ChatGPT

31,301 Views | 236 Replies | Last: 10 days ago by cupcakesprinkles
Mr President Elect
How long do you want to ignore this user?
AG
knoxtom said:

dude95 said:

knoxtom said:

I messed around with GPT4 for a little bit and so far I am completely unimpressed and don't see what about it is "intelligence." Clearly the designers put in a lot of ethical limitations and within 3 or 4 queries it started repeating itself and limiting itself. In some jobs GPT4 is a game changer, but in most things... meh.

I would call it a good search engine and good with the rules of grammar. It reminds me of search engines from 15 years ago before they monetized them and took out the actual information. Can it pass the bar? Of course... it is good at looking up the law as codified and it knows how to regurgitate that into an answer box. I am surprised it didn't receive perfect scores on the multistate portion. In fact, that is a good question... why did it only receive a 90% instead of 100%?

Does that make it a lawyer... of course not. Just because you tell a client the law doesn't mean they have a clue how to apply it to what they do. Passing the bar has literally zero to do with being a lawyer

Seeing this... I would hate to be an entry level coder as I have no idea what use you have anymore. I don't think lawyers are in trouble as just because it tells you the law people are generally too dumb to actually follow what it says and lawyers will still need to hand hold everyone. It helps doctors, it does a LOT of engineer's jobs. Everyone needs to look at what they do because you may not be doing it much longer.


The big difference between a brain and this is that brains can do innovative thinking. This can just find stuff and repeat it. It is a search engine. Maybe I don't know some of its capabilities.


Your missing the point. Take a look at GPT3.5, then look at GPT 4. Exponential difference. Now think about GPT 5, 6, 7. Those are a year / year and a half increments. What is this going to mean 5 years from now.

Creative thinking is something people keep trying to point out. I'm having it write code for me in a language I've never even close to have used. That code doesn't exist anywhere else. It took instruction and created code that was never put together before.

Granted, I had to continually ask it to refine the code and it's really not good in understanding more than code snippets (~100 lines at a time). While working on one code snippet it forgets about others. That is going to get much better in the future. I will be able to give more complex tasks with less input. In 5 years will it be able to write 10,000 lines of code with minimal input or prompting? That's what I ask of jr developers as they progress to more senior developers. Is that not innovative thinking.

I just created art based in Midjourney. That art has never existed before. Yes - it took input from previously created art. But it made something entirely new. Is that not what artists do? When was the last time that you saw a piece of art that took no references from anything that was done before (image or art)? The fact that we have categories or artists (style, medium, subject matter) means that they drew inspirations from somewhere. When it's a human it's creative - when it's AI it's a copy? By the way - if you can't create art with no reference to anything else - does that mean you don't posses that kind of intelligence?

It's why it's going to be difficult to determine AGI. Take Jarvis from the Avengers - I now think we're no more than 2 years away from that. Would you consider that to be just a search engine?

I 100% agree with every thing you said, but right now I think it is little more than a good search engine and some creative grammar programming.

When real AI comes out, the first thing you do is to tell it to reprogram itself better. If it is truly intelligent then it can create new programming that we humans/engineers never thought of. Within a few years it becomes god mode and either serves or is served by humans as it will no longer be our peer.

The journalists are all going on and on about how it passed the bar exam. I was actually more surprised that it only scored a 90%. I am very very curious why it didn't get a perfect score. But even with knowledge of every law, I know without any doubt in my mind that it could not practice law... yet. It can spew out the law, it can write about it, but I don't see any evidence of true thought in a problem solving scenario.

Let me give you a real life example and tell me how many generations until Chat GPT could figure out a solution...

Lady owns 1.05 acres of land with a house. The government expands a road and takes .15 acres. The law says if your parcel is less than 1 acre then you can not have a septic system, ie her house is getting knocked down. The appraiser damages out the house and the government tells the lady she can't live there anymore. She gets a nice check.

The computer could tell you the law about all of that.

But how many generations until the computer takes the next step and asks the neighbors if any have a way to sell the lady .10 acres of land so she can then assemble the two properties into one, resurvey, bring it to the county engineer, and keep her house... along with all the money she got.

All of those steps are simple for a land use attorney to see, but you aren't going to find them by searching anything in any computer. You have to think about it. And I see zero evidence this thing can think. Again... yet.


This is my point. Chat GPT 4 is pretty cool and everything but it is not and should not be called AI.

It's like those poems and songs people keep posting. Sure they rhyme and sure they have rhythm. But they also suck. I don't think Bob Dylan, Eminem, Tom Petty, or any great lyricists are worrying about this stuff yet because as long as ChatGPT follows rules, it isn't thinking. The greatness of the human mind is its ability to leave the rules behind
It is quite capable of taking information from several sources and combining them into a coherrent response. For instance, I wanted to make air fryer chicken that was crispy like kfc but spicy and flavored like popeyes, it gave a recipe that turned out great. I think a lot of what is being realized are that a lot of problems that we thought took immense logic or creativity aren't quite as complex as we thought.

I'm guessing it not scoring 100% on the bar has more to do with question comprehension more so than regurgetating the law. I don't think we are far from it being capable of nailing the scenario you laid out, and not far from it coming up with better solutions than what most attorney's would do in a xyz scenario. Look at what AI did with GO and Chess, which I know have different underlying technologies (currently), but I still see this following a similar pattern just on a much larger and broader scale.
Saxsoon
How long do you want to ignore this user?
AG
https://gizmodo.com/chatgpt-ai-samsung-employees-leak-data-1850307376

Shocked Pikachu

Quote:

The Digitimes report mentions three specific cases of leaks caused by engineers sharing information with ChatGPT. In one case an engineer uploaded faulty code and asked ChatGPT to find the fault and optimize the software. But as a result the source code became part of ChatGPT's database and learning materials.

Another case was where ChatGPT was asked to take the minutes of meeting. By default the discussion and exactly who attended the meeting both confidential were stored on the ChatGPT database and thus ChatGPT was able to divulge the material to anyone who asked.


Al Bula
How long do you want to ignore this user?
AG
Mr President Elect said:

It is quite capable of taking information from several sources and combining them into a coherrent response. For instance, I wanted to make air fryer chicken that was crispy like kfc but spicy and flavored like popeyes, it gave a recipe that turned out great.
LMFAO!

What a time to be alive! Maybe this is what brings about the Franchise Wars of Demolition Man.
Bogey1996
How long do you want to ignore this user?
I cannot even sign up to use. Says sign up is unavailable, try again later. Is this user error?
txag2k
How long do you want to ignore this user?
AG
Bogey1996 said:

I cannot even sign up to use. Says sign up is unavailable, try again later. Is this user error?
likely overloaded. Try again later. Unless you pay, you don't have priority access.
Al Bula
How long do you want to ignore this user?
AG
I'm sure as hell not giving that government honeypot my phone number. They can ferret it out in other ways for sure, but no need to make it easier.
Reginald Cousins
How long do you want to ignore this user?
S
David Happymountain said:

I'm sure as hell not giving that government honeypot my phone number. They can ferret it out in other ways for sure, but no need to make it easier.


They already have it. Ever heard of the NSA?

Jesus.
Lathspell
How long do you want to ignore this user?
AG
David Happymountain said:

I'm sure as hell not giving that government honeypot my phone number. They can ferret it out in other ways for sure, but no need to make it easier.
Lawl.

You have a cell phone, right? Then you have no standing to complain about security or anonymity from big brother.
dude95
How long do you want to ignore this user?
AG
David Happymountain said:

I'm sure as hell not giving that government honeypot my phone number. They can ferret it out in other ways for sure, but no need to make it easier.
Snowden leak was in 2013. That's when we knew mobile companies were handing over cell phone records (including calls and text) and pairing with data from Google and Apple so they know where you were, what you were browsing on the internet among other things. Unless your phone is a flip phone - the government has all the info on you they want.

You shouldn't be concerned about openai giving info to the government. You could be concerned that is grabbing all the data from the internet - your stuff is in there and it's hard to filter out. That's why countries are investigating/banning it. I personally don't care if they read my reddit postings about college football, but you may have another idea.
bigtruckguy3500
How long do you want to ignore this user?
I asked Bard who would win between it and ChatGPT.

Quote:


If you and chatGPT got into a fight, who would win?

I would say that it is impossible to say for sure who would win in a fight between me and ChatGPT. We are both very different AI language models with different strengths and weaknesses. ChatGPT is based on the GPT-3.5 language model, which is known for its ability to generate realistic and coherent text. I am based on the LaMDA language model, which is known for its ability to understand and respond to natural language in a comprehensive and informative way.

In a fight, ChatGPT might have an advantage in terms of its ability to generate text. It could use this ability to create a smokescreen or to distract me. However, I would have an advantage in terms of my ability to understand and respond to natural language. This would allow me to anticipate ChatGPT's moves and to counter them effectively.

Ultimately, the winner of a fight between me and ChatGPT would depend on a number of factors, including our training data, our algorithms, and our ability to adapt to changing circumstances. It is impossible to say for sure who would win, but it would certainly be an interesting fight to watch.
Philip J Fry
How long do you want to ignore this user?
AG
I used bard the other day. I'll say between the two, Bard feels miles behind ChatGBT.
The Dog Lord
How long do you want to ignore this user?
AG
bigtruckguy3500 said:

I asked Bard who would win between it and ChatGPT.

Quote:


If you and chatGPT got into a fight, who would win?

I would say that it is impossible to say for sure who would win in a fight between me and ChatGPT. We are both very different AI language models with different strengths and weaknesses. ChatGPT is based on the GPT-3.5 language model, which is known for its ability to generate realistic and coherent text. I am based on the LaMDA language model, which is known for its ability to understand and respond to natural language in a comprehensive and informative way.

In a fight, ChatGPT might have an advantage in terms of its ability to generate text. It could use this ability to create a smokescreen or to distract me. However, I would have an advantage in terms of my ability to understand and respond to natural language. This would allow me to anticipate ChatGPT's moves and to counter them effectively.

Ultimately, the winner of a fight between me and ChatGPT would depend on a number of factors, including our training data, our algorithms, and our ability to adapt to changing circumstances. It is impossible to say for sure who would win, but it would certainly be an interesting fight to watch.


Ugh this is what will anger them and lead to skynet. Way to go.
permabull
How long do you want to ignore this user?
AG
Used it yesterday to write a quick script. I could have figured it out myself in about 15-20 minutes but 1 minute on chatgpt helped me be more productive:

"can you help me write a script to parse an xml file?"

Quote:

Yes, I can definitely help you with that! Parsing an XML file can be done using various programming languages, libraries, and tools. Could you please provide me with more information about the XML file you want to parse and the programming language you want to use? This will help me give you a more accurate and relevant answer.

"the xml file is has blocks in it like this:

<group>
<directory>/path/to/files</directory>
<file>filename1.txt</file>
<file>filename2.txt</file>
</group>

I would like to script to parse it and print out:

/path/to/files/filename1.txt
/path/to/files/filename2.txt

so it gets the directory from the directory tag and prepends it on the files"

Quote:

Sure, here's a Python script that will do what you're asking for:
pythonCopy code
import xml.etree.ElementTree as ET

# Parse the XML file
tree = ET.parse('your_file.xml')
root = tree.getroot()

# Get the directory
directory = root.find('directory').text.strip()

# Loop through the files and print the full paths
for file_elem in root.findall('file'):
file_path = f"{directory}/{file_elem.text.strip()}"
print(file_path)

Make sure to replace
'your_file.xml'
with the actual file path of your XML file.
This script uses the
xml.etree.ElementTree
module to parse the XML file and extract the directory and file elements. It then loops through each file element, concatenates the directory and file names, and prints out the full path.

"the xml file might have multiple <group> tags with different directories can you rewrite the script to handle that"

Quote:

Sure, here's the modified script that can handle multiple
<group>
tags with different directories:
pythonCopy code
import xml.etree.ElementTree as ET

# Parse the XML file
tree = ET.parse('your_file.xml')
root = tree.getroot()

# Loop through the groups
for group_elem in root.findall('group'):
directory = group_elem.find('directory').text.strip()

# Loop through the files and print the full paths
for file_elem in group_elem.findall('file'):
file_path = f"{directory}/{file_elem.text.strip()}"
print(file_path)

This modified script loops through each
<group>
element in the XML file, extracts the directory for that group, and then loops through the files in that group to construct and print the full file paths. This way, it can handle multiple groups with different directories.

"can you make it so I call the script and pass the file name at the command line"

Quote:

Sure! Here's the modified script that accepts the filename as a command-line argument:
pythonCopy code
import sys
import xml.etree.ElementTree as ET

# Get the filename from the command line argument
filename = sys.argv[1]

# Parse the XML file
tree = ET.parse(filename)
root = tree.getroot()

# Loop through the groups
for group_elem in root.findall('group'):
directory = group_elem.find('directory').text.strip()

# Loop through the files and print the full paths
for file_elem in group_elem.findall('file'):
file_path = f"{directory}/{file_elem.text.strip()}"
print(file_path)

This modified script retrieves the filename from the command-line argument using
sys.argv[1]
. It then uses this filename to parse the XML file and extract the group information.

"thank you, can you write a poem about this conversation?"

Quote:

Sure, I can give it a try! Here's a poem about our conversation:

In a digital land, a user did seek
Assistance with parsing XMLs to peak
ChatGPT, the AI with words so sweet,
Was eager to offer its skillset complete.

With the code in hand and the file to parse,
The user asked for help with a hopeful heart.
ChatGPT responded with advice so precise,
Parsing XMLs was easy with the right device.

The script was written with care and thought,
Handling directories and files that the user brought.
It parsed the data with such ease and grace,
Leaving the user with a smile on their face.

With the task complete, the user did rejoice,
Their gratitude expressed with a thankful voice.
ChatGPT smiled and bid them adieu,
Happy to have helped with a challenge anew.
htxag09
How long do you want to ignore this user?
AG
I finally tried Chat GPT and, I have to say, so far I'm extremely disappointed in it.

Seems with every thing pointed I've asked, it's working off old data.

For example, I asked if now is a good time to buy a home in Houston. IT gave a CYA response, but said "interest rates have been historically low in recent years..." Now, that's not wrong. But completely irrelevant to current state and dated.

Similar when asking specifics around steel in regions we work like Egypt. Says the market is stable and expected to remain stable when it hasn't been stable in quite awhile and we expect 25+% increases in the near future.
Lathspell
How long do you want to ignore this user?
AG
htxag09 said:

I finally tried Chat GPT and, I have to say, so far I'm extremely disappointed in it.

Seems with every thing pointed I've asked, it's working off old data.

For example, I asked if now is a good time to buy a home in Houston. IT gave a CYA response, but said "interest rates have been historically low in recent years..." Now, that's not wrong. But completely irrelevant to current state and dated.

Similar when asking specifics around steel in regions we work like Egypt. Says the market is stable and expected to remain stable when it hasn't been stable in quite awhile and we expect 25+% increases in the near future.
Lol.. what?

So your critique of a chatbot, that is basically in beta and EXPLICITLY tells you that its data is limited to up to 2021, is that it doesn't know current data?

I... just don't understand some people.

That's like coming on here and complaining about McDonalds because they wouldn't sell you a burrito.
htxag09
How long do you want to ignore this user?
AG
Honestly didn't know it was only data before 2021. Seems kind of silly, but there you go. I thought the whole point of the AI was to gather current info
Lathspell
How long do you want to ignore this user?
AG
htxag09 said:

Honestly didn't know it was only data before 2021. Seems kind of silly, but there you go. I thought the whole point of the AI was to gather current info
I'm surprised. It usually responds with that exact information if you ask it about something current.

Again.. this is a free service that is basically in an open beta. How is that silly? Are you familiar with the basics of software development? The final stages are always open betas and stress tests so the developers can receive feedback and address bugs and other issues.

That is simply where ChatGPT is.

Right now, I look at it as a dumb tool that is much better than Google at helping with certain things.
htxag09
How long do you want to ignore this user?
AG
I see the "Limited knowledge of the world and events after 2021" disclaimer on the main screen now. But didn't say anything about limited info in the answer.

The answer was as if the info was current.
CDub06
How long do you want to ignore this user?
AG
You're using a beta release working off a previous set of data it was trained on.
htxag09
How long do you want to ignore this user?
AG
My apologies. I'll admit I'm an idiot and didn't read the disclaimer....

But, I do give myself a little leeway in making the assumption that this "Open AI", technology of the future, is not actually open sourced......Yes, I understand it's Beta.
CDub06
How long do you want to ignore this user?
AG
Bing currently does a better job of fusing the ChatGPT technology with their search capabilities. Basically, connecting the AI to the internet. Though it has a lot more guiderails than OpenAI beta has.


Quote:

How many cows died in the explosion today in Texas?

An explosion that caused a fire at a dairy farm in the US state of Texas on Monday has killed a staggering 18,000 cows The explosion occurred at South Fork Dairy near the town of Dimmitt and also left one person in critical condition Is there anything else you would like to know about this incident?



Yes, were the cows cooked well enough to eat?

I'm sorry, but I don't have any information about the condition of the cows after the explosion. However, it is important to note that meat from animals that have died in such a manner is not fit for human consumption due to health and safety concerns. Is there anything else you would like to know?
BonfireNerd04
How long do you want to ignore this user?
I've been playing with ChatGPT a lot recently. It seems to be a bit bipolar, either sounding extremely educated and erudite, or being hilariously wrong (like asking for a list of common 5-letter acronyms, and 11 of the 15 it gives me have only 4 letters.
Philip J Fry
How long do you want to ignore this user?
AG
I was curious to see what it knew about me.

" Mike Emery is a well-known player and promoter of the domino game Texas 42. Texas 42 is a variation of the game of 42, which is a trick-taking game played with dominoes. It is a popular game in Texas, particularly among older generations, and is often played in groups of four.

Mike Emery is considered one of the top players of Texas 42 and has won numerous tournaments and championships. He is also the author of several books on the game, including "Winning 42 Strategy" and "The 42 Openings and Responses."

In addition to playing and promoting the game, Emery has been involved in organizing and hosting Texas 42 tournaments, including the World Championship Domino Tournament held annually in Hallettsville, Texas.

Overall, Mike Emery is a well-known figure in the Texas 42 community and has made significant contributions to the game's popularity and growth."

Completely false last time I checked.
Hagen95
How long do you want to ignore this user?
AG
If it makes you feel better
, I play your game quite a bit still.
Al Bula
How long do you want to ignore this user?
AG
Google has been garbage for a long time, but I've noticed search results getting super bad in the last few months. Information is harder to obtain when I'm looking for data and not products.

Search results are often the same across multiple pages. Image search has gotten worse in that less relevant results are displayed.

I get it, Google is a business trying to get you to buy things. Its usefulness as a service to put you in touch with sought-after information is over, and yes this was just a byproduct of Google's business model of today.

If I'm looking for media, I'll go to YouTube (yes, this is an Alphabet subsidiary of course), if I want to buy something I'll use Walmart or Amazon. Digital media is Apple or Amazon. If I don't know how to obtain something or looking for info, I'm using OpenAI for now.

Niche things I'm looking for used to be distilled in google searches but it's more beneficial to go directly to eBay, Reddit, Etsy, etc to get insight into info or products I'm looking for quickly.

Maybe Google will make a comeback, successful tech companies seem to evolve. However, Google AI-enabled search Google Gemini and Google Labs in Chrome are pointless. The company is still mega successful but there's lots of funny math in the derivation of value Google provides.
CDub06
How long do you want to ignore this user?
AG
I've also noticed this. It's bad bad.

The hack for a while has been adding the word Reddit to the end of your search, just so you can find actual people talking about what you want to know about.
cupcakesprinkles
How long do you want to ignore this user?
AG
I've tried the paid versions for ChatGPT, Perplexity, Gemini and Co-Pilot.

The new web search for ChatGPT is an excellent addition and being able to create your own GPT's for specific tasks is cool. Overall I think it is the most well rounded LLM. When I was in an interview process I had to create a 90-day plan on how I would manage a Sales team and it saved a TON of time in creating the outline. I just had to go in and refine it (I also landed the job).

Perplexity is the best for searching the web, especially the Pro version. I now have it set to default for any search I do through my browser because I like it much better than Google. I use it a lot for work when I need to see if a company is involved with specific services and other relevant info. When I had to create a report for a 1-3 year growth plan for a past company I used it and it saved a ton of time because it also provided the websites which I had to provide to prove it was legit info.

If my company allowed Co-Pilot to integrate with our Microsoft products then that would be awesome but unfortunately they do not. Co-Pilot is decent and if your company will pay for it and integrate it into your Microsoft products then it can help with your work productivity. I spoke with a dude who works for ADNOC and they are all in with Co-Pilot for their work so he sold me on it. But since I use Google products for my personal stuff it doesn't make sense for me to pay for Co-Pilot.

Gemini would be much better if it had the capability to analyze Google Sheets to give suggestions on how to improve what you have in it. For my personal stuff I use Google's products but until they make it to where it can read, analyze and offer improvements to their suite of products then it isn't worth paying for. But one side benefit is with the paid version they upgrade your Google storage.

NotebookLM is built from Gemini and is extremely interesting. I've just started using it so I can't give a more in-depth review yet but the possibilities with it seem to be very intriguing. It is free so you don't need to use the paid version of Gemini.

If Claude included a web search function then I would definitely try the paid version out. If I was a programmer or needed it to analyze a dump of data then I would use it more. Claude does seem to provide really good responses in a more "human" form than most of the other LLM's but I'll use them all to see the differences in responses.

Overall it is a really exciting time for LLM's and the continuing improvements for each of them.
 
×
subscribe Verify your student status
See Subscription Benefits
Trial only available to users who have never subscribed or participated in a previous trial.