This might seem like a naysaying post, but it isn’t. In the 30 or more years working with technology, I’ve ridden atop a bubble or two myself. It’s something that the industry seems to need to do. Take a running jump at a market or problem or inflate itself until it bursts. I leave that up to you to determine. I want to talk about the current bubble, AI.
Throughout my career I’ve worked with people of varied knowledge in all the different areas of technology such as networking, software development, management and project management. For outsiders, it may come as a surprise (or not) that there are very few technical people in the average IT department. You generally divide IT departments into different areas. Administration, Development and Support but it depends on the organisation. Each sub department has in turn difference sections or levels to reflect the function it performs or supports. Support teams for instance typically have three levels consisting of the help desk staff (level one) rising up to the people who possess a great deal of knowledge and the passwords to the systems (level three). Development teams are the same. There are individuals at the top who have a great deal of knowledge and corresponding power to make changes system-wide. Sometimes there is cross-over between teams when you get people with skills everyone needs. But these people represent a very small percent of any IT team. Probably no more than 5 people in every 100.
These days with the maturity of commercial software and the sheer numbers of businesses running on the same technology (the genericization of business technology, another upcoming post) businesses hire people with specific skills usually in the technology they operate on. They overlook people with a broad understanding of technology because they don’t fit a preconceived mould. Many business system architectures are designed with layers to prevent critical parts of the software from being ‘fiddled with’ and broken. Some may describe this as ‘welding the bonnet shut’. Maybe that’s cynical but it doesn’t take away from the benefit to companies like Microsoft keeping their systems somewhat cryptic and customers reliant on their accredited consultants. Whether or not this leads to more reliable systems isn’t clear. I’ve worked with both opensource technologies and proprietary ones, I’ve experienced much greater levels of robustness with opensource technologies. More scrutiny, more people with passion. The dumbing down of the corporate IT department is now complete. Decades of IT departments being overseen by people from accounting backgrounds have driven anyone with any passion for technology out. The waves of irrelevant questions and performance metrics, the soulless technology platforms from salespeople with deep pockets for steak dinners. Today most businesses have the same software as their competitors and less opportunity to develop and advantage from building intellectual capital. The adults left the building years ago.
So along comes another technology revolution. Another pig sporting a different shade of lipstick that promises renewed employment contracts through hyperbole. All aboard everyone who doesn’t know or care what’s actually happening under the PowerPoints and service agreements. Yes, AI isn’t the real problem just another hyped up tool. It’s the salespeople and the dummies who bought the dog-food that need to be pushed out. And to honest, it’s the name that irks me most of all. It’s the use of the word intelligence. Since becoming a farmer, I have become acutely aware of the difference types of intelligence animals possess. Horses have incredible recall memory but are shit problem solvers while dogs can be impressively sneaky bastards. It’s because dogs don’t need to do something before understanding it. They can imagine it. To describe why AI isn’t really intelligent, I need to cover the more abstract forms of intelligence, reasoning and problem solving.
Intelligence
Although recall memory is classified as intelligence, almost any creature no matter how primitive can remember something for some period. And computers have always been brilliant in this context. Taking some information and storing it until its needed. But abstract forms of thought require diametrically greater brainpower. Proponents of AI will try to tell you AI has the ability to perform reasoning and problem solving. When the reality is, at best, AI mimics abstract thought. AI can’t actually perform abstract thought because the underlying hardware in unable to. Currently mankind is unable to build processors that can perform anything other than liner computations. AI is turbo charged recall memory. This is why I describe it as Advanced Automation or Machine Learning. The promised computers of tomorrow did not eventuate. In the end the research was too expensive and not producing anything that could be sold. Instead, the existing computers chips architecture sufficed as long as they increased in speed enough each iteration to warrant upgrades. Quantum Computing
It appears the great leap forward was taken out the back and shot in the head. I touched on quantum computing as part of another AI post here: Will AI lead to a revolution?
noting the recent Google announcement of a new 105-qubit 'Willow' quantum chip.
In that post I focussed on the impacts to employment. Here I want to focus on how AI is oversold and underestimated. Overselling is the hot air in the bubble. The underestimated risk is the pin. Unlike other technology trends AI is intrusive. Nearly all computers and mobile handsets currently being manufactured have dedicated AI hardware included. It’s not going to be like other trends where you could just opt out. You can come up with your own theories why the tech giants are including something at cost to them, when there is no demand from the consumer. I think its fairly easy to assume there is some significant benefit to them.
It’s the lack of oversight from governments, private developers and even systems administrators that keep AI from any scrutiny. AI is a service offered as a turnkey API that very few people have any inner knowledge on. The people making the decisions to implement AI would possess around the level of understanding as someone shopping around for a new car. AI is (among other things) a massive vacuum cleaner of information. And yes, current household smart devices and digital assistants gather information about you. But AI is less benign, and it has much more access. Even someone like me who builds his own smart devices will be powerless to stop current devices like an iPhone16 from sharing anything it listens to or sees. Then there is the data residing on the device or any connected services. You better damn well trust it.
Artificial
Today we have already witnessed alleged human deaths as a result of AI interactions. Most readers would be aware of the instances of fatalities in self-driving cars. If you are not... What is perhaps less known is a case where a US teen suicided after interacting with an AI bot. The chat log submitted as evidence was, at the very least, alarming to me.
and recently this: AI tells student from Michigan to die
People are interacting with entities under a guise it cares about their welfare. Are we sleepwalking into thinking these systems possesses moral and intellectual capabilities? As I have outlined AI cannot technically develop its own reasoning. The systems would need to be ‘taught to’. And I would be quick to point out some of the key players in AI development have displayed very questionable behaviour and a complete disregard for public safety. As they say, garbage in, garbage out. And we have allowed these systems into our lives with unbridled access to us. Good luck.