Welcome to your IBM i update for February/March 2025 – a snapshot of news and views on what’s been happening in the IBM i community in the past few weeks as reported by Proximity.
This month we visit the spiritual home of i-UG.
There, we learn about Hybrid AI. We also ponder what the ‘pending major announcement’ on IBM i could be and remind the community we only have months until the EOS for Power 9.
You can watch the video below or read the full article underneath.
i-UG February 2025
So, at the end of February, i-UG returned to its spiritual home of the Norton Grange hotel in Rochdale – a venue in which they’ve been ever present for (I think) the past three decades.
And, as always has been the tradition since its inception, started with tea, coffee and the obligatory bacon butties – all very good. As always, the i-UG team bought us a cracking schedule featuring expert speakers.
We kicked off the day with a ‘state of the nation’ presentation from the i-UG committee.
Here they explained that i-UG continues to go from strength to strength with new and returning delegates and continuing interest in their education programs – well done indeed to team i-UG.
Now, there were many topics that deserve your attention, but for the update, I’d like to concentrate on the conference keynote presentation on Hybrid AI.
This was presented by IBM Power Lead, Jay Diamond and principle, David Spurway – and I must say, it was a joy to see David back and well, with inimitable enthusiasm for all things IBM i.
So, in past IBM i Updates, we’ve discussed AI and the upcoming IBM i code assistant. But on the whole, discussions in the IBM i space have been mostly in and around the area of generative AI.
But, there’s more pillars besides generative AI that help provide better results, less hallucinations while providing a more focused business value model.
I’ll provide a summary here, BUT, should you wish to experience the full presentation and demonstration then you have another chance – join the i-UG team in London on the 13th March and experience David live.
You can find i-UG here – https://i-ug.co.uk/
Hybrid AI for the IBM i
David started by explaining just what Hybrid AI is. Referencing an article from Forbes, David explained that it’s a fusion of the three existing pillars of AI.
These are Neural (or traditional machine learning), Generative (of which includes LLM), and Symbolic (rules based reasoning).
By combining these three pillars, you engage a powerful toolset that better enables you to leverage AI, without the inherent weakness of the individual strands.
‘But that’s surely not possible Andy?” I hear you say… and yes, I’ve been one of those many players eulogising over the potential that AI can bring business.
Recently, I’ve spoken about the IBM i Coding Assistant and what that may bring to the table, but it’s worth taking a moment to look at the bigger picture.
For example, the IBM i Coding Assistant will come under the ‘Generative AI’ strand. You transact with this type of AI using human, conversational text to produce [in the case of a coding assistant] RPG code.
This will allow the user to ask, in plane english, for the assistant to write a program, resolve a problem or analyse code.
Using it’s collective knowledge [the LLM], it will generate programs to your heart’s content but one of the most significant limitations is in its propensity to generate seemingly plausible code that is actually incorrect, fabricated, or doesn’t function as intended, essentially “making up” parts of the code that aren’t based on real logic or requirements.
These, ‘hallucinations’ (as they’re called) causes a real risk when relying on AI code generation tools as experienced developers will be needed to carefully review and verify the generated code to avoid potential issues.
Neural AI, or deep learning is at the heart of this new wave of AI rollout. However, it’s expensive to train, demanding huge datasets and even larger processors.
It’s therefore not suitable for small or specialised environments and, as the old adage goes – garbage in, garbage out. If training is incomplete or the data biased, the model will reflect these in its outputs.
Traditional or Symbolic AI also has weakness. It’s rigid, tightly bound by the predefined rules presented.
Sure, if you’re running looking to automate business processes – this is a great step forward but it lacks adaptability and simple isn’t able to handle new or unexpected situations without new rules or [in the case of deep learning] further extensive training.
But by combining these three pillars, Hybrid AI can balance each individual weaknesses.
For example:
- ✅ Rule-based AI adds explainability and structure to machine learning.
- ✅ Machine learning adds pattern recognition and adaptability to rules-based AI.
- ✅ Generative AI adds creativity & automation, while RAG (Retrieval-Augmented Generation.. More on this later) prevents hallucinations.
Imagine it like your Sunday roast. Dinner is never complete without your meat and two veg.
Small language models (SLMs), Agentic and RAG for AI on the IBM I
And sticking with that Sunday Lunch metaphor, what is the ‘gravy’ that bring the dinner together? And are there any trimmings to make it even better.?
Well, yes there is plenty of gravy in boat. One such item was printed in an article published on LinkedIn by Jay Diamond [of whom introduced and prefaced the session], he explained that SLMs [Small Language Models] could fit better with more specialised applications, not only would they be quicker to train but they also consume less energy and are easier to be fine tuned for specialist tasks.
Next, is the big new talking point in AI technology and that with Agentic AI. This is the automation layer of Hybrid AI combining:
- ✅ Neural AI (Deep Learning, Generative AI) → To process text, voice, and images.
- ✅ Symbolic AI (Rule-Based Systems) → To ensure structured decision-making.
- ✅ Machine Learning (Predictive Models) → To adapt and improve over time.
- ✅ RAG (Retrieval-Augmented Generation) → To retrieve real-time, fact-based data before taking action.
Speaking of RAG, Retrieval-Augmented Generation is the bridge between traditional AI and generative AI, allowing LLMs and SLMs to integrate with structured enterprise data and real-time information sources…
To paraphrase David, he explained that RAG works similarly to human interaction… but more akin to woking with a consultant who is able to reference and retrieve information from external data sources rather than relying solely on pre-trained knowledge.
He went on to explain why RAG was needed. He explained that traditional Large and Small Language Models work like a closed book exam.
Here the AI only has access to pre-trained data, similar to a student taking an exam with no outside help —they can only answer based on the data they were trained on and as such, this means responses can be outdated or inaccurate since the model cannot verify or update its knowledge.
RAG on the other hand is similar to that of an open book exam in that the AI can look up relevant information, like a student using reference materials during an exam.
As such, it fixes the issue by allowing AI to pull in real-time, validated data before generating an answer. It does this by retrieving relevant information from external sources (e.g., databases, documents, or your IBM i enterprise systems) and then processes this new data before generating a response, ensuring accuracy.
And, unlike standard AI models, no additional training is required, saving on compute power and GPU costs.
How does RAG AI affect the IBM i community?
But why does all this matter for the IBM i community? Well, RAG will ensure that your AI responses are accurate, verifiable, and cost-effective by eliminating the need for constant model retraining.
And, as we’re running on IBM Power server, RAG will efficiently process real-time enterprise data without requiring expensive external GPU resources.
So, you can see how this could be a real game changer for businesses running on IBM Power.
But there was more and David provided a real-time demonstration of how RAG can support your enterprise with Hybrid AI and what this would mean for your business, providing examples and elements of agentic AI too. It was an enlightening presentation delivered with pace and virve.
And, should you wish to learn more and see the presentation yourself, you still have the opportunity to do so…
As Jay and David are presenting the very same presentation at the i-UG Conference South on Thursday 13th March at IBM, York Road, London. I’ll be there, and so should you.
Upcoming announcements for the IBM i
Moving on, rumours are rife on what IBM’s major announcement on IBM i will be? Steve Will provided a snippet of information during his ‘Strategice Preview’ IBM i guided tour.
In this, Steve said, “All of you should be expecting a major release this year, because we do three years between major releases,” he continued. “I can’t announce everything or anything specifically right now, but you all should be expecting that.”
So, it’s clear that IBM will be releasing a major version… a version 8 perhaps. Or, with all the new AI hot sauce, perhaps (god forbid) we’ll see a rebrand?
Indeed, Will gave a nod in this direction… he said, “We want to embed AI into Db2 and make it easy for Db2 data to be shared with AI environments… We want to make it easier for people to deploy and manage AI workloads that are connected to IBM i.”
Lots of exciting stuff upcoming from IBM and you’ll be sure to learn more about this major release on the IBM i Update.
IBM will end standard service for Power 9 Systems
And finally, a reminder for all who have a Power 9. IBM will end standard service for Power9 systems on January 31, 2026. This means that IBM will no longer provide updates, patches, or maintenance for these systems after this date.
Now, while IBM have history in providing extended support, should you have such a system, now would be a good time to consider your options for a Power 10 and updating to OS 7.5.
And that’s it for this month’s update. For more like this, please see our blog on the Proximity website, but for now, I’m Andrew Nicholson, we’re Proximity, the IBM i partners that are in your corner.
Should I not see you at i-UG London on the 13th March, I’ll see you next time on the update!
Catch our previous January IBM i Update here. Alternatively, if you’d like to receive the latest IBM i Update hot off the press to your inbox, subscribe to the newsletter version on LinkedIn.
