I would like to say that we should be positive but also cautious on AI. In one of my PHA, we had some technical questions such as flash points, etc. I ask ChatGPT for help. In two days, it gave me two different answers. Only one of them is correct. So my takeaway is so far human cannot be too lazy to let AI to be SME. It may save you one hour but takes you another 3 for troubleshooting.
Although I agree with your first comment, I'm not so sure about the second. The summary provided by Chat GPT is a useful listing of 'management of change' related issues and , in my view, isn't limiting itself by use of the phrases 'wide range of activities' and 'including but not limited to'.
Your summary is really too limited and would lead (less informed) readers to think that 'MoC' doesn't apply to engineering / equipment changes, for example.
Yes, the program did leave some room for interpretation. Actually, I often consider MoC to be the ‘most important’ because virtually all accidents result from someone (either a designer or an operations person) taking the facility outside its safe conditions. Its why it is so important to define, as well as possible, just what those limits are, as we discuss in https://psmreport.substack.com/p/the-principles-of-process-safety.
I'm coming from the same direction in reality. For a number of years, I was responsible for an engineering 'MoC' process for a COMAH operator in the UK. Introduced (a few years later admittedly) after the 1974 Flixborough disaster. Our 'MoC' procedure pre-dated a more joined up approach to process safety and the introduction of a more holistic 'Process Safety Management System' (ref OHSA 1910.119 et al) where engineering 'MoC' is just an element. So, that's why I thought the original Chat GPT list (although open to some challenge) was helpful as it pointed to other areas that need consideration as part of 'change management'
Anyway, we can have a useful 'back and forth' about 'PSMS's I'm sure, but getting back to the crux of your post - Chat GPT, in this context, is helpful but not definitive. And needs to be treated with a certain amount of caution. I'm sure that your concerns in this regard are similar to mine - that it could be used to provide 'de facto' and definitive answers to complex, technical and safety related questions. Chat GPT is nowhere near that yet, in my my view.
I will probably start a second post on this topic. Another commenter noted that ChatGPT can be flat out wrong when objective information is called for. I was asked if AI can help emergency responders. The preliminary conclusion seems to be ‘not yet’.
I would like to say that we should be positive but also cautious on AI. In one of my PHA, we had some technical questions such as flash points, etc. I ask ChatGPT for help. In two days, it gave me two different answers. Only one of them is correct. So my takeaway is so far human cannot be too lazy to let AI to be SME. It may save you one hour but takes you another 3 for troubleshooting.
We have planned shutdown, during draining of flammable product fire catch up.
Is it process safety event?
Although I agree with your first comment, I'm not so sure about the second. The summary provided by Chat GPT is a useful listing of 'management of change' related issues and , in my view, isn't limiting itself by use of the phrases 'wide range of activities' and 'including but not limited to'.
Your summary is really too limited and would lead (less informed) readers to think that 'MoC' doesn't apply to engineering / equipment changes, for example.
The reality is that none of the elements are the most important because they constitute a system. They all link to one another.
Yes, the program did leave some room for interpretation. Actually, I often consider MoC to be the ‘most important’ because virtually all accidents result from someone (either a designer or an operations person) taking the facility outside its safe conditions. Its why it is so important to define, as well as possible, just what those limits are, as we discuss in https://psmreport.substack.com/p/the-principles-of-process-safety.
I'm coming from the same direction in reality. For a number of years, I was responsible for an engineering 'MoC' process for a COMAH operator in the UK. Introduced (a few years later admittedly) after the 1974 Flixborough disaster. Our 'MoC' procedure pre-dated a more joined up approach to process safety and the introduction of a more holistic 'Process Safety Management System' (ref OHSA 1910.119 et al) where engineering 'MoC' is just an element. So, that's why I thought the original Chat GPT list (although open to some challenge) was helpful as it pointed to other areas that need consideration as part of 'change management'
Anyway, we can have a useful 'back and forth' about 'PSMS's I'm sure, but getting back to the crux of your post - Chat GPT, in this context, is helpful but not definitive. And needs to be treated with a certain amount of caution. I'm sure that your concerns in this regard are similar to mine - that it could be used to provide 'de facto' and definitive answers to complex, technical and safety related questions. Chat GPT is nowhere near that yet, in my my view.
I will probably start a second post on this topic. Another commenter noted that ChatGPT can be flat out wrong when objective information is called for. I was asked if AI can help emergency responders. The preliminary conclusion seems to be ‘not yet’.