Just want to add that there is a difference between lending and investing.
There’s an old saying that actually has some truth to it: “equity is a pillow but debt is a stick”. That’s because lenders can—and do—force companies to shut down because they cannot meet the payment or balance sheet requirements of a loan but investors, as long as they have the appetite, can keep injecting capital into a money-losing company they believe in (classic example: Amazon).
In theory, at least. The current data center construction craze is a lot like what goes on in professional sports. Teams pit different cities against each other, demand tax breaks, receive public funding for lavish stadiums…the list goes on and on and on!
I wonder how similar this AI development is to what we observed with ride shares many years ago.
Uber and Lyft initially appeared dirt cheap to customers so everybody started using them. But it turned out, these companies were just burning through heaps of loaned $ that de facto were subsidizing a service sold below cost. At some point the VCs et al. got tired of paying for everybody’s cheap cabs so they forced changes. Uber and Lyft got expensive and rider numbers dipped. They never became as cheap again, but now I guess they are actually making a profit.
Meanwhile, competition to their nice comfortable little duopoly did arrive in the form of driverless cabs (Waymo et al.) and because there’s no driver (especially in areas where drivers make a min wage and need to be provided at least some benefits) they held the promise of being able to undercut the now expensive Uber/Lyft duopoly. But gosh big surprise, that didn’t happen. You don’t get a driver but you still pay top $ for your ride. So by now I guess stock markets are reaping some benefit from all this, but customers still seem stuck paying essentially what they paid back in the day, at least now with the benefit of no longer sitting in dirty smelly Crown Vics with card readers that supposedly always just went out of service. 
The basic idea is similar. Start out with lowball prices that lose money but squeeze out the original competitors, and when they’re gone, jack up the prices and coin money. The actual formula for getting rich quick is more like “Pump it up, Sell it off, and run away very fast”. The “smart” investor wants to get in early, raise the stock price, and cash out before they have to actually run a business, which is hard work.
Don’t forget there are two sides to every trade: a seller and a buyer. If something has an inflated current value or unpredictable future value, then it’s on the buyer if things don’t work out.
An example of the problems with AI is in the attached screenshot. I searched in Google for fibromyalgia causes and Google gave me this not helpful summary. If your pain signals are being processed by the Mayo clinic then it is a problem. I tried it again a few hours later and it gave me sensible results.
I am a heart attack and cancer survivor and a victim of a false diagnosis for an ICD and a missed blood clot in the heart for seven years that grew to the size and weight of a golf ball and now that is gone has damaged my heart to the point where I am now classified was suffering heart failure although the situation could improve through time.
Given all that, without AI, I would be a goner.
My medical situation is complex. Just asking AI, “duh what should I do to survive with heart failure?” is nowhere near enough. Neither would that work to ask a random cardiologist. Context is everything. And context is probably the biggest task when using AI.
One great place to start is to upload medical reports and blood test results - the more you have over time the better.
But be prepared to question the results. For example, I had to switch from long-term use of aspirin to another blood thinner, which created a certain result in the blood test (don’t remember what off hand) that could have been concerning if it was not explained.
AI was able to rip through my medical file in another language and learn that back in 2019 one cardiologist falsified my LVEF and also missed the blood clot because she used software to evaluate a CT scan.
Meanwhile, another hospital got the same better result saw the falsified result from the first hospital and then falsified their results to match that of the first hospital.
When I told AI to dig further, it found a strong working relationship between the offending cardiologists of both hospital.
Almost had another heart attack when I learned this.
Let’s keep this focused on issues surrounding the AI bubble, not the utility or lack thereof of different AI systems.
