With the results I got, I would be surprised if it was useful.
It would be nice when someone asks to say, yah, just make your code work like this:
Then provide code you requested and reviewed from the bot to save time generating it.
Ah, I don't think I am being helpful right now. Just so discouraged about that bot.
Sorry
I would say that ChatGPT will soon be a legitimate threat to google as a search engine, but from a code development standpoint, competent developers will not be threatened for quite some time.
I think it would still scale something crudely like this:
10 devs * 10 hours * 1 good development /10 hours = 10 good developments
10 devs * 10 hours * the best bot developing 1 good development / 1 hour = 100 good developments
Just came across this in my weekly reading, and it seems somewhat relevant to this thread: AI.com now links directly to ChatGPT
Yes, ChatGPT can generate a Perspective View. In this post I fed it a lot of information to get the results for a basic Perspective View. The next post will be using the GPT-4 framework to build the View from scratch and then teach ChatGPT how to get the right format, JSON nodes, etc.
I don't think it will replace Ignition developers anytime soon for a variety of reasons, I think it will ultimately become an orders of magnitude better version of what developers have used Google for over the last 25 or so years.
" 3. ChatGPT will not create everything from scratch
Seems like a function of time and effort before it can do more.
I don't understand how you got it to work though.
I have had poor results with ChatGPT for a long time now since they dumbed it down.
edit
Seems that ChatGPT4 is still smart for people who buy plus.
Whatever version I am getting as a free user, it is dumb compared to the free version in January imo.
Must have learned the format for the JSON from my initial experiments. I asked it to create a Perspective View for me with a text field and label and it gave me detailed instructions on how to bind everything, then I asked it what information I would need to provide to build a view and then it gave me the JSON for a view right there. Had to tell it to tweak a couple things and re-generate and was able to get a view I could import directly into Ignition in about 2 minutes from start to finish including the time it took ChatGPT to type it out along with all the instructions.
You definitely need to feed it a lot of information to know what is going on, it doesn't know about the various components, extrapolates poorly on component types "ia.display.button" instead of ia.input.button" for example, and sometimes it nests the properties in really weird ways but it fixes them and remembers the correct format after you tell it how to fix it.
Excellent post. I didn't know there was a continue statement for long code blocks until I read this; that's a useful thing to know.
This week, I read this article that readers of this thread could find useful:
How to Use ChatGPT to Write Code
It mentions the continue statement multiple times in the opening paragraphs which caused me to remember this thread and prompted me to post the article here.
Humane - cofounded by Imran Chaudhri had a TedTalk 2 days ago.
The Disappearing Computer: An Exclusive Preview of Humane’s Screenless Tech | Imran Chaudhri | TED - YouTube]
That company is making a device that has a camera and mic to receive the information we do, and then AI on top of that to process the information and respond.
The next evolution I think is to combine with something like Google Remote Desktop where the remote device could interface with our computer. We could pair program solo?
I am pretty excited.
I can't find anything on when the device will be on the market though.
Hope it comes with a lanyard option. I don't like pockets on my shirt.
Yup. I believe he is correct about the direction we are trending.
https://www.lightfieldlab.com]
He is projecting onto his hand though.
Solidlight holographics are more practical.
Not sure why there aren't page displays from those displays already.
The 3d objects seem harder than 2d to me, and less practical.
I ordered a device on Amazon for like $90 to see if I can make it display a webpage.
Then with the revolver plugin, I could have my Ignition pages showing at my desk in a fun way.
But maybe the lights don't have that kind of flexibility.
My takeaway was that the screens are getting less visible, and the idea he's working towards is a world where the technology disappears from view. From my perspective, both holograms and projections seem to be a poor fit for this vision.
...even so, the hologram projectors do look really cool 
Because reading is faster than hearing, I expect some kind of projection will be required.
I was participating in a survey for Stack Overflow, when I encountered the following lists of AI tools:
Most of these, I've never heard of.
There are 100s of new tools each day so I'm not surprised lol. It's impossible to keep up
IIRC, the search for alternatives to chatGPT has come up a few times in this thread, so when I encountered a list of such things, I felt compelled to share it. I figure if they were prominent enough to make the survey list, they would probably be worth checking out by those who are interested in such things. So far, I have only tested Bard, Bing, chatGPT 3.5, and chatGPT 4.0, and personally, I don't feel any inclination to invest anymore time in testing these technologies at the moment.
If anybody is interested in my assessments, here they are:
bard PROs:
None
bard CONs:
• I estimate its overall abilities to be about a quarter of what GPT 3.5 can do.
• Its responses exhibit a clear political bias that goes far beyond what could incidentally occur as a natural result of the material it has been trained on.
• Google is making no attempt to comply with privacy regulations, and as a consequence, they are intentionally withholding the technology from all European Union countries
• When asked even basic coding questions, this produced code that was totally unusable. Generally, the code wasn't even close enough to what was actually needed to be useful to an experienced developer with good debugging abilities.
chatGPT 3.5/4.0 PROs:
• It's really good at text generation: essays, technical documentation, text messages, etcetera.
• I've heard that it's even really good at generating love letters, romantic texts, and nice things to write in cards, but obviously, I've never tested this, and anything that I've written to my wife since this technology has become available has come straight from my own heart.
• During my testing, it was surprisingly good at math. I tested it with large systems of equations, complex [imaginary] numbers, and other advanced algebra problems, and while 3.5 went off the deep end a couple of times, 4.0 almost always produced the correct answers, and unlike a calculator, it shows the work involved in getting the answer. If this had been available when I was taking calculus in college while working a full time job, I probably would have gotten a lot more sleep.
chatGPT 3.5/4.0 CONs:
• If it doesn't know the answer, it will fabricate one and present it with a confidence that makes the answer seem both plausible and correct.
• When used for coding, the code is rarely usable as is, but generally it does produce code that is close enough to what is actually needed to be useful for an experienced developer to debug.
Bing PROs:
• Unlike google, the search results are straight and to the point. There are not 10 to 15 advertisements to search through, and perhaps most importantly, it doesn't automatically cross out critical search parameters like google does, so the results are far more relevant.
Bing CONs:
• Bing itself still feels clunky, but clicking the chat icon does clear all of the in your face advertisements and loads the clean AI chat interface
• The conversation style choice is confusing, and I'm not sure how it actually affects the core results that are produced.
• Even though I prefer this over google, some part of my brain knows that this is a microsoft product, so there is some well earned distrust. Nevertheless, I haven't encountered anything in my usage so far that would validate these feelings.
@zacharyw.larson
How about this one:
Scientists develop an A.I. that turns peoples’ thoughts into text
Reading today means writing tomorrow.
That tech will evolve into real-time writing to many brains.
Hope it's not worse than watching TV.
Nextsense is making an over-ear brainwave reader.
Google had interviews saying they don't know how their AI did X and Y.
It is unclear if Google is hyping AI dangers or doesn't know how their AI works.
AI companies hype AI dangers to push licensing to block additional AI startups.
If Google is unsure how their AI works though, Google should probably reset their project.
The internet hasn't converged to one-way media like TV, Radio, and printing.
Tutorials on constructing language AI bots are available.
Seems unlikely licensing and regulations will prevent startups.
Probably would be more effective to publicize good standard practices.
It is not very competitive in nature though.
I use ChatGPT specifically daily. It helps me write python and SQL daily.
I am using it even on weekends.
I have been able to do some things I didn't realize I could.
Some of the softwares I am using are being upgraded on the development side with ChatGPT.
It is making me faster and it making people faster at making software that makes me faster.
Though some days, ChatGPT is effectively stupid.
Sometimes ChatGPT just spits out useless incorrect information and code.
Recently, it has been doing this thing where it only checks one message prior in the conversation.
It isn't too terrible, I have to compound all my previous requests of ChatGPT into the next request till it finally gives me what I am trying to get.
It still isn't making queries for me.
It still isn't making pages for me.
Though I am sure it could if the OpenAI devs had not dumbed it down since January.
I don't think I have an incorrect view of how it was in January, but maybe I was asking it simpler things.
I cant be alone in saying I am never putting hardware in or constantly near my brain. Feels way to dystopian. There's not a person on earth I trust enough to make that hardware, not to mention the issue of who gets to decide to do what with the data that is collected. I know kids in high school are apparently looking forward to this (according to friends who are teachers) so this must be my "I'm getting old" moment.
Regarding the 'don't know how an AI did something' since you seem very much into the promise AI you should try making a deep neural network from scratch if you know linear algebra. Theres a few tutorials around about how to do it and setting up the gradient descent, and training your model to be less wrong over iterations. Then train it to solve a problem like predicting if a titanic passenger survived or not like in this Titanic - Machine Learning from Disaster | Kaggle. Now imagine you were asked to explain why your model said this person did survive and another person did not and all you have essentially are a bunch of neural nodes of different weights, its very hard to parse "why" though with the example I gave there are tools now and people can explain why - but I imagine the complexity multiples exponentially when you are making an AI like generalized language model that ChatGPT is.
You should check out the late philosopher Marshall Mcluhan if you're interested in how new media affects old and for a different perspective on technology in general. You're right it doesn't necessarily eliminate old media but it does tend to change it. The book "The Shallows" I think it was, mentions how every issue of Newsweek had pages of pure text in the middle for the main story. Then since the advent of the internet, now magazine (and newspaper) articles are more setup for the quick perusal that internet articles had been doing - big block quotes in the middle, pictures relatively often, and keeping articles in general on the shorter side.
I do use ChatGPT sometimes for hard problems that I assume have already been solved on StackOverflow (since ChatGPT was trained on that) but you should always keep in mind all it's really good at is using statistics to guess what the next word should be with pretty good accuracy, but it itself cannot then parse and determine any truth value of what it just told you.
I won't wear implants like the Neuralink thing.
I would wear a device like Nextsense is making.
I drink pop, but refuse to drink bottled water from recycled plastics.
Bottle manufacturers have been proven to not know what's in the recycled plastics.
A lot of lines will be drawn on AI.
I don't plan to make an AI bot myself.
I do like to watch the tutorials.
I am not familiar with the term the promise AI.
I would avoid using ChatGPT if I had a better alternative.
I am not sure that group knows what they are doing anymore than Google.
The way the companies talk sometimes, it seems like they are just trying algorithms instead of designing what they think would be good to use.
I suppose I do need to read more about the things that have happened that were bad already.
It is possible I underestimate the damages.
https://www.analyticsinsight.net/famous-ai-gone-wrong-examples-in-the-real-world-we-need-to-know/]
