Should NI drop investment in AI for now?
In the Thursday session led by Elijah Kerry at GDevCon#5 regarding industrial protocols, the session expanded to a wide-ranging discussion of NI priorities.
While discussing the ten Investment Themes, an audience member suggested dropping item #6, the use of AI in generating code, to make more resources available in the other areas which sorely needed work.
This generated an strong audience reaction, and when Elijah informally polled the audience, the majority of the participants supported dropping or delaying the AI investment!
Elijah remarked that this was exactly the opposite reaction from a recent meeting with NI users in the USA, who strongly supported NI moving into AI.
One audience member at GDevCon responded that his company had already done significant work investigating use AI, and had found negligible improvement. And many in the audience showed skepticism regarding the use of AI to make LabVIEW code generation more efficient.
As a member of “the old farts” community, having cut my teeth on Fortran and job control cards, I was struck, but not surprised, by this great Atlantic divide.
One obvious reason is that Europeans are more skeptical about “overly-hyped” products from Silicon Valley. We recall with horror the .com bust, and fear a similar AI bubble is in the making. There is little critical reporting about AI, and most CEOs seem forced to announce that they are on the bandwagon, perhaps fearing their market value will be punished if they are left behind.
A remarkable exception to this is Apple, which has been cautious in not over-hyping its approach, instead asking, how does it make your life better? And that in the face of competitors like Google, who is widely recognized having much more advanced AI tools but with well-known side effects and issues.
When I listened to Kevin Schultz’s presentation at NI-Connect 2024, I was impressed by his conservative approach to AI, how he noted that NI would be careful in testing the results of AI generated output. Again, an approach that exploits AI and finds the use cases that make a real contribution. I would suggest that underneath the banner of AI hype, his presentation was a very reasoned, “European” approach to the subject. And thus, was an excellent example of a classic Dr. T. ism “The genius of the And”, in this case being the biggest optimist and pessimist at the same time.
Which brings me back to Apple and Silicon Valley. In his last years before his death, Steve Jobs told Walter Isaacson that his greatest fear was that Apple would suffer the same fate as Hewlett-Packard, which seemed to have lost its way. He feared that Apple would be destroyed by the short-term focus that came in the early 1980s when fast money and fat stock options, drove the industry to a “get rich quick” focus. “Throw money at the wall and see if it sticks.” Steve Jobs remembered, as do I, that before the 1980s, HP was considered the “dean of Silicon Valley”. The most highly-respected, innovative company in the valley, whose products known for the trustworthiness and reliability. Hallmarks inherent in the test and measurement industry. HP lived by the mantra: “Under promise, and over deliver”.
My take-away from the meetings at GDevCon#5, is that there were open and frank discussions, where NI humbly recognized mistakes in recent years, along with their commitment to get things back on track. And most of all, I was amazed by the loyalty, enthusiasm, and love for LabVIEW of the conference participants, despite having been along for a bumpy ride!
Written by Carsten Thomsen
PS:
You can vote on NI priorities here.
3 replies on “AI and NI: The great Atlantic Divide”
AI definitely is overhyped and there is a lot wrong with it. For one it’s name is simply miss leading. We had a similar attempt at trying to hype computers as AI 40 years ago already when people were trying to sell to the public programs like Eliza as AI and the next big thing in human development to the dismal of Eliza’s creator. Nowadays LLM are much better and can actually deduce the context of written text instead of just acting on some trigger words But what it generates from that is still not much more than what a statistical parrot could do.
Yes it can search its internal language model much faster for existing patterns than a human could do in his often rather imperfect memory but if the answer is not there already, it will with the highest authority, simply put together whatever seems statistically most relevant and put it into eloquently formulated but usually quite empty sounding sentences.
It’s a perfect marketing speech machine but is that what we really need? Selling more junk to people who didn’t know they needed that junk?
And that’s not even talking about the energy consumption of those huge mega data clusters powering these Algorithmic Intelligence machines!
I remember thinking during a company meeting that Dr. T was too obsessed with this “Google” thing … making sure that our page was the first one that would come up. I’m sure grateful NI made those website investments. In that era, ni.com was best in class. I don’t think NI can afford to “wait and see” on AI.
I personally am primarily excited about the fact that LabVIEW’s future is once again a major focus at NI — regardless of the specific priorities.
“NI humbly recognized mistakes in recent years”
Has this included a re-think of the huge cost increases for using LabVIEW? The annual license fee is an immense disincentive to using LabVIEW.