AI and quantum computing – the Skynet scenario

In the original Terminator film, Skynet, an AI controlled defensive system, became self-aware at 2:14am EDT, on 29 August 1997. And that led to the frightening scenario in the film and all its sequels. I’m not really suggesting that we are all doomed. You know as well as I do that headlines are only there to grab your attention. What I am suggesting is that these emerging technologies (and like all emerging technologies, they’ve been around for a while) will make huge changes to our lives.

Let’s look back at personal computing. Many people can remember the clunky IBM PCs available in the early 1980s. You typed into them and remembered to save your work (usually). Nowadays, using Word or Google Docs, you type in, and it tells you if it thinks you have misspelled a word. It also suggests that your grammar may be wrong – you have used the wrong form of the verb to go with the subject of your sentence, The software has, in many ways, deskilled the job of writing. You no longer have rooms full of typists at companies. Most people no longer need a secretary to write their letters. It also means that teachers marking essays, shouldn’t be spending their time correcting grammar errors in their students’ work because the software should have done it already before the work was submitted. Teachers will be marking the content of the essay, the information it contains, and the application of the facts and the ideas.

In a similar way, when things like ChatGPT, Bard from Google, and Copilot from Microsoft, become commonly available, they will have a huge impact on the essays written by students. It will be possible to tell the AI what you want written and how many words, and the AI will write the essay. To begin with, these will need to be checked to ensure that the facts are all true (they aren’t always!), but that will only be half the story. The teacher marking the essay, will, probably, have already received 30 very similar essays from the other students in the class. What will be needed is for the students to show how those fact can be applied, and their own ideas. The simple regurgitation of facts won’t be enough. People will still write essays, but the important content will be different. I’m suggesting that things will have to change. And, slowly they will.

ChatGPT and the others are examples of Large Language Models (LLMs). These are AIs that can generate natural language texts from large amounts of data. Their responses make it seem like you are conversing with an intelligent piece of software. However, in many ways, they are simply choosing the most likely word to come next in a conversation or essay or poem, etc.

So, AIs could produce contracts for the merger of two companies. Both sides will use their own Ai to create very similar contracts. And those companies will use AIs to ensure that the contracts protect their investments. And, to begin with, qualified and experienced staff will need to double-check those contracts. What won’t be needed are the staff that usually write those contracts in the first place.

AI could write programs in whatever language you choose. And another AI could test whether that new program works with the existing applications on, for example a mainframe, without causing any problems. But, to begin with, expert programmers and experienced systems programmers will need to cast an eye over the code for any issues or anything that is missing. This is how the AI will learn.

The biggest issue, for some people, with AI is when the software creates artwork, for example a sonnet, a painting, or a work of fiction. Is the book, poem, or picture really art? Does that mean our definitions are going to have to change?

At the more practical level, AI could be used to perform a ransomware attack on your mainframe. It could also carry out the best pen test you’ve ever had on your system, and guide you through the steps you need to take to be as secure as possible. Again, this could lead to a dystopian future (like Terminator) because the figure often quoted is that 20 percent of cyberattacks come from disgruntled employees. Get rid of all employees and cut your risk of attack by a fifth!

All of that can take place on our existing computing platforms. What if you add quantum computing into the mix?

Quantum computers use chips that are housed in cooling units just above absolute zero (0^oK), where electrons can move through the superconductors with no resistance. They use qubits, which can simultaneously be a 1 or a 0. IBM says that a qubit places the quantum information it holds into a state of superposition, which represents a combination of all possible configurations of the qubit. Groups of qubits in superposition can create complex, multidimensional computational spaces. Complex problems can be represented in new ways in these spaces. In contrast, all current computing platforms use 0s and 1s to perform addition quite quickly. The bottom line is that quantum computers are hugely faster than conventional computers.

The one negative use of quantum computing that everyone is talking about is their ability to quickly decrypt encrypted data. What people conventionally think would take decades to decrypt, could be decrypted on a quantum machine in hours. That would make online banking a very risky business for everyone. It means that encrypted messages sent over the Internet could be read. It makes the world a very unsafe place. IBM recently took steps to make their mainframes quantum-safe – as far as possible.

On the more positive side, putting AI and quantum computing together would be brilliant for our health. It would be able to identify that we are taking more pain killer medication than usual and identify, from other sources, what our condition might be, and recommend which medical specialist to see for further tests.

It could control traffic flows on busy roads and in towns and cities. It could check planning applications meet local bylaws. So many mundane tasks, could be performed incredibly speedily by AI on quantum computers, and simply checked by qualified humans, that there would never be any need to wait for months for anything to happen again.

Moving forward further, it would become possible for AI running on quantum computers to control robotic equipment to perform very delicate operations on humans – like heart surgery. And, because it never gets tired, these could run 24 hours a day. No-one needs to wait for an operation.

As always, the future is uncertain, but there are many huge benefits of combining AI and quantum computing. But there will be misuse of the technology that needs to be confronted and controlled. I don’t believe Skynet will be become self-aware. I do believe that there will be huge changes in the way our children or grandchildren live their lives from the way we do currently.

Size isn’t important – z16 rack-mounted versions

Following on from last April’s announcement of its z16 mainframe, IBM has announced its new rack-mounted version, which can be used with standard 19-inch racks and power distribution units. And the advantage of that for customers is that they don’t need a massive, dedicated computer room to house their new mainframe, they can just slide it into the racking they already have and start getting all the advantages of a modern mainframe.

Not only are there z16 configurations, there are also IBM LinuxONE 4 configurations. That means, organizations that are currently using lots of Linux racks to power their corporate computing (with lots of people to manage those configurations) can save money by replacing them with a single new unit. It takes far fewer people to look after as many Linux systems running on a mainframe than x86 devices. It also takes less power, so it is a greener option for organizations looking at their current carbon footprint. Specifically, IBM is saying that consolidating Linux workloads on a single Rockhopper 4 would let customers replace at least 36 x86 servers. It would also reduce energy consumption by 75 percent and space usage by 67 percent. The IBM LinuxONE Emperor 4, was launched in September last year,

These new rack-mounted mainframes use the Telum processors that everyone was talking about last year with their almost real-time AI inferencing at scale and quantum-safe cryptography. In addition, they provide high availability levels designed to help companies maintain customer access to bank accounts, medical records, and personal data. With the on-going growth in cyberattacks, including ransomware attacks on all companies, including mainframe users, these new mainframes include important security capabilities such as confidential computing, centralized key management, and quantum-safe cryptography.

The z14 was the first mainframe I saw as a rack-mounted system, and the z15 was also offered in that format. So, IBM must know that the rack mounted mainframe idea works with customers.

You may ask how powerful the new processors are? Well, the technical specifications aren’t as high as last year’s z16 models – and that’s been IBM’s model for selling mainframes over quite a few years, to bring out a high-spec machine one year followed by a less powerful variant the following year. I always think of it like a hardback copy of a book coming out one year followed by the paperback edition the following year.

According to Marcel Mitran, IBM's CTO of cloud platform and IBM zSystems, IBM is seeing the decentralization of IT infrastructure so users can bring compute power closer to where their data is, especially for hybrid cloud environments”. And that’s what using the new single frame and rack-mounted systems can achieve. There is now an easy way to bring high-level computing power to what was the edge of a company’s computing infrastructure.

Of course, if you move processing away from the centre, you’re going to increase the risk of security breaches, which is why potential customers would do well to take a look at these new models because, as well as strong security, they provide high availability, transaction processing at scale, and greater resilience. The new models provide 99.9999999% availability. Someone else has calculated that to be equivalent to three seconds of downtime over the course of a year.

IBM is saying that the newly-launched single frame and rack mount configurations increase client infrastructure choices. Like any big business, IBM needs to sell what customers want to buy. And that’s clearly the thinking behind this move. It certainly gives customers one more great choice about how to modernize their computing environment.

The new IBM z16 and the LinuxONE 4 servers both become available on 17 May

 

Mainframes – expect the best

Too often, when I have conversations about mainframes with non-mainframe IT people, the picture in my head of the mainframe is a pimped-out rack with flashing lights – you know, the kind of box that IBM takes round to mainframe conferences etc. It’s exciting and modern to look at. Whereas the picture of a mainframe that these other IT people have in their heads is some giant box with hundreds of peripherals filling a computer room and everyone staring at green screens. Plus, like movies from the 1970s, they picture lots of tape drives with the tapes whirring backwards and forwards like there’s some kind of parity check problem. Although we’re using the same words, we have completely different images in our heads.

We also have completely different emotions about the topic in hand. I rather like mainframes and see them as a large part of the answer to the question about the best way to stay in business with a successful modern IT strategy. They see them as being about as interesting as a box containing last-week’s pizza, and as relevant as whether sound will every catch on in movies!

Let me very quickly tell you about your brain and how it works. Most people think that your brain simply responds to stimuli, which it does, but it also acts like a prediction machine. It builds up expectations about what will happen next. So, if you predict that the next Zoom meeting you have to attend is going to be a complete yawn-fest and Bob from accounts is going to drone on about the predicted year-end figures, you will start to act as though those events have already happened. You will, in fact, be creating a self-fulfilling prophecy. And, if you want to beat Bob from accounts at a game of tennis at lunch time, you need walk on to the court like a winner because that’s what you expect to happen, and it probably will happen.

So, the point of telling you that is because when I’m talking to non-mainframers about the mainframe, they expect me to be discussing a box built in the late 1960s running an operating system written about the same time. Their assumptions are so different to mine, their expectations are totally non-congruent with mine, that we could be talking a totally different language because of the lack of communication that is going on.

For example, I was talking to a small group of people the other day and saying that I could run mainframe software on a laptop. You would have thought that I’d just said that I had discovered Atlantis and there was a tunnel to it from the middle of Swindon. Everyone thought that it was a huge joke. And yet, that’s what IBM Z Development and Test Environment (ZD&T) does. It runs z/OS on a PC running Linux (Ubuntu). It doesn’t need mainframe hardware! The idea is that so long as you have an Intel or compatible chip on your PC or workstation, you can run z/OS, middleware, and other z/OS software, and emulate z/Architecture with virtual I/O and devices.

There are three variants. ZD&T Personal Edition lets a single user to run IBM Z on their PC. ZD&T Enterprise Edition comes with a web-based interface. And ZD&T Parallel Sysplex can be used to enable a Sysplex environment that is running within z/VM.

It seems like a useful idea to try out your mainframe software on your laptop without using any mainframe cycles. As well as testing, you could also develop applications, and you could demonstrate them to colleagues or potential customers. Or, I guess, you could train new application programmers using an environment that they can’t break – or at least the consequences of breaking are comparatively minimal.

Alternatively, you could run mainframes in the cloud with IBM’s Wazi as a Service. Again, it’s a development and testing option that doesn’t actually involve any mainframes – because, as I just mentioned, it’s cloud-based.

All this came to mind because I see that Pop-Up Mainframe, a company I know nothing about, has just announced PopUp Mainframe 2.0, which provides virtual environments running ZD&T Version 14.0.0 and IBM’s latest z/OS operating system, IBM z/OS V2.5. They are offering it as a quick, low-cost way for organizations to test business applications on z/OS 2.5, with zero risk to the physical mainframe.

PopUp environments can be spawned and retired in minutes, allowing IT teams to create an environment, perform their testing, then destroy the environment with no long-term maintenance requirements. The company is claiming that it’s a great opportunity to test the new features available in z/OS 2.5.

Like I say, I don’t have any connection to the company, but the idea of running mainframe software off the mainframe (whether that being Intel-chip devices or the cloud) seems like a great way to get those DevOps program developments under way and overcome the development backlog that has built up at some companies.

There are so many ways that the mainframe (and, yes, I know I’ve just been talking about working without a mainframe being present) can be flexibly used by organizations. It just seems so sad and limiting that non-mainframe specialists in the IT team don’t realize what a powerful and significant contribution mainframes can make to the success of a project and the company they work for.

Perhaps, when we speak to these non-mainframers, we should take a leaf our of Socrates book. He would challenge the assumptions of the people he was talking to (it’s called Socratic questioning), and he would keep doing it. If we did that, they might recognize that what they think about the mainframe – their expectations – are old-fashioned, or just plain wrong. And we can help them understand what benefits the mainframe can offer to their organization.