Sticking To The Backroads On This Journey
August 7, 2023 Timothy Prickett Morgan
Things are getting progressively weird in the computer business, and with lots of travel over the past two weeks up and down the East coast for vacation and for visiting family, I have had a lot of time to think about things. It’s so weird that I really don’t want to play the radio. I just want to keep us safe on the highways and have my thoughts. Parse a few things, suss them out, see how they might fit together and in what particular configuration.
Because time is of the essence, I have been largely on the interstates, but the best thinking is on the backroads and has always been. I plan to get some such time, just to slow my brain down so I can feel a different rhythm.
What I can tell you is this. We have absorbed a lot of change in the four decades that I have been involved in the industry, first as an amateur enthusiast and then gradually becoming a professional who knows a thing or two about how systems work and how they affect culture, society, and business. I don’t think that the pace of change has been exponential as some people are fond of saying. But the rate of change has been constant. The PC revolution of the middle 1980s and early 1990s, the commercialization of the basic Internet and Web in the middle 1990s and early 2000s, the rise of the Internet giants and data analytics on a commensurate scale in the 2010s, and now the advent of large language models and generative AI that does more than play games in the 2020s.
With each revolution, changes came to the backend systems like the AS/400 and its progeny, and despite our frustrations sometimes because we have wanted this platform to always be on the cutting edge, Big Blue has done a pretty good job at making sure the relevant bits for business computing have been integrated into the platform.
It is not at all clear how generative AI will be integrated into the IBM i platform, and I expressed some of my misgivings about it back in January. There is no question in my mind that every aspect of business will be touched by generative AI, although I also expect some resistance. The writers I know, for instance, are adamantly against using it, much as the Screen Actors Guild is taking a hard stance. But I also know that some editors are using ChatGPT to clean up the bad copy that some writers have turned in for years and for which, honestly, they should have always been ashamed. No such use case exists at IT Jungle, of course. Alex and I are too proud to have some machine do our work for us. I do use transcription services for my interviews, and anything that improves the speed and quality of the speech-to-text conversion is fine by me. I still believe that the interview is vital, and that I need to ask the questions, new questions in pursuit of new data and new connections, and that I can do this better than any chatbot. And then the weaving together of a story after that interview is something that also requires the human touch for it to have life and not sound tinny.
I like a world where all people matter, and I am determined to be a person and to matter. And I think that, given the situation that all white collar and pink collar workers are facing – the blue collar workers faced the robots back in the 1980s, and my Dad was one of them that had to stare down a Unimate robot in the foundry where he busted his ass all day. Ultimately, the robot won, but it was not better than him until years after he retired.
The AS/400 and its progeny live largely on the backroads and on the outskirts of the cities, and that is going to serve us all well on this next – and perhaps most dramatic – of transitions in the IT sector. The companies that design, make, distribute, or sell stuff are still going to need to do that. Those that keep track of our money or insurance will have to do the same. We can – and should – take a conservative approach to generative AI. None of us are trying to compete with Google and Microsoft and Meta Platforms and Amazon Web Services directly at generative AI. That’s a billionaire’s game.
Take it slow. This is not the Dot-Com Boom. You don’t need to figure out your presence on the Web in the next six months. You need to figure out how to train and keep the people you have and how to please and keep the customers you have. None of these things is easy anymore. Perhaps they never were. But they sure do seem to matter more with so many livelihoods at stake.
This wave of technology demands more caution than the prior ones because the unintended consequences loom so large. And that is saying a lot given the amount of damage that social networks and what passes as media has already done. Twitter, now X, is becoming a platform in its own right, but it does not have the pristine image it used to have. LinkedIn is still, as far as I can tell, a good platform so long as people keep it to business. Remember: This is business, nothing personal.
Other than that. We will just have to see what develops with generative AI and react calmly but purposefully to rein it in a little. That’s all the wisdom from the road that I have this day.
Generative AI is based on parsing and modelling a lot of human work. Incorporating even, of course, the incidental bias and the prejudice of such work. See the issue risen by holders of copyrights of books… the AI is trained using millions of works and articles… if you sell then the AI model, do you need to recognize a fee for the underlying owners?
If then AI will start training on AI generated articles, I see something circular happening with the current models. With the possibility to “poison them” willingly. Still one needs human to break the circle.
Regarding in-house company business use, I think we will just access such services and tech via webservices and stop.
However, the “non-generative” AI field and machine learning and statistics is not so fashionable but still vast and frequently even more relevant for internal decision processes and optimization.
Even “basic” and classical things like building decision-trees from relational data can be powerful. And I see the case of local inference engines to speed up things. For businesses is fundamental that the model must be transparent with the possibility to introspect such model: i.e. explain what have driven the model decision. Compare that to tracing the reasoning flow of a neural net with billion parameters… you end up treating that as a “black box” and problems will normally start then if you use such tech for critical things… it’s easy to create monsters.