IBM Z and LinuxONE IBM Z

The initial question was raised as:

"have you ever been involved in mainframe modernization I wanted to learn like how to use mainframe with GEN AI the modern tech; is there any demo or use case available where we can try and test".

I spoke to Amit who asked me the question and we exploded the question into.

1. Generative AI (genAI) is a broad brush definition to describe any type of artificial intelligence (AI) that is capable of producing or generating text, images, video, or audio clips. {different approaches using AI to consider}.

2. Generative AI (genAI), learns using patterns gained from training data to create new, unique outputs with the same statistical properties, which means that this is quite an open area {any tips on training data or where to find it, how to work with it}.

3. Are there demo's available systems that can be used to build and familiarise yourself with Generative AI (genAI) technology.

4. What would be the best use cases for this: here I can think of many examples of where you may want to apply Generative AI (genAI) such as SMF data, log data, error logs, against the JIRA database, capacity planning. I think there are too many use cases and unless as a community there is a really strong use case, a disclaimed would be "Generative AI (genAI) can be used in a variety of different ways based on user needs".

5. Would ZOWE / Open mainframe project be another source of inspiration or is this outside of scope?


Those questions arose after a quick follow-up call with Amit, I know this has been quite a hot topic as I attended a briefing on an almost identical presentation to the IBM Champions community, I am curious what advice we can offer or suggest.

Is there anything I have missed as I was heading to the hospital and didn't have time to do a really deep dive with Amit but I believe I captured with him the essence of what he is looking for in our quick chat.

Thank you fellow IBM Champions :)

Hello fellow IBM Champions....

I have had a quick look into this topic and I have found two free Coursera courses from IBM

&

Google suggests a wealth of free information is available, I have suggested to look at Google for now.

If anyone has any specific Mainframe suggestions I am happy to add them. A quick dig on this topic (GenAI) shows it is quite a wide topic, my advice to Amit is to look on Google and narrow his search via the Google platform.


Please feel free to tell me different if you think this advice could be improved :)

The initial question was raised as:

"have you ever been involved in mainframe modernization I wanted to learn like how to use mainframe with GEN AI the modern tech; is there any demo or use case available where we can try and test".

I spoke to Amit who asked me the question and we exploded the question into.

1. Generative AI (genAI) is a broad brush definition to describe any type of artificial intelligence (AI) that is capable of producing or generating text, images, video, or audio clips. {different approaches using AI to consider}.

2. Generative AI (genAI), learns using patterns gained from training data to create new, unique outputs with the same statistical properties, which means that this is quite an open area {any tips on training data or where to find it, how to work with it}.

3. Are there demo's available systems that can be used to build and familiarise yourself with Generative AI (genAI) technology.

4. What would be the best use cases for this: here I can think of many examples of where you may want to apply Generative AI (genAI) such as SMF data, log data, error logs, against the JIRA database, capacity planning. I think there are too many use cases and unless as a community there is a really strong use case, a disclaimed would be "Generative AI (genAI) can be used in a variety of different ways based on user needs".

5. Would ZOWE / Open mainframe project be another source of inspiration or is this outside of scope?


Those questions arose after a quick follow-up call with Amit, I know this has been quite a hot topic as I attended a briefing on an almost identical presentation to the IBM Champions community, I am curious what advice we can offer or suggest.

Is there anything I have missed as I was heading to the hospital and didn't have time to do a really deep dive with Amit but I believe I captured with him the essence of what he is looking for in our quick chat.

Thank you fellow IBM Champions :)

I see that there's not been a lot of take-up on this, I'll venture a few points:

• The question is not very well formed.  It seems akin to "I hear there's a cool new energy called electricity.  Are there any use cases I can try it out on?"
• There are lots of "free" implementations of Generative AI out there right now, e.g. ChatGPT, Claude, DALL-E, that can answer questions, respond to requests, etc.
• A specific mainframe use case is "IBM watsonx Code Assistant for Z" which assists a developer in understanding and refactoring COBOL and, where desired, helps re-implement refactored COBOL in Java.  This is not generative AI "on the mainframe," it is generative AI "for the mainframe."
• If you are interested in potential use cases for generative AI, I'd suggest an interwebz search for "use cases for generative ai."  There are many articles, blogs, etc. on the subject.
• In terms of doing generative AI "on the mainframe," my personal opinion is that, without a compelling reason to run it on a mainframe, it's better run on commodity (read that as "cheap") hardware.  That doesn't mean that I wouldn't run other forms of AI on a mainframe, where speed and latency matter, just that generative AI tends to have a voracious appetite for data and CPU (and GPU) when it's in use.  I'm of the opinion that "in-transaction" use cases (like classification - is this transaction likely to be fraudulent?) are going to win the day for a while "on the mainframe."

Hello fellow IBM Champions....

I have had a quick look into this topic and I have found two free Coursera courses from IBM

&

Google suggests a wealth of free information is available, I have suggested to look at Google for now.

If anyone has any specific Mainframe suggestions I am happy to add them. A quick dig on this topic (GenAI) shows it is quite a wide topic, my advice to Amit is to look on Google and narrow his search via the Google platform.


Please feel free to tell me different if you think this advice could be improved :)

The initial question was raised as:

"have you ever been involved in mainframe modernization I wanted to learn like how to use mainframe with GEN AI the modern tech; is there any demo or use case available where we can try and test".

I spoke to Amit who asked me the question and we exploded the question into.

1. Generative AI (genAI) is a broad brush definition to describe any type of artificial intelligence (AI) that is capable of producing or generating text, images, video, or audio clips. {different approaches using AI to consider}.

2. Generative AI (genAI), learns using patterns gained from training data to create new, unique outputs with the same statistical properties, which means that this is quite an open area {any tips on training data or where to find it, how to work with it}.

3. Are there demo's available systems that can be used to build and familiarise yourself with Generative AI (genAI) technology.

4. What would be the best use cases for this: here I can think of many examples of where you may want to apply Generative AI (genAI) such as SMF data, log data, error logs, against the JIRA database, capacity planning. I think there are too many use cases and unless as a community there is a really strong use case, a disclaimed would be "Generative AI (genAI) can be used in a variety of different ways based on user needs".

5. Would ZOWE / Open mainframe project be another source of inspiration or is this outside of scope?


Those questions arose after a quick follow-up call with Amit, I know this has been quite a hot topic as I attended a briefing on an almost identical presentation to the IBM Champions community, I am curious what advice we can offer or suggest.

Is there anything I have missed as I was heading to the hospital and didn't have time to do a really deep dive with Amit but I believe I captured with him the essence of what he is looking for in our quick chat.

Thank you fellow IBM Champions :)

Thank you for the advice for Amit.

I had drafted something a similar reply to yourself, in my reply to Amit I have included both messages, we both reached the same conclusion that the request would have benefited from being more specific in its nature.

Whilst the question wasn't great it's good to see people asking questions as it gives you an idea of what others in our community are thinking about with their mainframe. 

I see that there's not been a lot of take-up on this, I'll venture a few points:

• The question is not very well formed.  It seems akin to "I hear there's a cool new energy called electricity.  Are there any use cases I can try it out on?"
• There are lots of "free" implementations of Generative AI out there right now, e.g. ChatGPT, Claude, DALL-E, that can answer questions, respond to requests, etc.
• A specific mainframe use case is "IBM watsonx Code Assistant for Z" which assists a developer in understanding and refactoring COBOL and, where desired, helps re-implement refactored COBOL in Java.  This is not generative AI "on the mainframe," it is generative AI "for the mainframe."
• If you are interested in potential use cases for generative AI, I'd suggest an interwebz search for "use cases for generative ai."  There are many articles, blogs, etc. on the subject.
• In terms of doing generative AI "on the mainframe," my personal opinion is that, without a compelling reason to run it on a mainframe, it's better run on commodity (read that as "cheap") hardware.  That doesn't mean that I wouldn't run other forms of AI on a mainframe, where speed and latency matter, just that generative AI tends to have a voracious appetite for data and CPU (and GPU) when it's in use.  I'm of the opinion that "in-transaction" use cases (like classification - is this transaction likely to be fraudulent?) are going to win the day for a while "on the mainframe."

Hello fellow IBM Champions....

I have had a quick look into this topic and I have found two free Coursera courses from IBM

&

Google suggests a wealth of free information is available, I have suggested to look at Google for now.

If anyone has any specific Mainframe suggestions I am happy to add them. A quick dig on this topic (GenAI) shows it is quite a wide topic, my advice to Amit is to look on Google and narrow his search via the Google platform.


Please feel free to tell me different if you think this advice could be improved :)

The initial question was raised as:

"have you ever been involved in mainframe modernization I wanted to learn like how to use mainframe with GEN AI the modern tech; is there any demo or use case available where we can try and test".

I spoke to Amit who asked me the question and we exploded the question into.

1. Generative AI (genAI) is a broad brush definition to describe any type of artificial intelligence (AI) that is capable of producing or generating text, images, video, or audio clips. {different approaches using AI to consider}.

2. Generative AI (genAI), learns using patterns gained from training data to create new, unique outputs with the same statistical properties, which means that this is quite an open area {any tips on training data or where to find it, how to work with it}.

3. Are there demo's available systems that can be used to build and familiarise yourself with Generative AI (genAI) technology.

4. What would be the best use cases for this: here I can think of many examples of where you may want to apply Generative AI (genAI) such as SMF data, log data, error logs, against the JIRA database, capacity planning. I think there are too many use cases and unless as a community there is a really strong use case, a disclaimed would be "Generative AI (genAI) can be used in a variety of different ways based on user needs".

5. Would ZOWE / Open mainframe project be another source of inspiration or is this outside of scope?


Those questions arose after a quick follow-up call with Amit, I know this has been quite a hot topic as I attended a briefing on an almost identical presentation to the IBM Champions community, I am curious what advice we can offer or suggest.

Is there anything I have missed as I was heading to the hospital and didn't have time to do a really deep dive with Amit but I believe I captured with him the essence of what he is looking for in our quick chat.

Thank you fellow IBM Champions :)

It's not "just" about speed or performance, but also about privacy and security. Suppose the bank is using Generative AI to craft you a nice explanation about why your loan is not approved. It would probably have to use information about your account, your credit history, etc. A secure architecture would be much simpler to implement when it doesn't involve external services that need to access your confidential information. It would then send the composed letter through approved secure channels to print or publish to the client.

Thank you for the advice for Amit.

I had drafted something a similar reply to yourself, in my reply to Amit I have included both messages, we both reached the same conclusion that the request would have benefited from being more specific in its nature.

Whilst the question wasn't great it's good to see people asking questions as it gives you an idea of what others in our community are thinking about with their mainframe. 

I see that there's not been a lot of take-up on this, I'll venture a few points:

• The question is not very well formed.  It seems akin to "I hear there's a cool new energy called electricity.  Are there any use cases I can try it out on?"
• There are lots of "free" implementations of Generative AI out there right now, e.g. ChatGPT, Claude, DALL-E, that can answer questions, respond to requests, etc.
• A specific mainframe use case is "IBM watsonx Code Assistant for Z" which assists a developer in understanding and refactoring COBOL and, where desired, helps re-implement refactored COBOL in Java.  This is not generative AI "on the mainframe," it is generative AI "for the mainframe."
• If you are interested in potential use cases for generative AI, I'd suggest an interwebz search for "use cases for generative ai."  There are many articles, blogs, etc. on the subject.
• In terms of doing generative AI "on the mainframe," my personal opinion is that, without a compelling reason to run it on a mainframe, it's better run on commodity (read that as "cheap") hardware.  That doesn't mean that I wouldn't run other forms of AI on a mainframe, where speed and latency matter, just that generative AI tends to have a voracious appetite for data and CPU (and GPU) when it's in use.  I'm of the opinion that "in-transaction" use cases (like classification - is this transaction likely to be fraudulent?) are going to win the day for a while "on the mainframe."

Hello fellow IBM Champions....

I have had a quick look into this topic and I have found two free Coursera courses from IBM

&

Google suggests a wealth of free information is available, I have suggested to look at Google for now.

If anyone has any specific Mainframe suggestions I am happy to add them. A quick dig on this topic (GenAI) shows it is quite a wide topic, my advice to Amit is to look on Google and narrow his search via the Google platform.


Please feel free to tell me different if you think this advice could be improved :)

The initial question was raised as:

"have you ever been involved in mainframe modernization I wanted to learn like how to use mainframe with GEN AI the modern tech; is there any demo or use case available where we can try and test".

I spoke to Amit who asked me the question and we exploded the question into.

1. Generative AI (genAI) is a broad brush definition to describe any type of artificial intelligence (AI) that is capable of producing or generating text, images, video, or audio clips. {different approaches using AI to consider}.

2. Generative AI (genAI), learns using patterns gained from training data to create new, unique outputs with the same statistical properties, which means that this is quite an open area {any tips on training data or where to find it, how to work with it}.

3. Are there demo's available systems that can be used to build and familiarise yourself with Generative AI (genAI) technology.

4. What would be the best use cases for this: here I can think of many examples of where you may want to apply Generative AI (genAI) such as SMF data, log data, error logs, against the JIRA database, capacity planning. I think there are too many use cases and unless as a community there is a really strong use case, a disclaimed would be "Generative AI (genAI) can be used in a variety of different ways based on user needs".

5. Would ZOWE / Open mainframe project be another source of inspiration or is this outside of scope?


Those questions arose after a quick follow-up call with Amit, I know this has been quite a hot topic as I attended a briefing on an almost identical presentation to the IBM Champions community, I am curious what advice we can offer or suggest.

Is there anything I have missed as I was heading to the hospital and didn't have time to do a really deep dive with Amit but I believe I captured with him the essence of what he is looking for in our quick chat.

Thank you fellow IBM Champions :)

I'd argue that horse has left the barn a loooooooooong time ago.  Much of what you describe has some its "system of record" data maintained on z/OS, but much of the other processing (e.g., credit checks, other customer suitability analysis) has long since been peeled off and put on distributed/cloud (primarily for the cost and low speed requirements I mentioned before).  To move it back to Z, you'd have to make a pretty compelling argument that their hybrid infrastructure is somehow not secure enough and that the cost is worth it.

It's not "just" about speed or performance, but also about privacy and security. Suppose the bank is using Generative AI to craft you a nice explanation about why your loan is not approved. It would probably have to use information about your account, your credit history, etc. A secure architecture would be much simpler to implement when it doesn't involve external services that need to access your confidential information. It would then send the composed letter through approved secure channels to print or publish to the client.

Thank you for the advice for Amit.

I had drafted something a similar reply to yourself, in my reply to Amit I have included both messages, we both reached the same conclusion that the request would have benefited from being more specific in its nature.

Whilst the question wasn't great it's good to see people asking questions as it gives you an idea of what others in our community are thinking about with their mainframe. 

I see that there's not been a lot of take-up on this, I'll venture a few points:

• The question is not very well formed.  It seems akin to "I hear there's a cool new energy called electricity.  Are there any use cases I can try it out on?"
• There are lots of "free" implementations of Generative AI out there right now, e.g. ChatGPT, Claude, DALL-E, that can answer questions, respond to requests, etc.
• A specific mainframe use case is "IBM watsonx Code Assistant for Z" which assists a developer in understanding and refactoring COBOL and, where desired, helps re-implement refactored COBOL in Java.  This is not generative AI "on the mainframe," it is generative AI "for the mainframe."
• If you are interested in potential use cases for generative AI, I'd suggest an interwebz search for "use cases for generative ai."  There are many articles, blogs, etc. on the subject.
• In terms of doing generative AI "on the mainframe," my personal opinion is that, without a compelling reason to run it on a mainframe, it's better run on commodity (read that as "cheap") hardware.  That doesn't mean that I wouldn't run other forms of AI on a mainframe, where speed and latency matter, just that generative AI tends to have a voracious appetite for data and CPU (and GPU) when it's in use.  I'm of the opinion that "in-transaction" use cases (like classification - is this transaction likely to be fraudulent?) are going to win the day for a while "on the mainframe."

Hello fellow IBM Champions....

I have had a quick look into this topic and I have found two free Coursera courses from IBM

&

Google suggests a wealth of free information is available, I have suggested to look at Google for now.

If anyone has any specific Mainframe suggestions I am happy to add them. A quick dig on this topic (GenAI) shows it is quite a wide topic, my advice to Amit is to look on Google and narrow his search via the Google platform.


Please feel free to tell me different if you think this advice could be improved :)

The initial question was raised as:

"have you ever been involved in mainframe modernization I wanted to learn like how to use mainframe with GEN AI the modern tech; is there any demo or use case available where we can try and test".

I spoke to Amit who asked me the question and we exploded the question into.

1. Generative AI (genAI) is a broad brush definition to describe any type of artificial intelligence (AI) that is capable of producing or generating text, images, video, or audio clips. {different approaches using AI to consider}.

2. Generative AI (genAI), learns using patterns gained from training data to create new, unique outputs with the same statistical properties, which means that this is quite an open area {any tips on training data or where to find it, how to work with it}.

3. Are there demo's available systems that can be used to build and familiarise yourself with Generative AI (genAI) technology.

4. What would be the best use cases for this: here I can think of many examples of where you may want to apply Generative AI (genAI) such as SMF data, log data, error logs, against the JIRA database, capacity planning. I think there are too many use cases and unless as a community there is a really strong use case, a disclaimed would be "Generative AI (genAI) can be used in a variety of different ways based on user needs".

5. Would ZOWE / Open mainframe project be another source of inspiration or is this outside of scope?


Those questions arose after a quick follow-up call with Amit, I know this has been quite a hot topic as I attended a briefing on an almost identical presentation to the IBM Champions community, I am curious what advice we can offer or suggest.

Is there anything I have missed as I was heading to the hospital and didn't have time to do a really deep dive with Amit but I believe I captured with him the essence of what he is looking for in our quick chat.

Thank you fellow IBM Champions :)