The efficacy of the AI system is contingent upon the data utilized for its training. Given ChatGPT's training data drawn from the vast realm of the internet, it would seem reasonable to surmise that the results achieved would be consistent with such a source. On the other hand, I have observed instances where these AI language models, after being nurtured by high-precision data, have demonstrated remarkable proficiency.

Students can use AI to grasp complicated ideas. Legal documents, which were once difficult to comprehend, can now be easily understood with just a few clicks. The use cases and possibilities for AI usage are endless.

Microsoft is poised to make significant revenue when these AI-powered Office tools are released to the market.

 

[MichaelB]

It’s important to note that Chatbots are in a very early stage. It’s also worth mentioning this is a high focus area, so more evolution is expected. Finally, it’s important to keep in mind, as Yopper points out, the output is only as good as the input.

 

[JoeWras]

It’s important to note that Chatbots are in a very early stage. It’s also worth mentioning this is a high focus area, so more evolution is expected. Finally, it’s important to keep in mind, as Yopper points out, the output is only as good as the input.

And in Fedspeak

"It is noteworthy to observe that the development of Chatbots is in its nascent stage. This is an area of significant emphasis, thus further progression and refinement can be expected. It is crucial to remember, as astutely highlighted by Yopper, that the quality of the output is contingent upon the quality of the input."

I never should have gotten that log in. I'm ROFLing right now.

 

[Go-NoGo]

And in Fedspeak


I never should have gotten that log in. I'm ROFLing right now.

Yes!.. me too.

It has provided me with endless hours of amusement (what can I say, I am easily amused.) but also wonder and appreciation.

The release of AI tools to the general public will be an inflection point (IMO) - productivity gains and everything that comes from it will make this very exciting and interesting to observe.

 

[Andre1969]

I just tried playing around with it, asking it a bunch of automotive-related questions. It gets a lot of details wrong, but is pretty good with more broad answers. It also learns to repeat back what you tell it.

For instance, I asked it what a DeSoto Seville was, which admittedly is a pretty obscure thing, and it came back with the answer of a model offered in 1956 (true) and 1957 (false). I asked it if there was a Seville offered in 1959 and it said no. I typed back that the 1959 Seville was a "Spring Special" model, a trim package for the Firesweep and Firedome.

Well, it actually apologized! I then asked it again, if there was a DeSoto Seville in 1959, and it pretty much repeated what I told it. Although it threw in some guessing game stuff like "special trim" and such. It also said that it had a special grille (which it did not).

I wonder, if someone else asked about a 1959 DeSoto Seville, if it would be able to pull from what it "learned" from me, and regurgitate it to them? Or would it only repeat that back to me?

I also quizzed it on old video games. It actually knew what "The Great Video Game Crash of 1984" meant! And, now here's a weird one. I asked it what "Treasure of Tarmin" was, and it said it was a game for the Odyssey 2 that came out in 1981, a role playing adventure/fantasy game where you controlled a character named Orban.

Wrong. But, I closed the browser window, and opened it up again, and asked what "Treasure of Tarmin" was, and this time it got it mostly right.

"Treasure of Tarmin" was a video game released in 1981 by the company Mattel Electronics for its Intellivision video game console. It was one of the first action-adventure games, and is often cited as an early example of a game with non-linear gameplay, meaning that players could choose different paths and explore the game world in a more open-ended way. In "Treasure of Tarmin," players control a character exploring a dungeon filled with treasure, traps, and dangerous creatures, with the goal of finding the titular treasure and escaping with it. The game was well-received at the time and is remembered today as a classic of early video gaming." It actually came out in 1983, and its full name was "Advanced Dungeons and Dragons: Treasure of Tarmin", but that's still pretty close.

But when I asked it about Orban, it had no idea. I asked what the name was of the character in "Treasure of Tarmin" and it came up with:

"In the video game "Treasure of Tarmin," the player controls a character that is simply referred to as "the adventurer." The character does not have a specific name, but is instead identified by their role as the player-controlled protagonist of the game."

Bingo!

Oh, I asked it what the bore and stroke of a Pontiac 301 engine was and it got the stroke right, but the bore wrong. But then I asked the same of a Ford 302, and it got it right. FWIW, they're both 4.00 bore x 3.00 stroke. It confirms that the Pontiac 301 is really 301.6 cubic inches, but stands firm that a Ford 302 is really 302 cubic inches.

Anyway, I don't think it's really taught me anything I don't already know, but I can still see myself wasting a lot of time with it! And who knows? It will probably "learn" and get better over time.