Absolute Insanity

This all started a few days ago and I had to mull a few things over. You see AI does not exist, no matter how strong the hype and the presentations are. Now we see also the term ‘spatial AI’, another joust towards hype and revenue grabbing (the easy way). There are a few issues with all this. You see machine learning and deeper machine learning are great, they are awesome. In addition the growth of Large language models (LLM) are adding to the mixture but here is the snafu (situation normal, all fucked up). It is all still in the hands of programmers and verifiers. The issue of human error comes into play.

So when we realise this The BBC Article (at https://www.bbc.com/news/articles/c977elr6veno) called ‘Airline to ‘better manage’ flights with AI use’ should get some investors worried. The start is seen with “The use of artificial intelligence (AI) at easyJet’s new control centre has allowed its operations teams to better manage flights, the airline said.” It reminds me of an old setting in the 90’s when someone produced a program called Goldmine. Don’t get me wrong it was a good program but it relied on standardisation. That means that exceptions aren’t dealt with. The programmers never anticipated the exception thy were given, so alternative fields were used and in AI the use of alternative solutions tend to be devastating on data models. So when we see “More than 250 staff work in the control centre, managing easyJet’s daily programme of about 2,000 flights.” We might see the initial problem. Last minute changes (pilot gets food poisoning) or perhaps the flight attendant got stuck in location X. It does not matter what the issues are, things will go pear shaped. And that is before they are confronted with the ‘oversight’ of the programmer. 

Now there is the recognition that a system like this can reduce stress on these 250 staff members, but it will need human verification and that is not what an AI system needs (if it existed). In the end I reckon that investors will see in 6-12 months that operating costs have exploded. I reckon that Johan Lundgren talks a good talk, and there are benefits to Deeper Machine Learning and it will help any corporation but the missing part in this are the programmers. You see these solutions aren’t AI, they required a programmer and that programmer makes mistakes. It might be simple, it might be complex and when that is found it tends to be in the most inconvenient way possible. 

Interesting that the BBC didn’t see this part. It would have been the first step I would take. Which firm was involved in this system? How many programmers? What previous assignments did they have? I reckon that the investors might have some questions on all this and I hope for Johan Lundgren that he has answers.