Re: [PSES] AI & Regulatory Compliance

[Douglas Nix]

Hey, Ralph,

Listen, I get being excited about AI and the possibilities. The problem I have 
with this discussion is the approach that some people are taking using this 
tool. There are some excellent AI-based academic tools available:

▸Wisio.app (limted free version)
▸Jenni.ai
▸OpenRead.academy
▸Other research tools include:
-Paper Digest (https://www.paper-digest.com )
-Elicit (https://elecit.org )
- Litmaps visual literature search (https://www.litmaps.com 
)
-Research Rabbit (https://www.researchrabbit.ai 
)
-Connected Papers (https://www.connectedpapers.com 
)
-writefull (https://www.writefull.com )
-Penelope.ai (https://www.penelope.ai )


All of these tools can be very helpful. I’m a fan of Litmaps and Elicit in 
particular. I’m just starting to poke around at Wisio, Jenni, and OpenRead.

Doug Nix
d...@mac.com

"The man who does not read good books has no advantage over the man who can't 
read these books." Mark Twain 

> On Oct 10, 2023, at 17:01, Ralph McDiarmid  wrote:
> 
> That analogy was not intended, and I’m merely trying to point out that AI is 
> here to stay and is expected to become a more useful tool in our industry as 
> it has been proven to be in the medical profession.
>  
> From: Douglas Nix  
> Sent: Tuesday, October 10, 2023 1:03 PM
> To: rmm.priv...@gmail.com
> Cc: EMC-PSTC@LISTSERV.IEEE.ORG
> Subject: Re: [PSES] AI & Regulatory Compliance
>  
> You cannot compare ChatGPT to a medical AI that has been tailored for a 
> specific task like medical image analysis. That’s like saying a 1968 Mini 
> Cooper and Dodge 440 HEMI Charger are comparable because they are both cars. 
> Nope.
>  
> Using ChatGPT to summarize a paper, produce an abstract from uploaded text, 
> or produce a set of points as a starting point is perfectly fine. You are 
> giving the software the specific input material from which to generate the 
> output. You cannot ask it research questions an expect a valid response 
> because ChatGPT has no parameters for correctness. It only wants to give you 
> a plausible sounding answer. It will give you an authoritative answer with no 
> reference to anything resembling truth.
>  
> Doug Nix
> d...@mac.com 
> http://www.dougnix.net 
> (519) 729-5704
>  
> "All animals except man know that the ultimate joy of life is to enjoy it."  
> -- Samuel Butler
>  
>  
> 
> 
>> On Oct 10, 2023, at 14:15, Ralph McDiarmid > > wrote:
>>  
>> Physicians have used AI (expert systems) in their offices for many years as 
>> a tool to help them diagnose a problem more accurately and will greater 
>> speed.  It’s a tool to speed productivity, and that’s how I use it today.
> 
>  
> This message is from the IEEE Product Safety Engineering Society emc-pstc 
> discussion list. To post a message to the list, send your e-mail to All 
> emc-pstc postings are archived and searchable on the web at: 
> https://www.mail-archive.com/emc-pstc@listserv.ieee.org/
> 
> Website: https://ewh.ieee.org/soc/pses/  
> Instructions: https://ewh.ieee.org/soc/pses/list.html (including how to 
> unsubscribe) 
> List rules: https://ewh.ieee.org/soc/pses/listrules.html
> 
> For help, send mail to the list administrators:
> Mike Sherman at: msherma...@comcast.net 
> Rick Linford at: linf...@ieee.org 
> For policy questions, send mail to:
> Jim Bacher at: j.bac...@ieee.org 
> To unsubscribe from the EMC-PSTC list, click the following link: 
> https://listserv.ieee.org/cgi-bin/wa?SUBED1=EMC-PSTC&A=1 
> 


-

This message is from the IEEE Product Safety Engineering Society emc-pstc 
discussion list. To post a message to the list, send your e-mail to 


All emc-pstc postings are archived and searchable on the web at:
https://www.mail-archive.com/emc-pstc@listserv.ieee.org/

Website:  https://ewh.ieee.org/soc/pses/
Instructions:  https://ewh.ieee.org/soc/pses/list.html (including how to 
unsubscribe)
List rules: https://ewh.ieee.org/soc/pses/listrules.html

For help, send mail to the list administrators:
Mike Sherman at: msherma...@comcast.net
Rick Linford at: linf...@ieee.org

For policy questions, send mail to:
Jim Bacher:  
_
To unsubscribe from the EMC-PSTC list, click the following link: 
http

Re: [PSES] AI & Regulatory Compliance

[Ralph McDiarmid]

That analogy was not intended, and I’m merely trying to point out that AI is 
here to stay and is expected to become a more useful tool in our industry as it 
has been proven to be in the medical profession.

 

From: Douglas Nix  
Sent: Tuesday, October 10, 2023 1:03 PM
To: rmm.priv...@gmail.com
Cc: EMC-PSTC@LISTSERV.IEEE.ORG
Subject: Re: [PSES] AI & Regulatory Compliance

 

You cannot compare ChatGPT to a medical AI that has been tailored for a 
specific task like medical image analysis. That’s like saying a 1968 Mini 
Cooper and Dodge 440 HEMI Charger are comparable because they are both cars. 
Nope.

 

Using ChatGPT to summarize a paper, produce an abstract from uploaded text, or 
produce a set of points as a starting point is perfectly fine. You are giving 
the software the specific input material from which to generate the output. You 
cannot ask it research questions an expect a valid response because ChatGPT has 
no parameters for correctness. It only wants to give you a plausible sounding 
answer. It will give you an authoritative answer with no reference to anything 
resembling truth.

 

Doug Nix

d...@mac.com  

http://www.dougnix.net

(519) 729-5704

 

"All animals except man know that the ultimate joy of life is to enjoy it."  -- 
Samuel Butler

 

 





On Oct 10, 2023, at 14:15, Ralph McDiarmid mailto:rmm.priv...@gmail.com> > wrote:

 

Physicians have used AI (expert systems) in their offices for many years as a 
tool to help them diagnose a problem more accurately and will greater speed.  
It’s a tool to speed productivity, and that’s how I use it today.

 


-

This message is from the IEEE Product Safety Engineering Society emc-pstc 
discussion list. To post a message to the list, send your e-mail to 


All emc-pstc postings are archived and searchable on the web at:
https://www.mail-archive.com/emc-pstc@listserv.ieee.org/

Website:  https://ewh.ieee.org/soc/pses/
Instructions:  https://ewh.ieee.org/soc/pses/list.html (including how to 
unsubscribe)
List rules: https://ewh.ieee.org/soc/pses/listrules.html

For help, send mail to the list administrators:
Mike Sherman at: msherma...@comcast.net
Rick Linford at: linf...@ieee.org

For policy questions, send mail to:
Jim Bacher:  
_
To unsubscribe from the EMC-PSTC list, click the following link: 
https://listserv.ieee.org/cgi-bin/wa?SUBED1=EMC-PSTC&A=1

Re: [PSES] AI & Regulatory Compliance

[Douglas Nix]

You cannot compare ChatGPT to a medical AI that has been tailored for a 
specific task like medical image analysis. That’s like saying a 1968 Mini 
Cooper and Dodge 440 HEMI Charger are comparable because they are both cars. 
Nope.

Using ChatGPT to summarize a paper, produce an abstract from uploaded text, or 
produce a set of points as a starting point is perfectly fine. You are giving 
the software the specific input material from which to generate the output. You 
cannot ask it research questions an expect a valid response because ChatGPT has 
no parameters for correctness. It only wants to give you a plausible sounding 
answer. It will give you an authoritative answer with no reference to anything 
resembling truth.

Doug Nix
d...@mac.com
http://www.dougnix.net
(519) 729-5704

"All animals except man know that the ultimate joy of life is to enjoy it."  -- 
Samuel Butler



> On Oct 10, 2023, at 14:15, Ralph McDiarmid  wrote:
> 
> Physicians have used AI (expert systems) in their offices for many years as a 
> tool to help them diagnose a problem more accurately and will greater speed.  
> It’s a tool to speed productivity, and that’s how I use it today.


-

This message is from the IEEE Product Safety Engineering Society emc-pstc 
discussion list. To post a message to the list, send your e-mail to 


All emc-pstc postings are archived and searchable on the web at:
https://www.mail-archive.com/emc-pstc@listserv.ieee.org/

Website:  https://ewh.ieee.org/soc/pses/
Instructions:  https://ewh.ieee.org/soc/pses/list.html (including how to 
unsubscribe)
List rules: https://ewh.ieee.org/soc/pses/listrules.html

For help, send mail to the list administrators:
Mike Sherman at: msherma...@comcast.net
Rick Linford at: linf...@ieee.org

For policy questions, send mail to:
Jim Bacher:  
_
To unsubscribe from the EMC-PSTC list, click the following link: 
https://listserv.ieee.org/cgi-bin/wa?SUBED1=EMC-PSTC&A=1

Re: [PSES] AI & Regulatory Compliance

[Ralph McDiarmid]

Which is another way of saying that ChatGPT and other similar AI are tools used 
by professionals to assist with improving their productivity, but not something 
to be used in lieu of a professional.  (at least not in this decade).  That 
time may come, but I suspect engineering jobs are safe for the foreseeable 
future.

 

Ralph

 

From: Dan Roman <0d75e04ed751-dmarc-requ...@listserv.ieee.org> 
Sent: Tuesday, October 10, 2023 10:51 AM
To: EMC-PSTC@LISTSERV.IEEE.ORG
Subject: Re: [PSES] AI & Regulatory Compliance

 

Other than playing around with text queries the most useful thing I have found 
either Chat GPT or Bard (the two I have used regularly) good for is Python 
coding.  While I can never take the code as is, it does offer tips on libraries 
that I did not even know existing to more easily perform a task.  I think 
others have hit on a similar use case for presentations and other output.  The 
AI can be useful to give you some ideas that you might not have already had and 
help you think about a problem or task in a different way  As with my Python 
example, you can’t just take what it gives you, you have to do the work and 
perhaps use some clues or ideas from the AI.

 

Dan

 

 

From: Brian Gregory [mailto:brian_greg...@netzero.net] 
Sent: Tuesday, October 10, 2023 12:16 PM
To: EMC-PSTC@LISTSERV.IEEE.ORG  
Subject: Re: [PSES] AI & Regulatory Compliance

 

 

Chat GPT is essentially a BS generator.  A very smart friend, a very successful 
entrepreneur, finds it quite useful for writing add copy, which I think proved 
my point.

 

The one time I tried challenging GPT with a question on power factor with 
generator convention (where positive power generation is represented as 
negative), it failed miserably. 

 

Echoing Dan's point;  what other AI's are more suitable to electrical 
engineering concepts, including testing and reviewing lists of standards?  If I 
had to pick one question for an AI I think might qualify would be:  " does this 
requirement apply to a residential unit? "  

 

 Colorado Brian 



-- Forwarded Message --
From: Dan Roman <0d75e04ed751-dmarc-requ...@listserv.ieee.org 
 >
To: EMC-PSTC@LISTSERV.IEEE.ORG  
Subject: Re: [PSES] AI & Regulatory Compliance
Date: Mon, 9 Oct 2023 22:21:26 -0400

Has anyone tried feeding a standard or set of standards, into Chat GPT for 
example, and ask it product construction or testing questions? I've seen demos 
of Chat GPT digesting complex equipment manuals and being able to generate what 
amounts to a quick start guide and answer questions about operating or 
servicing the equipment. I can't try this with the free version.

--

Dan Roman

danp...@verizon.net  

On Oct 9, 2023, at 7:31 PM, Douglas Powell mailto:doug...@gmail.com> > wrote: 

I've done some playing around with ChatGPT, Claude.AI, Perplexity.AI, BIng 
Chatbot, and a few others.  

 

My first test was to see if these Large Language Models (LLM) could accurately 
answer questions where I feel I have good expertise, such as "How do I select a 
complete list of safety-critical components using UL XYZ?" or "Write a 
comprehensive Test Protocol for  a  kitchen appliance using IEC 
60335-*X-X." For the most part they failed in several points when the questions 
got just a little technical.  Sometimes, I would challenge the AI on answers it 
gave, and occasionally it would backpedal. 

 

Even so, I do find them useful as a starting point for writing White Papers and 
PowerPoint slides or procedural documents at the paragraph-level, for internal 
distribution.  Try prompting an AI to generate 20 PowerPoint slides for basic 
Risk Assessment Training using ISO 31000.  In every case however, I only use AI 
to do the initial grunt work and then personally edit everything for accuracy. 
So far, my favorite is Perplexity.AI because it gives attribution of sources 
for all its answers, second place is possibly Claude.AI.  And been using 
QuillBot.AI to check grammar and paraphrase complex paragraphs.  But when it 
comes to life-safety issues, I will always be very careful and check the work.

 

Incidentally, I've tried a few fun things as a lunchtime diversion.  I'll 
prompt the AI with something like: "Entropy simply isn’t what it used to be."  
or "Schrödinger’s cat walks into a bar. And doesn’t." And of course the AI 
never gets the joke.  

 

-Doug

 

Douglas E Powell

Laporte, Colorado, USA

  LinkedIn

 

(UTC-06:00, US-MDT)

 

 

 

 

 

On Fri, Oct 6, 2023 at 3:59 PM Regan Arndt mailto:reganar...@gmail.com> > wrote:

Greetings fellow members,

Our industry will not be immune to this new era of AI technology and want to 
bring up this topic again to obtain some more insight from the forum into:

"How do you think AI will play a future role in ou

Re: [PSES] AI & Regulatory Compliance

[Ralph McDiarmid]

I would add that about 9 out of 10 power engineers I’ve worked with over the 
years are equally confused with the “convention” of power factor and sign 
convention of real and reactive power flow.  Microelectronics engineers also 
often get the sign convention wrong when considering current source or sink 
from a logic gate or an op-amp.   I wouldn’t beat up on ChatGPT for struggling 
with that very specific topic.

 

Ralph

 

From: Brian Gregory  
Sent: Tuesday, October 10, 2023 9:16 AM
To: EMC-PSTC@LISTSERV.IEEE.ORG
Subject: Re: [PSES] AI & Regulatory Compliance

 

 

Chat GPT is essentially a BS generator.  A very smart friend, a very successful 
entrepreneur, finds it quite useful for writing add copy, which I think proved 
my point.

 

The one time I tried challenging GPT with a question on power factor with 
generator convention (where positive power generation is represented as 
negative), it failed miserably. 

 

Echoing Dan's point;  what other AI's are more suitable to electrical 
engineering concepts, including testing and reviewing lists of standards?  If I 
had to pick one question for an AI I think might qualify would be:  " does this 
requirement apply to a residential unit? "  

 

 Colorado Brian 



-- Forwarded Message --
From: Dan Roman <0d75e04ed751-dmarc-requ...@listserv.ieee.org 
 >
To: EMC-PSTC@LISTSERV.IEEE.ORG  
Subject: Re: [PSES] AI & Regulatory Compliance
Date: Mon, 9 Oct 2023 22:21:26 -0400

Has anyone tried feeding a standard or set of standards, into Chat GPT for 
example, and ask it product construction or testing questions? I've seen demos 
of Chat GPT digesting complex equipment manuals and being able to generate what 
amounts to a quick start guide and answer questions about operating or 
servicing the equipment. I can't try this with the free version.

--

Dan Roman

danp...@verizon.net  

On Oct 9, 2023, at 7:31 PM, Douglas Powell mailto:doug...@gmail.com> > wrote: 

I've done some playing around with ChatGPT, Claude.AI, Perplexity.AI, BIng 
Chatbot, and a few others.  

 

My first test was to see if these Large Language Models (LLM) could accurately 
answer questions where I feel I have good expertise, such as "How do I select a 
complete list of safety-critical components using UL XYZ?" or "Write a 
comprehensive Test Protocol for  a  kitchen appliance using IEC 
60335-*X-X." For the most part they failed in several points when the questions 
got just a little technical.  Sometimes, I would challenge the AI on answers it 
gave, and occasionally it would backpedal. 

 

Even so, I do find them useful as a starting point for writing White Papers and 
PowerPoint slides or procedural documents at the paragraph-level, for internal 
distribution.  Try prompting an AI to generate 20 PowerPoint slides for basic 
Risk Assessment Training using ISO 31000.  In every case however, I only use AI 
to do the initial grunt work and then personally edit everything for accuracy. 
So far, my favorite is Perplexity.AI because it gives attribution of sources 
for all its answers, second place is possibly Claude.AI.  And been using 
QuillBot.AI to check grammar and paraphrase complex paragraphs.  But when it 
comes to life-safety issues, I will always be very careful and check the work.

 

Incidentally, I've tried a few fun things as a lunchtime diversion.  I'll 
prompt the AI with something like: "Entropy simply isn’t what it used to be."  
or "Schrödinger’s cat walks into a bar. And doesn’t." And of course the AI 
never gets the joke.  

 

-Doug

 

Douglas E Powell

Laporte, Colorado, USA

  LinkedIn

 

(UTC-06:00, US-MDT)

 

 

 

 

 

On Fri, Oct 6, 2023 at 3:59 PM Regan Arndt mailto:reganar...@gmail.com> > wrote:

Greetings fellow members,

Our industry will not be immune to this new era of AI technology and want to 
bring up this topic again to obtain some more insight from the forum into:

"How do you think AI will play a future role in our Regulatory compliance 
world?"

 

I can see some real benefits to this when applying it to a risk assessment, EMC 
redesign including change out of critical components, & possibly to help 
support engineering judgments in lieu of testing (i.e. temperature modeling), 
etc.…though I think we would have to include the usage of AI as one of the 
potential risks in the risk assessment. lol

UL touched base on this topic during their UL Innovations summit held in 
Fremont, CA yesterday.

I know there is the ISO/IEC DTR 5469 in development, but this appears, I 
believe, to only address functional safety of the AI systems in question, and 
not addressing AI in normal compliance assessments for products under the 
category of laboratory, IT, audio, etc.

Also, is there someone out there that will be speaking to this during the 2024 
ISPCE in Chicago

Re: [PSES] AI & Regulatory Compliance

[Ralph McDiarmid]

Very interesting demo of its capabilities.

 

I am wondering if the $20 per month version of ChatGPT might be worthwhile.

 

Ralph

 

From: Dan Roman <0d75e04ed751-dmarc-requ...@listserv.ieee.org> 
Sent: Monday, October 9, 2023 7:21 PM
To: EMC-PSTC@LISTSERV.IEEE.ORG
Subject: Re: [PSES] AI & Regulatory Compliance

 

Has anyone tried feeding a standard or set of standards, into Chat GPT for 
example, and ask it product construction or testing questions? I've seen demos 
of Chat GPT digesting complex equipment manuals and being able to generate what 
amounts to a quick start guide and answer questions about operating or 
servicing the equipment. I can't try this with the free version.

--

Dan Roman

danp...@verizon.net  

On Oct 9, 2023, at 7:31 PM, Douglas Powell mailto:doug...@gmail.com> > wrote:

I've done some playing around with ChatGPT, Claude.AI, Perplexity.AI, BIng 
Chatbot, and a few others.  

 

My first test was to see if these Large Language Models (LLM) could accurately 
answer questions where I feel I have good expertise, such as "How do I select a 
complete list of safety-critical components using UL XYZ?" or "Write a 
comprehensive Test Protocol for  a  kitchen appliance using IEC 
60335-*X-X." For the most part they failed in several points when the questions 
got just a little technical.  Sometimes, I would challenge the AI on answers it 
gave, and occasionally it would backpedal. 

 

Even so, I do find them useful as a starting point for writing White Papers and 
PowerPoint slides or procedural documents at the paragraph-level, for internal 
distribution.  Try prompting an AI to generate 20 PowerPoint slides for basic 
Risk Assessment Training using ISO 31000.  In every case however, I only use AI 
to do the initial grunt work and then personally edit everything for accuracy. 
So far, my favorite is Perplexity.AI because it gives attribution of sources 
for all its answers, second place is possibly Claude.AI.  And been using 
QuillBot.AI to check grammar and paraphrase complex paragraphs.  But when it 
comes to life-safety issues, I will always be very careful and check the work.

 

Incidentally, I've tried a few fun things as a lunchtime diversion.  I'll 
prompt the AI with something like: "Entropy simply isn’t what it used to be."  
or "Schrödinger’s cat walks into a bar. And doesn’t." And of course the AI 
never gets the joke.  

 

-Doug

 

Douglas E Powell

Laporte, Colorado, USA

  LinkedIn

 

(UTC-06:00, US-MDT)

 

 

 

 

 

On Fri, Oct 6, 2023 at 3:59 PM Regan Arndt mailto:reganar...@gmail.com> > wrote:

Greetings fellow members,

Our industry will not be immune to this new era of AI technology and want to 
bring up this topic again to obtain some more insight from the forum into:

"How do you think AI will play a future role in our Regulatory compliance 
world?"

 

I can see some real benefits to this when applying it to a risk assessment, EMC 
redesign including change out of critical components, & possibly to help 
support engineering judgments in lieu of testing (i.e. temperature modeling), 
etc.…though I think we would have to include the usage of AI as one of the 
potential risks in the risk assessment. lol

UL touched base on this topic during their UL Innovations summit held in 
Fremont, CA yesterday.

I know there is the ISO/IEC DTR 5469 in development, but this appears, I 
believe, to only address functional safety of the AI systems in question, and 
not addressing AI in normal compliance assessments for products under the 
category of laboratory, IT, audio, etc.

Also, is there someone out there that will be speaking to this during the 2024 
ISPCE in Chicago next May? Perhaps there should be an IEEE PSES technical 
committee created for this?

Side note: Nordcloud puts out a good summary blog on this:  
 
https://nordcloud.com/blog/how-ai-can-help-you-obtain-regulatory-compliance/

There are other AI blogs out there in the ether but more tuned for the 
financial sector and other industries….

Looking forward to the discussion. 

 

P.S.

I am now the Chair for the IEEE PSES Risk assessment technical committee and am 
looking for any volunteers to join our group. This topic will be one to discuss 
soon within our group. Please email me at   
regan.ar...@ieee.org or   reganar...@gmail.com if 
you are interested in joining.

Cheers!

Regan Arndt


  _  


This message is from the IEEE Product Safety Engineering Society emc-pstc 
discussion list. To post a message to the list, send your e-mail to All 
emc-pstc postings are archived and searchable on the web at: 
https://www.mail-archive.com/emc-pstc@listserv.ieee.org/ 

Website: https://ewh.ieee.org/soc/pses/   
Instructions: https://ewh.ieee.org/soc/pses/l

Re: [PSES] AI & Regulatory Compliance

[Ralph McDiarmid]

“Even so, I do find them useful as a starting point for writing White Papers 
and PowerPoint slides or procedural documents at the paragraph-level, for 
internal distribution. “

 

I do the same.  It’s useful now, even though not perfect, and AI is only going 
to get better and that will happen rapidly.

 

Physicians have used AI (expert systems) in their offices for many years as a 
tool to help them diagnose a problem more accurately and will greater speed.  
It’s a tool to speed productivity, and that’s how I use it today.

 

Ralph

 

From: Douglas Powell  
Sent: Monday, October 9, 2023 4:30 PM
To: EMC-PSTC@LISTSERV.IEEE.ORG
Subject: Re: [PSES] AI & Regulatory Compliance

 

I've done some playing around with ChatGPT, Claude.AI, Perplexity.AI, BIng 
Chatbot, and a few others.  

 

My first test was to see if these Large Language Models (LLM) could accurately 
answer questions where I feel I have good expertise, such as "How do I select a 
complete list of safety-critical components using UL XYZ?" or "Write a 
comprehensive Test Protocol for  a  kitchen appliance using IEC 
60335-*X-X." For the most part they failed in several points when the questions 
got just a little technical.  Sometimes, I would challenge the AI on answers it 
gave, and occasionally it would backpedal. 

 

Even so, I do find them useful as a starting point for writing White Papers and 
PowerPoint slides or procedural documents at the paragraph-level, for internal 
distribution.  Try prompting an AI to generate 20 PowerPoint slides for basic 
Risk Assessment Training using ISO 31000.  In every case however, I only use AI 
to do the initial grunt work and then personally edit everything for accuracy. 
So far, my favorite is Perplexity.AI because it gives attribution of sources 
for all its answers, second place is possibly Claude.AI.  And been using 
QuillBot.AI to check grammar and paraphrase complex paragraphs.  But when it 
comes to life-safety issues, I will always be very careful and check the work.

 

Incidentally, I've tried a few fun things as a lunchtime diversion.  I'll 
prompt the AI with something like: "Entropy simply isn’t what it used to be."  
or "Schrödinger’s cat walks into a bar. And doesn’t." And of course the AI 
never gets the joke.  

 

-Doug

 

Douglas E Powell

Laporte, Colorado, USA

  LinkedIn

 

(UTC-06:00, US-MDT)

 

 

 

 

 

On Fri, Oct 6, 2023 at 3:59 PM Regan Arndt mailto:reganar...@gmail.com> > wrote:

Greetings fellow members,

Our industry will not be immune to this new era of AI technology and want to 
bring up this topic again to obtain some more insight from the forum into:

"How do you think AI will play a future role in our Regulatory compliance 
world?"

 

I can see some real benefits to this when applying it to a risk assessment, EMC 
redesign including change out of critical components, & possibly to help 
support engineering judgments in lieu of testing (i.e. temperature modeling), 
etc.…though I think we would have to include the usage of AI as one of the 
potential risks in the risk assessment. lol

UL touched base on this topic during their UL Innovations summit held in 
Fremont, CA yesterday.

I know there is the ISO/IEC DTR 5469 in development, but this appears, I 
believe, to only address functional safety of the AI systems in question, and 
not addressing AI in normal compliance assessments for products under the 
category of laboratory, IT, audio, etc.

Also, is there someone out there that will be speaking to this during the 2024 
ISPCE in Chicago next May? Perhaps there should be an IEEE PSES technical 
committee created for this?

Side note: Nordcloud puts out a good summary blog on this:  
 
https://nordcloud.com/blog/how-ai-can-help-you-obtain-regulatory-compliance/

There are other AI blogs out there in the ether but more tuned for the 
financial sector and other industries….

Looking forward to the discussion. 

 

P.S.

I am now the Chair for the IEEE PSES Risk assessment technical committee and am 
looking for any volunteers to join our group. This topic will be one to discuss 
soon within our group. Please email me at   
regan.ar...@ieee.org or   reganar...@gmail.com if 
you are interested in joining.

Cheers!

Regan Arndt


  _  


This message is from the IEEE Product Safety Engineering Society emc-pstc 
discussion list. To post a message to the list, send your e-mail to All 
emc-pstc postings are archived and searchable on the web at: 
https://www.mail-archive.com/emc-pstc@listserv.ieee.org/ 

Website: https://ewh.ieee.org/soc/pses/   
Instructions: https://ewh.ieee.org/soc/pses/list.html (including how to 
unsubscribe)  
List rules: https://ewh.ieee.org/soc/pses/listrules.html 

For help, s

Re: [PSES] AI & Regulatory Compliance

[Dan Roman]

Other than playing around with text queries the most useful thing I have found 
either Chat GPT or Bard (the two I have used regularly) good for is Python 
coding.  While I can never take the code as is, it does offer tips on libraries 
that I did not even know existing to more easily perform a task.  I think 
others have hit on a similar use case for presentations and other output.  The 
AI can be useful to give you some ideas that you might not have already had and 
help you think about a problem or task in a different way  As with my Python 
example, you can’t just take what it gives you, you have to do the work and 
perhaps use some clues or ideas from the AI.

 

Dan

 

 

From: Brian Gregory [mailto:brian_greg...@netzero.net] 
Sent: Tuesday, October 10, 2023 12:16 PM
To: EMC-PSTC@LISTSERV.IEEE.ORG
Subject: Re: [PSES] AI & Regulatory Compliance

 

 

Chat GPT is essentially a BS generator.  A very smart friend, a very successful 
entrepreneur, finds it quite useful for writing add copy, which I think proved 
my point.

 

The one time I tried challenging GPT with a question on power factor with 
generator convention (where positive power generation is represented as 
negative), it failed miserably. 

 

Echoing Dan's point;  what other AI's are more suitable to electrical 
engineering concepts, including testing and reviewing lists of standards?  If I 
had to pick one question for an AI I think might qualify would be:  " does this 
requirement apply to a residential unit? "  

 

 Colorado Brian 



-- Forwarded Message --
From: Dan Roman <0d75e04ed751-dmarc-requ...@listserv.ieee.org 
 >
To: EMC-PSTC@LISTSERV.IEEE.ORG  
Subject: Re: [PSES] AI & Regulatory Compliance
Date: Mon, 9 Oct 2023 22:21:26 -0400

Has anyone tried feeding a standard or set of standards, into Chat GPT for 
example, and ask it product construction or testing questions? I've seen demos 
of Chat GPT digesting complex equipment manuals and being able to generate what 
amounts to a quick start guide and answer questions about operating or 
servicing the equipment. I can't try this with the free version.

--

Dan Roman

danp...@verizon.net  

On Oct 9, 2023, at 7:31 PM, Douglas Powell mailto:doug...@gmail.com> > wrote: 

I've done some playing around with ChatGPT, Claude.AI, Perplexity.AI, BIng 
Chatbot, and a few others.  

 

My first test was to see if these Large Language Models (LLM) could accurately 
answer questions where I feel I have good expertise, such as "How do I select a 
complete list of safety-critical components using UL XYZ?" or "Write a 
comprehensive Test Protocol for  a  kitchen appliance using IEC 
60335-*X-X." For the most part they failed in several points when the questions 
got just a little technical.  Sometimes, I would challenge the AI on answers it 
gave, and occasionally it would backpedal. 

 

Even so, I do find them useful as a starting point for writing White Papers and 
PowerPoint slides or procedural documents at the paragraph-level, for internal 
distribution.  Try prompting an AI to generate 20 PowerPoint slides for basic 
Risk Assessment Training using ISO 31000.  In every case however, I only use AI 
to do the initial grunt work and then personally edit everything for accuracy. 
So far, my favorite is Perplexity.AI because it gives attribution of sources 
for all its answers, second place is possibly Claude.AI.  And been using 
QuillBot.AI to check grammar and paraphrase complex paragraphs.  But when it 
comes to life-safety issues, I will always be very careful and check the work.

 

Incidentally, I've tried a few fun things as a lunchtime diversion.  I'll 
prompt the AI with something like: "Entropy simply isn’t what it used to be."  
or "Schrödinger’s cat walks into a bar. And doesn’t." And of course the AI 
never gets the joke.  

 

-Doug

 

Douglas E Powell

Laporte, Colorado, USA

  LinkedIn

 

(UTC-06:00, US-MDT)

 

 

 

 

 

On Fri, Oct 6, 2023 at 3:59 PM Regan Arndt mailto:reganar...@gmail.com> > wrote:

Greetings fellow members,

Our industry will not be immune to this new era of AI technology and want to 
bring up this topic again to obtain some more insight from the forum into:

"How do you think AI will play a future role in our Regulatory compliance 
world?"

 

I can see some real benefits to this when applying it to a risk assessment, EMC 
redesign including change out of critical components, & possibly to help 
support engineering judgments in lieu of testing (i.e. temperature modeling), 
etc.…though I think we would have to include the usage of AI as one of the 
potential risks in the risk assessment. lol

UL touched base on this topic during their UL Innovations summit held in 
Fremont, CA yesterday.

I know there is the ISO/IEC DTR 5469 in development, but this appears, I 
b

Re: [PSES] AI & Regulatory Compliance

[Brian Gregory]

 Chat GPT is essentially a BS generator.  A very smart friend, a very 
successful entrepreneur, finds it quite useful for writing add copy, which I 
think proved my point. The one time I tried challenging GPT with a question on 
power factor with generator convention (where positive power generation is 
represented as negative), it failed miserably.  Echoing Dan's point;  what 
other AI's are more suitable to electrical engineering concepts, including 
testing and reviewing lists of standards?  If I had to pick one question for an 
AI I think might qualify would be:  " does this requirement apply to a 
residential unit? "Colorado Brian 

-- Forwarded Message --
From: Dan Roman <0d75e04ed751-dmarc-requ...@listserv.ieee.org>
To: EMC-PSTC@LISTSERV.IEEE.ORG
Subject: Re: [PSES] AI & Regulatory Compliance
Date: Mon, 9 Oct 2023 22:21:26 -0400


Has anyone tried feeding a standard or set of standards, into Chat GPT for 
example, and ask it product construction or testing questions? I've seen demos 
of Chat GPT digesting complex equipment manuals and being able to generate what 
amounts to a quick start guide and answer questions about operating or 
servicing the equipment. I can't try this with the free version.

--Dan Romandanpses@verizon.netOn Oct 9, 2023, at 7:31 PM, Douglas Powell 
 wrote:I've done some playing around with ChatGPT, 
Claude.AI, Perplexity.AI, BIng Chatbot, and a few others.   My first test was 
to see if these Large Language Models (LLM) could accurately answer questions 
where I feel I have good expertise, such as "How do I select a complete list of 
safety-critical components using UL XYZ?" or "Write a comprehensive Test 
Protocol for  a  kitchen appliance using IEC 60335-*X-X." For the most part 
they failed in several points when the questions got just a little technical.  
Sometimes, I would challenge the AI on answers it gave, and occasionally it 
would backpedal.  Even so, I do find them useful as a starting point for 
writing White Papers and PowerPoint slides or procedural documents at the 
paragraph-level, for internal distribution.  Try prompting an AI to generate 20 
PowerPoint slides for basic Risk Assessment Training using ISO 31000.  In every 
case however, I only use AI to do the initial grunt work and then personally 
edit everything for accuracy. So far, my favorite is Perplexity.AI because it 
gives attribution of sources for all its answers, second place is possibly 
Claude.AI.  And been using QuillBot.AI to check grammar and paraphrase complex 
paragraphs.  But when it comes to life-safety issues, I will always be very 
careful and check the work. Incidentally, I've tried a few fun things as a 
lunchtime diversion.  I'll prompt the AI with something like: "Entropy simply 
isn’t what it used to be."  or "Schrödinger’s cat walks into a bar. 
And doesn’t." And of course the AI never gets the joke.   -Doug
 Douglas E PowellLaporte, Colorado, USALinkedIn (UTC-06:00, US-MDT)
On Fri, Oct 6, 2023 at 3:59 PM Regan Arndt  
wrote:Greetings fellow members,
Our industry will not be immune to this new era of AI technology and want to 
bring up this topic again to obtain some more insight from the forum into:
"How do you think AI will play a future role in our Regulatory compliance 
world?"
 
I can see some real benefits to this when applying it to a risk assessment, EMC 
redesign including change out of critical components, & possibly to help 
support engineering judgments in lieu of testing (i.e. temperature modeling), 
etc.…though I think we would have to include the usage of AI as one of 
the potential risks in the risk assessment. lol
UL touched base on this topic during their UL Innovations summit held in 
Fremont, CA yesterday.
I know there is the ISO/IEC DTR 5469 in development, but this appears, I 
believe, to only address functional safety of the AI systems in question, and 
not addressing AI in normal compliance assessments for products under the 
category of laboratory, IT, audio, etc.
Also, is there someone out there that will be speaking to this during the 2024 
ISPCE in Chicago next May? Perhaps there should be an IEEE PSES technical 
committee created for this?
Side note: Nordcloud puts out a good summary blog on this: 
https://nordcloud.com/blog/how-ai-can-help-you-obtain-regulatory-compliance/
There are other AI blogs out there in the ether but more tuned for the 
financial sector and other industries….
Looking forward to the discussion.
 
P.S.
I am now the Chair for the IEEE PSES Risk assessment technical committee and am 
looking for any volunteers to join our group. This topic will be one to discuss 
soon within our group. Please email me at regan.ar...@ieee.org or 
reganar...@gmail.com if you are interested in joining.
Cheers!
Regan Arndt
This message is from the IEEE Product Safety Engineering Society emc-pstc 
discussion list. To post a message to the list, send your e-mail to All 
emc-pstc postings are archived and searchable on the

Re: [PSES] AW: [PSES] Directives that require the CE Marking

[Charlie Blackham]

Me again 😊

The opening sentence tells you to tread carefully:

The CE mark is a certification mark that indicates conformity with health, 
safety, and environmental protection standards for products sold within the 
European Economic Area (EEA


  *   It’s not a “certification” mark
  *   Not all CE marking Directives cover safety or health
  *   There are no CE marking Directives that cover environmental protection
  *   The rules don’t apply to products “sold”, they apply to products “placed 
on the market” – there doesn’t need to be a monetary exchange, and the lack of 
one does not put you out of scope

Best regards
Charlie

Charlie Blackham
Sulis Consultants Ltd
Tel: +44 (0)7946 624317
Web: https://sulisconsultants.com/
Registered in England and Wales, number 05466247

From: rmm.priv...@gmail.com 
Sent: Sunday, October 8, 2023 4:56 PM
To: Charlie Blackham 
Cc: EMC-PSTC@LISTSERV.IEEE.ORG
Subject: RE: [PSES] AW: [PSES] Directives that require the CE Marking

I agree that tools like ChatGPT have a long way to go before their response can 
be trusted without question, but the response below is from the free version.  
The $20 a month version might provide a clearer answer.  I use ChatGPT as a 
starting point for topics I have some familiarity with, then I drill down on 
the details.

I asked ChatGPT “under what conditions should I not apply the CE Mark ?”, and 
it answered:

=
The CE mark is a certification mark that indicates conformity with health, 
safety, and environmental protection standards for products sold within the 
European Economic Area (EEA).  It demonstrates that a product complies with the 
relevant EU directives and regulations. However, there are certain conditions 
or situations where you should not apply the CE mark:

  1.  Non-conforming products: If your product does not meet the essential 
requirements or standards specified by the relevant EU directives or 
regulations, you should not apply the CE mark. It is essential to ensure that 
your product complies with all applicable requirements before affixing the CE 
mark.
  2.  Non-covered products: Some products or product categories may not be 
covered by specific EU directives or regulations, and therefore, the CE mark 
may not be applicable to them. Always check whether your product falls under 
any relevant EU legislation before considering the CE mark.
  3.  Self-declaration without applicable legislation: In some cases, a 
manufacturer may incorrectly self-declare conformity and affix the CE mark to a 
product without any applicable EU directive or regulation. This is not allowed, 
and the CE mark should only be applied when there is clear and relevant 
legislation to comply with.
  4.  Non-EEA sales: The CE mark is primarily intended for products intended 
for sale within the EEA. If your product is not intended for the EEA market, 
applying the CE mark may not be necessary. However, you should comply with the 
regulatory requirements of the specific market where you intend to sell your 
product.
  5.  Unauthorized use: The CE mark should only be applied by manufacturers or 
authorized representatives who have fulfilled the necessary conformity 
assessment procedures and legal obligations. Unauthorized use of the CE mark is 
not allowed and can result in legal consequences.
  6.  False or misleading information: Providing false or misleading 
information, documentation, or test results to obtain the CE mark is illegal 
and should not be done under any circumstances.

It's crucial to consult with regulatory authorities, experts, or legal counsel 
knowledgeable about the specific product and its applicable regulations to 
ensure compliance with CE marking requirements. Incorrectly applying the CE 
mark can lead to legal consequences, including fines and product recalls.
==

Ralph

From: Charlie Blackham 
mailto:char...@sulisconsultants.com>>
Sent: Saturday, October 7, 2023 1:17 AM
To: rmm.priv...@gmail.com; 
EMC-PSTC@LISTSERV.IEEE.ORG
Subject: RE: [PSES] AW: [PSES] Directives that require the CE Marking

A nice example of why ChatGPT won’t put us all out of a job just yet 😊

Point 9 is wrong – you may not use the CE mark to demonstrate compliance with a 
“non CE marking” Directive, that is one that does not contain the provision for 
CE marking within its Articles or Annexes

Point 3 does not make it clear that the Directive also includes products that 
are “likely, under reasonably foreseeable conditions, to be used by consumers 
even if not intended for them” – an example of this would be something like a 
corporate branded battery pack phone charger which is supplied only on a B2B 
basis, but is likely to be taken home by the recipient

Best regards
Charlie

Charlie Blackham
Sulis Consultants Ltd
Tel: