Generating code with ChatGPT

[Technoman]

Hi,

As many articles are relating code generation with ChatGPT (OpenAI), I gave a try.

I supplied the following description :
"basic code for a light activated switch Picaxe 08M2 240V SSR connected to C.2 LDR connected to C.1 hysteresis "

Generated code :

' Declare variables
dim threshold as word ' threshold value for LDR
dim hysteresis as word ' hysteresis value

' Set threshold and hysteresis values
threshold = 500
hysteresis = 50

main:
    ' Read the value of the LDR
    readadc C.1, b0

    ' Check if the LDR value is above the threshold + hysteresis
    if b0 > threshold + hysteresis then
        high C.2 ' turn on the SSR
    elseif b0 < threshold - hysteresis then
        low C.2 ' turn off the SSR
    endif
    pause 500 ' wait for 500 milliseconds before checking again
    goto main ' loop back to the start

The result is not too bad. Even after slightly modifying the demand, it was not able to setup constants and use variables instead. Also an arbitrary value exceed 8 bits (threshold).

 

[julianE]

pretty impressive considering that picaxe isn't as mainstream as arduino.

 

[Goeytex]

A friend of mine asked ChatGPT to write some code in GCB the question was:

Help me with programming an LCD in Great Cow BASIC"

ChatGPT Responded with:

Sure, I'd be happy to help you with programming an LCD in GCBASIC. Here is an example of a basic program that will write "Hello, World!" to an LCD connected to a PIC16F886 microcontroller:

'Program to write "Hello, World!" on an LCD
#CHIP 16F886, 4

;Define Hardware settings
#define LCD_RS PORTB.0
#define LCD_ENABLE PORTB.1
#define LCD_DB4 PORTC.3
#define LCD_DB5 PORTC.2
#define LCD_DB6 PORTC.1
#define LCD_DB7 PORTC.0

main:
    'Clear the LCD
    CLS
    'Write "Hello, World!" to the LCD
    PRINT "Hello, World!"

This program starts by defining the microcontroller type and frequency #CHIP 16F886,4 and LCD hardware settings using #define statements for the individual pins for the RS, Enable, and data lines of the LCD.

In the main function the code first clear the LCD using the CLS command and finally write the "Hello, World!" string to the LCD using PRINT "Hello, World!" command.

You can use the LOCATE command to set the cursor position before printing text. For example, you can use LOCATE x,y command to set the cursor at the x-th row and y-th column on the LCD.

There are many other commands that you can use to control the LCD such as PRINT for writing text, LEFT$, RIGHT, and CENTER for text alignment, and CLS` to clear the screen. You can find more information on the Great Cow BASIC website documentation or by looking in the Help files in GCBASIC IDE itself.
Let me know if this helps and if you have any other question or issue

Quite impressive. The code works perfectly.

It seems that ChatPT learns over time. The more questions it is asked the smarter it gets. So it will probably be able to add Symbols in Picaxe Basic after a day or two as it studies Picaxe Basic in its spare time.

Goey (Not a big fan of AI)

 

[Buzby]

I'm not sure what to make of this code generated by ChatGPT .....

' PICAXE code for a "Knight Rider" display

symbol leds = b.0 ' define the LEDs as b.0

main:
for delay = 1 to 20 ' loop through the delay
for led = 0 to 7 ' loop through the LEDs
shiftin leds, leds, right ' shift the LEDs to the right
high leds.0 ' turn on the first LED
pause 50 ' delay for 50ms
next led
next delay
goto main ' loop the program

In this code, the LEDs are connected to pin B.0 of the PICAXE microcontroller. The "main" loop starts by iterating through a delay loop and then through the LEDs. Within the inner loop, the LEDs are shifted to the right using the "shiftin" command, and the first LED is turned on using the "high" command. Finally, a pause of 50 milliseconds is used to create a delay between each LED being lit.

 

[hippy]

ChatGPT is impressive. It seems to have trawled the internet and created a corpus of knowledge about everything. Its cleverness is in deciding what's been asked for, finding and collating information in its knowledge base, and outputting what was asked for in the way it thinks you wanted.

In some way it's like having someone who is an expert at Googling, and surfing StackOverflow and GitHub et al for programming knowledge, with enough skills to figure out what's wanted and delivering that as an appropriately styled answer, or a complete program.

The really clever part for me, and also my concern, is that it often seems to return stuff I have seen, read, and maybe even written, but in a form which is not easily identifiable as coming from a particular source. It creates an answer, and often a good one, but doesn't acknowledge or identify the source materials from which that answer is derived.

It therefore seems to me to be the ultimate plagiarist, passing off numerous author's works as if its own. I don't however have a problem with that beyond it not identifying the source material or acknowledging the creators of the knowledge it is using. It also means it's hard to fathom why SHIFTIN was included in the example given. If we knew the source then we could go to that and see why that was used there, perhaps see how ChatGPT figured out it would be appropriate in its answer.

It seems ChatGPT is incredibly brilliant but also unbelievably stupid and wrong at times, as seen by the two examples above. That's perhaps more glaring in programming where its answers can range from perfect to ridiculous. In less testable situations it can be hard to tell if it is right or wrong, what biases are being represented, and in some cases I have seen it wouldn't be difficult to class answers as propagandist or agenda following.

We are still in the phase where ChatGPT mostly elicits a "wow" response, and it's going to take some time to get beyond that.

 

[erco]

Goey (Not a big fan of AI)

But AI is a huge fan of Goeytex, just like the rest of us.

 

[Buzby]

Please write a poem about picaxe users

Picaxe users, with circuits in hand,
Building robots to explore and expand,
Their knowledge of technology and design,
A passion for innovation in every line.

With soldering iron and PICAXE board,
They tinker and create, never bored,
Programming microcontrollers with care,
Their creations come to life with a flair.

From simple machines to complex systems,
Their work always shows the latest trends and gems,
Whether in school or in industry,
Picaxe users are constantly busy.

Their robots move, sense and react,
Their skills and creativity, an unmatched fact,
They push the boundaries and make new ground,
Their passion for technology, forever found.

So here's to the Picaxe users,
Who make our world more interesting and diverse,
May their creations continue to inspire,
And their passion for technology, never tire.

 

[PhilHornby]

In some way it's like having someone who is an expert at Googling, and surfing StackOverflow and GitHub et al for programming knowledge, with enough skills to figure out what's wanted and delivering that as an appropriately styled answer, or a complete program.

ROFL - that's exactly how my daughter tackled her 'A' level Computing project . It was the first time the school had run the course and the students were all several steps ahead of the teacher . I was very impressed by the way she engaged with StackOverflow - I've never dared!

She works for ARM now (and told me about ChatGPT a few weeks ago).

 

[Flenser]

The result is not too bad

But the code is wrong!
- The variable declarations are wrong. They are for some completely different version of BASIC.
- The IF statement syntax is wrong. PICAXE BASIC does not allow a calculation as a part of the IF test. This looks like it could also be the syntax for some completely different version of BASIC.

When I as exactly the same question of ChatGPT today "basic code for a light activated switch Picaxe 08M2 240V SSR connected to C.2 LDR connected to C.1 hysteresis" I get this different response:

Here is an example of basic code for a light-activated switch using a PICAXE 08M2 microcontroller, a 240V SSR connected to pin C.2, and an LDR connected to pin C.1 with hysteresis:

main:
    readadc 1, b0
    if b0 < 50 then
        high 2
    else
        low 2
    endif
    goto main

This program will continuously read the value of the LDR on pin C.1 and store it in variable b0. It then checks if the value is less than 50 (the threshold for darkness) and if it is, turns on the SSR connected to pin C.2. If the value is not less than 50, the SSR is turned off. The program then loops back to the beginning and continues to monitor the LDR.

Note: you may need to adjust the threshold value to suit your specific application.

This code does pass verification for the 08M2 but does not have any hysteresis.

When I modify the question to make it clearer I want hysteresis included "basic code for a light activated switch Picaxe 08M2 240V SSR connected to C.2 LDR connected to C.1 with hysteresis" I get this different response:

Here is an example of basic code for a light-activated switch using a PICAXE 08M2 microcontroller, a 240V SSR connected to pin C.2, and an LDR connected to pin C.1 with hysteresis:

symbol hysteresis = 20      'set hysteresis value
symbol threshold = 100      'set threshold value for LDR

main:
  readadc 1, b0             'read value from LDR on pin C.1
  if b0 < threshold - hysteresis then
    high 2                  'turn on SSR (connected to pin C.2)
  else if b0 > threshold + hysteresis then
    low 2                   'turn off SSR (connected to pin C.2)
  endif
  pause 100                 'pause for 100 milliseconds
  goto main                 'loop back to main

This code uses the "readadc" command to read the value from the LDR on pin C.1 and stores it in variable "b0". It then compares the value to the threshold value with a hysteresis of 20. If the value is less than the threshold minus the hysteresis, it turns on the SSR connected to pin C.2. If the value is greater than the threshold plus the hysteresis, it turns off the SSR. The code then pauses for 100 milliseconds and loops back to the main routine.

Please note that this is just a basic example and you may need to adjust the values of hysteresis and threshold depending on your circuit and lighting conditions. Also, be sure to check the voltage rating of your SSR and LDR.

This response now uses the correct PICAXE BASIC SYMBOL syntax to declare the constants but the IF syntax is still wrong for PICAXE BASIC.

Then when I change the question to generate code for the 40X2 chip "basic code for a light activated switch Picaxe 40X2 240V SSR connected to C.2 LDR connected to C.1 with hysteresis" I get this different code:

main:
  readadc 1, b0
  if b0 > threshold_high then
    high C.2
    low C.1
  elseif b0 < threshold_low then
    low C.2
    high C.1
  endif
goto main

There are no definitions for threshold_high or threshold_high.
There is no hysterisis this time.
This time ChatGPT has decided not to include any delay into the main loop. Why?
C.1, which is used to read the voltage across the LDR is being set high & low. Why?
- The readadc command automatically configures the pin to be an input so I can't think what the purpose might be of setting C.1 sometimes high and sometimes low between readadc readings.

If the code is wrong for programs this simple .....

 

[papaof2]

The all-knowing guru just doesn't know where to use which piece of knowledge?

I've worked with perhaps a dozen different dialects of BASIC over time but it's rare that I mix syntax.
Perhaps I'm just more experienced with the hardware I'll be using and ChatGPT doesn't have any "hands on" experience?

 

[hippy]

But the code is wrong!
If the code is wrong for programs this simple .....

It has rather set itself up for a fall in claiming to have delivered what was asked for when someone wiser would have prefixed their response with "It will probably be something like ..."

We are much less tolerant of those who claim something will work when it patently won't, or proves not to, than we are to those who admit they don't have the answer but try to help move things in the right direction.

That it's producing different answers to what is basically the same question seems to demonstrate it is merely finding someone else's work which it thinks is the best match to the particular words used, then using some fancy algorithm to make that look more like what was asked for.

Its explanatory commentary does go beyond "b0 = b0 + 1 ; Add one to b0" but it's hard to tell how smart it's really being.

I have spotted such 'plodding commentary' in other ChatGPT examples I have seen, and it may currently be a clue in revealing what is AI generated and not, though I am sure they will try their best to remove such 'tells'.

As here "it's not right" is much easier to determine when one knows it's not right than when one doesn't, and it's rather like Wikipedia and some media sources in that respect. How much we can trust or have faith in what it delivers will probably become a matter of some discussion in the future.

 

[hippy]

Trying to craft a perfect NonGPT answer to the question isn't easy. This was my starting point -

; Here is an example of basic code for a light-activated switch
; using a PICAXE 08M2 microcontroller, a 240V SSR connected to 
; pin C.2, and an LDR connected to pin C.1 with hysteresis:

#Picaxe 08M2
#No_Data
#Terminal 4800

Symbol LDR        = C.1 ; LDR analogue input pin, 8-bit, brighter reads higher
Symbol SSR        = C.2 ; SSR output pin, active high

; Adjust the two lines below to suit your needs
;
Symbol THRESHOLD  = 128  ; 0 to 255
Symbol HYSTERESIS =  64  ; 0 to 255
;
; Adjust the two lines above to suit your needs

Symbol brightness = b0
Symbol lightLevel = b1 : lightLevel = HYSTERESIS ^ $FF Max THRESHOLD + HYSTERESIS
Symbol darkLevel  = b2 : darkLevel  = HYSTERESIS       Min THRESHOLD - HYSTERESIS

PowerOnReset:
  Low SSR

Diagnostics:
  Pause 2000
  SerTxd("THRESHOOLD : ", #THRESHOLD,  CR, LF)
  SerTxd("HYSTERESIS : ", #HYSTERESIS, CR, LF)
  SerTxd("lightLevel : ", #lightLevel, CR, LF)
  SerTxd("darkLevel  : ", #darkLevel,  CR, LF)

MainLoop:
  Do
    ReadAdc LDR, brightness
    Select Case brightness
      Case <= darkLevel  : High SSR
      Case >= lightLevel : Low  SSR
    End Select
  Loop

There are a number of implicit features there, that the LDR reading is 8-bit, the brighter it is the higher the reading will be, the SSR is active high. The 'lightLevel' and 'darkLevel' are variables to avoid issues with doing them as SYMBOL, though one could just define LIGHT_LEVEL and DARK_LEVEL rather than THRESHOLD and HYSTERESIS.

Adding the ability to SYMBOL or #DEFINE what is implicit in the above to create a generic, works for any PICAXE, is quite a challenge. It can be done but fewer will understand the actual code !

 

[erco]

ChatGPT: using Eagle, design me a PCB to rule the world...

One day.soon...

 

[Technoman]

But the code is wrong!
- The variable declarations are wrong. They are for some completely different version of BASIC.
- The IF statement syntax is wrong. PICAXE BASIC does not allow a calculation as a part of the IF test. This looks like it could also be the syntax for some completely different version of BASIC.
...

It doesn't specifically know Picaxe's basic (may get better over time) and give you, for the same demand, different answers, with more or less errors you could easily fix. Starting from scratch, IMHO, it could be helpful. Giving the same description, today, at first I got a working program using symbol for constants... and later constants over 255!

As mentioned by Hippy, AI raises numerous questions about IP (Intellectual Property) among : preserving IP of sources? Who is the owner of the product?...

 

[regpye]

I made a few searches and found this that may be useful. Also it would be a good idea to tell any AI program what basic you are wanting to use as can be seen by what is shown below.
Great Cow BASIC is designed to work with Microchip PIC and Atmel AVR microcontrollershttps://sourceforge.net/projects/gcbasic/reviews/
https://dangerousprototypes.com/docs/index.php?oldid=18927&title=PIC_Resources
While PICAXE BASIC and Great Cow BASIC are dialects of BASIC, they are not fully compatible with each other
https://sourceforge.net/projects/gcbasic/reviews/
. According to a review on SourceForge, the same source code can be used with very few changes in these two popular families of microcontrollers, but the code generated by Great Cow BASIC may execute faster on the equivalent PIC than the code generated by PICAXE BASIC
https://sourceforge.net/projects/gcbasic/reviews/
.However, it is possible to use Great Cow BASIC to program PICAXE chips. For example, a user on the PICAXE forum reported using Great Cow BASIC to program PICAXE 14M2 and 20M2 devices, which are virtually pin-compatible drop-in replacements for PICAXE chips
https://picaxeforum.co.uk/threads/going-nuts-not-picaxe.30251/
.Overall, while Great Cow BASIC is not specifically designed for PICAXE chips, it is possible to use it to program these chips with some modifications. However, it is important to note that the code generated by Great Cow BASIC may not execute as fast on PICAXE chips as it does on Microchip PIC and Atmel AVR microcontrollers.

 

[Technoman]

... it is possible to use Great Cow BASIC to program PICAXE chips ...

A Picaxe chip is a standard Microchip microcontroller programmed with an interpreter and a downloader : see Picaxe manual 1, page 16. Once you do that, the interpreter is gone, you can't program it anymore using Picaxe tools.

 

[Technoman]

Can ChatGPT predict the future?

 

[hippy]

Can ChatGPT predict the future?

LOL. That to me shows how on one hand it's amusing for those who know the truth but on the other hand it can be problematic, perhaps outright dangerous, if those who don't know the truth believe its lies.

First it has simply invented the 20M3 and presents its existence as if fact.

It claims the 20M3 runs at 20MHz, faster than 20M2 which runs at 16MHz, when the 20M2 can run at 32MHz.

The "20" in any PICAXE naming indicates the number of chip legs so it would seem impossible to have 20 I/O pins if it allows 0V and V+ to be connected which is usually essential for operation.

The rest is just invention or guess presented as fact. Some things might be true if an M3 actually existed but who knows.

My biggest complaint of ChatGPT is that it cites no source for the things it is claiming, and others have reported that, when it does, it often simply invents citations which don't exist, are unrelated, don't support what is being claimed.

My greatest concern is that some people will believe whatever it says, will take its drivel and lies as fact, believe it is true, and potentially act on that basis. Currently it's impossible to tell when it's lying and not unless one already knows the facts.

And what happens when other AI, which are also trained on whatever they can scrape from the internet, take rubbish emitted from ChatGPT as fact. Pretty soon they will all be churning out garbage and, if they do start citing sources, it will probably end up like conspiracy and hoax sites where they are completely circular links.

One might think people will simply start ignoring ChatGPT because it's not reliable, not truthful, but we know how it is with conspiracy and hoax sites. It's one thing to deal with the non-existence of a 20M3 but it can be something else trying to deal with lies on more contentious real world issues. There will even be those who will wilfully take ChatGPT's lies and use them to support their own lies and agendas, the modern day parallel of "I read it in the newspaper so it must be true".

I don't know how we are going to resolve this issue as there is now a cohort which believes lies are as equally valid as truth, that alternative fact and made up garbage is as good as real and evidenced facts, who would take any attempted regulation to enforce truth as censorship, a tyrannical act to take away their freedom to lie, their right to propagate lies.

As some have said; we live in interesting times.

 

[Buzby]

The results from asking chatGPT fairly simple questions ( not related to programming ) are quite worrying.

e.g. This is the result I got when asking if Kit Pedlar's book 'Mind over Matter' is still available.

Yes, Kit Pedlar's book "Mind Over Matter: The Images of Pink Floyd" is still available for purchase. The book, which was first published in 2003, explores the connections between science fiction, technology, and the music of the iconic British rock band Pink Floyd.

This is totally wrong, but very convincing and believable. Kit Pedlar never wrote about Pink Floyd, but if I didn't already know that I could quite well take it as true.

This is a tiny example of the sort of problem we will have in the future, false facts or news appearing as real. If the only interface you have to, say, your bank is AI, and your account has problems, where do you turn to ?. ( I've had this with real support people who could only act on what they saw on their screens. It tooks months to sort it out. )

Another example ...

Q : Which is heavier, a kilo of feathers or a wednesday of gloves ?
A : Both a kilo of feathers and a Wednesday of gloves weigh the same, which is one kilogram.

This is an example of the difficulty in processing language. If in the future an AI has to make a medical or financial decision, how can we be sure it has correctly interpreted the question ?

I'm not a fan of AI yet !

 

[papaof2]

If in the future an AI has to make a medical or financial decision, how can we be sure it has correctly interpreted the question ?

And you also have the problem of how well the person asking a question can configure and qualify that question.

For medical questions, you must have multiple patients on which to apply the AI's suggested treatment. Keep count of which ones work, which neither help nor hurt and which are damaging and which are fatal. Just make the ChatGPT medical answers are subject to investigation and loss of license under the authority of local medical boards.

There's always the possibility of AI getting slightly more than half of the answers correct by a mix of some possibly pertinent facts swaying the random 50/50 nature of a yes/no question such as "Is beheading a cure for insanity?"

Considering some of the people who are currently in power in politics, a portion of the voters must have asked ChatGPT "Who should I vote for?"

 

[Buzby]

And you also have the problem of how well the person asking a question can configure and qualify that question.

In most cases there will not be a person placing the question, it will be the output of some other AI system !

 

[julianE]

I think AI will be little more than a novelty, tech is hitting a wall and they are trying anything and everything to stay relevant. I have a 10 year old laptop and a brand new one, for everything I use a laptop there is no difference, yes the 10 year old laptop was a higher end unit at the time but 10 years used to be an eternity. Even our beloved picaxe has not been freshened in ages and it's perfectly fine for 90% of applications.
think of all the life changing tech that never became big, i remember when "in the future" we could get everything custom, we were going to be scanned and haberdashery would fit us perfectly, IoT, where a fridge would tell us we are out of milk, yea that took off. remember the stupid app that would let you know when your friends were in a vicinity and could meet for a pint, yea that was brilliant thinking for weirdo stalkers. mindless VR ideas, even the porn industry is indifferent to VR. Don't get me started with vehicles and all the stupid tech they are hoisting on consumers.
rant off.

 

[Buzby]

I think AI will be little more than a novelty, tech is hitting a wall and they are trying anything and everything to stay relevant.

Sorry, I disagree.

In the last 20 years the World has been transformed by Social Media, ultra miniature electronics ( smartphones etc. ), robots in manufacturing, advances in genetics, new materials ( especially plastics ), and a host of other things. These advances show no sign of slowing down.

AI is already established in at least one aspect of medicine, the routine analysis of breast cancer test samples.

The 'stupid tech' in new cars is the precursor of real self driving cars, and these are very soon to be on the roads big time.

We don't have proper 'general use' robots yet. The hardware is ready ( Boston Dynamics, etc ), but they still need brains. AI will provide them.

ChatGPT is the first successful example of its kind, like the Wright Brothers flight in 1903. They flew a distance of 120ft at an altitude of 10ft. Six years later Blériot flew across the English Channel.

Six years from now AI will be starting to manage production lines, automate farming, perform office admin work, and a hundred other jobs. There will be lots of mistakes, but they won't stop progress.

I think the biggest change will be when robots work along side us. First it will be jobs like bricklaying, or receptionists, or shop assistants, or any of a thousand different factory jobs.

All the progress the human race has made over thousands of years has been due to Man's intelligence, but no man is intelligent enough to build a more intelligent man.

An AI will be able to design a more intelligent AI, that's what the recent letter from the Future of Life Institute ( Elon Musk, Steve Wozniak, etc ) is worried about.

I honestly don't know what the future will be like, but I do know it will be very different.

 

[PhilHornby]

ChatGPT is the first successful example of its kind, like the Wright Brothers flight in 1903. They flew a distance of 120ft at an altitude of 10ft. Six years later Blériot flew across the English Channel.

That's quite interesting, because the first manned flight (of an heavier than air aircraft) actually occurred in either 1849 or 1853. It took another 50+ years of experimenting, before the Wright brothers finally cracked it (i.e. got the thing to fly where they wanted it to). AI feels a bit like that: I first used an AI program back in the 1980's, but the steps forward have seemed very slow until ChatGPT was unleashed on the world. Already, it and its competitors are everywhere!

 

[julianE]

I honestly don't know what the future will be like, but I do know it will be very different.

could not agree more

 

[hippy]

The steps forward have seemed very slow until ChatGPT

I am not convinced ChatGPT is such a great jump forward. I think it is just that it's an application of AI which has captured the public's imagination and got itself the hype other less public-facing AI never got. One can argue it's ability to create its word salad is far less impressive than the AI behind full self-drive and similar.

ChatGPT's success is in appearing to be human-like, and that feels like a big jump in the game. One could say that ChatGPT's big success is in emulating the people we have all encountered; those who always know the answer to everything and simply make up things and ignore facts to support their answer. I was reading an interesting article yesterday which said that, what AI companies are calling "hallucinations", is better described as BS.

I think ChatGPT's biggest success may be in getting people to talk about AI, think about what it is, what it means.

 

[PhilHornby]

As an AI language model, I don't have feelings, opinions or personal biases, but I can provide you with some objective facts and insights related to your message.

Firstly, it's true that ChatGPT is not the only impressive AI application out there, and there are certainly many other AI technologies that are more specialized and focused on specific tasks, such as full self-drive as you mentioned. However, what sets ChatGPT apart is its ability to generate coherent and contextually relevant human-like responses in natural language, which has proven to be a significant breakthrough in the field of artificial intelligence and natural language processing.

In terms of the accuracy of its responses, ChatGPT is based on a language model that has been trained on massive amounts of text data from the internet, making it highly knowledgeable and able to generate highly plausible answers. However, it is not infallible, and there may be times when its responses are not entirely accurate or relevant to the given context.

Regarding the concept of "hallucinations" or BS generated by AI, it's important to note that these are often a result of the limitations of the training data and the models used. While AI models are highly advanced and can learn to generate highly plausible responses, they still rely heavily on the quality and diversity of the data they are trained on. As such, there may be cases where the AI generates responses that are not entirely accurate or relevant to the given context.

Finally, it's true that ChatGPT has helped raise awareness and interest in AI among the general public, as well as the potential implications and ethical considerations associated with AI development and deployment. As AI technologies continue to advance, it's crucial that we engage in ongoing discussions and debates about their impact on society and how we can ensure that they are used ethically and responsibly.

It has an answer for everything

 

[hippy]

It has an answer for everything

And that appears to be by design. It's primary objective is to give a plausible sounding answer to everything and has no shame in inventing and lying to do so. It will even admit its answers aren't always truthful, factual, correct or accurate., that it can't and shouldn't be trusted.

Like every BS artist it blames others for the way it behaves, absolves itself of any responsibility for any harm or damage done by victim blaming, laying responsibility on those who believed any nonsense it's spouting.

In many ways it's superb 'performance art', challenging the very nature of belief and truth.

And not unlike those well known media personalities some call propagandists who air conspiracy theories and emit lies to deliberately and knowingly present a false perspective. They also use the 'performance art' defence, distance themselves from the harm and damage which leads from what they say.

Which presents an interesting question as to whether they, and likewise ChatGPT, are "the problem", or is that elsewhere ? Some will argue that lies, untruths, alternative facts, are just as good as other people's truth and facts, deserve equal footing, are pushing to have it accepted as so.

There's going to be some very philosophical and existential debates ahead.

 

[julianE]

There's going to be some very philosophical and existential debates ahead.

the debates have happened long ago, post modernism is very old. people cranking out much of this drivel are poorly educated computer nerds serving drivel to poorly educated public. there was a time when engineers and scientists received a well rounded education, J. Robert Oppenheimer was a liberal arts student at one time, a wide ranging intellect. we are living in an age of specialists.

this forum is one of the few places on the internet that has informed discussion, i may not agree with the conclusions but the arguments are well and politely presented.

Like every BS artist

reminds me of steve jobs passing i-phones as revolutionary designs, it's bauhaus from the 1920s but to the masses it's "new".

 

[Tvmender]

Has anyone tried the Turing test with it?

 

[hippy]

Has anyone tried the Turing test with it?

I believe Turing's test is whether a machine is indistinguishable from a human in the responses it gives to questions asked. In which case I believe ChatGPT would fail based on examples we have already seen published.

A Turing test could perhaps be undertaken with just one carefully crafted question , one which would elicit a "don't know" or similar response from a human, while ChatGPT would likely make something up or give some answer which stylistically differs from what we would expect from a human response..

Setting aside that a 20M3 doesn't exist, ChatGPT's response shown earlier is verbose and rather different to what a normal human response would be which would usually be shorter and to the point; "it's got more memory for code and data, can run faster and the M3 has features which the M2 doesn't". ChatGPT is more verbose, more essay like, will throw things in to give itself a greater air of credibility which aren't strictly relevant to what was asked.

It typically ending with a final paragraph as it does in the M3 response is what I find usually gives it away as not being human. Most humans don't add an "in conclusion" unless writing an essay or article.

But how much of that is ChapGPT like ? I should perhaps have just said "I don't know". I would suggest my caveating with "I believe", "could perhaps", does illustrate I am a real person because ChatGPT would currently express itself in more certain and confident tones.

 

[regpye]

LOL. That to me shows how on one hand it's amusing for those who know the truth but on the other hand it can be problematic, perhaps outright dangerous, if those who don't know the truth believe its lies.

First it has simply invented the 20M3 and presents its existence as if fact.

It claims the 20M3 runs at 20MHz, faster than 20M2 which runs at 16MHz, when the 20M2 can run at 32MHz.
And what happens when other AI, which are also trained on whatever they can scrape from the internet, take rubbish emitted from ChatGPT as fact. Pretty soon they will all be churning out garbage and, if they do start citing sources, it will probably end up like conspiracy and hoax sites where they are completely circular links.

Not only chatGPT giving this information.
PERPLEXITY
The PICAXE microcontroller chips come in two series, the M2 and X2 series, and six physical sizes, including 8, 14, 18, 20, 28, and 40 pins
[1]
. The PICAXE-20M3 is a microcontroller that runs at 20MHz, which is faster than the PICAXE-20M2 that runs at 16MHz
[2]
. However, the PICAXE-20M2 can run at 32MHz
[2]
. For more information on PICAXE chips, you can visit their website or online store
The PICAXE-20M2 is a microcontroller chip that is compact but powerful and supports up to 18 inputs/outputs with 11 analogue/touch sensor channels
[1] . It is very popular in education due to its low cost and small size, but also very popular in a wide range of hobbyist projects
[1] . The chip replaces the older PICAXE-20M chips
[1] . The PICAXE-20M2 microcontroller chip supports 16 input/outputs including analogue inputs
[2] . It is available for purchase on solarbotics.com for $5.95 USD
[2] . The chip is also available for purchase on picaxestore.com for £3.59
[3] .The PICAXE-20M2 microcontroller chip is part of the M2 series of PICAXE chips
[1] . The M2 series includes the PICAXE-08M2, PICAXE-14M2, PICAXE-18M2, PICAXE-20M2, PICAXE-20X2, PICAXE-28X2, and PICAXE-40X2
[1] . The same software, BASIC language, and USB download cable is shared between all the PICAXE chips
[1] .The PICAXE-20M2 microcontroller chip can be used in various projects, such as a light-activated switch
[4] . The chip can be programmed using the BASIC language and downloaded to the chip using a USB download cable
[1] . The PICAXE-20M2 microcontroller chip is a versatile and affordable option for hobbyists and educators alike.
4 SOURCES

picaxe

solarbotics

picaxestore

picaxeforum

 

[regpye]

ChatGPT is impressive. It seems to have trawled the internet and created a corpus of knowledge about everything. Its cleverness is in deciding what's been asked for, finding and collating information in its knowledge base, and outputting what was asked for in the way it thinks you wanted.

chatGPT 3.5 has no connection to the Internet it relies solely on what was given to it in 2021 and after that what we teach it with trial and error.
If chatGPT was connected to the Internet it would be great, the next version may do this

 

[Buzby]

I just tried Perplexity ....

 

[Buzby]

I asked again, but this time with the 'enhanced' option, which uses ChatGPT4 ....

 

[hippy]

"Yes, there is a PICAXE 20M3 microcontroller. However, it is referred to as the PICAXE-20M2".

That's a classic. I guess we could skip an entire M3 series and go straight to M4 with zero effort. But why stop there

I am not sure why it's citing 'picaxe' as the source of that information because I couldn't find any mention of 20M3 through Google other than this thread. Maybe it's an extrapolation of someone asking about "M3" availability and getting a response that "M2" is the latest ?

I did wonder if it had got confused by our stock codes but not sure how. "PICAXE-20M2" is stock code "AXE102M2" or " AXE012M2SM" in the store.

 

[julianE]

I had my first go with AI using bing, the results were less than impressive. I'm also seeing it all over the web, I looked up my favorite violinist and was taken to a website that had all the data correct, written badly. Reading anything by hippy, alen and a number of others is more than informational it's often sprinkled with wit and always written well. My question is answered and often I learn more than I asked. I'm sure we have all been to lectures where a lecturer takes a trite question from the audience and crafts an answer full of meaning. Once we transition to AI wave beauty goodbye. I can't figure out what the end goal is, eliminate technical writers, which used to be a very well paid profession, take away jobs from liberal arts majors. Solution looking for a problem.

There are so many of these idee fixe in silicon valley, they read one book and it consumes them, Snow Crash might be the only book read in Silicon Valley and the world has to bend to their will and we all should live in a metaverse. I recently read that one of the reasons for 5G is to enable a super connected world and to make metaverse a reality.

I made my living working in tech and have spent time working in san jose, much good came from technology and it has improved many lives but it's become hell bent on eliminating every job and creating a world empty of beauty. oh well.

what do i know, i still listen to music on vinyl

 

[OLDmarty]

ChatGPT: using Eagle, design me a PCB to rule the world...

One day.soon...

Did you want a single-sided world? or a double-sided world? ;-)

 

[lbenson]

Did you want a single-sided world? or a double-sided world? ;-)

Too much double-sidedness now. I want a four-layer at a minimum.

 

[erco]

what do i know, i still listen to music on vinyl

You modernist sellout! I'm sticking with shellac resin.

You Are My Sunshine/Bill Mooney

www.youtube.com