Tuonelan sorsa

mutta

114 people died in the Hyatt Regency collapse, and in the US it's the third largest structural collapse fatality count, behind 9/11 and the Pemberton Mill collapse in 1860.

I've learned about this tragedy in my physics classes, to demonstrate tensile strength, and as a reminder about the importance of calculations being done right. I've also learned about it in my legal classes as an example of construction defect lawsuits. I've seen it referenced in disaster response classes.

Between AI and the current Presidential administration, we're barrelling right back towards this nightmare.

There are multiple errors that resulted in this collapse, but these stand out to me:

1. Kansas City was facing high unemployment and needed to attract jobs and business into the city. So the planning and inspection departments may have looked too closely at the designs.

2. An engineering firm too lazy to double check their designs or design changes by the manufacturer before approving them. The error that resulted in the collapse was one that the owner of the engineering firm said that a "first year engineering student" would spot.

3. The steel manufacturer treating preliminary plans as final plans, not verifying the math on their end.

The bridges' original design could only hold 60% of the minimum load required by city code. The design changes recommended by the manufacturer halved that. Less than a year and 3 weeks from opening to the public, the whole thing collapse.

Articles about the collapse say that everyone "trusted" the other party to have done the calculations correctly.

A significant portion of the population trusts what the computer or AI tells them, without checking. Imprecisely calibrated AI hallucinate information. The US economy is going into a downturn and federal regulatory agencies are being gutted.

We are going to see the Hyatt Regency Collapse repeat over and over for decades, not just in buildings, but in medicine, manufacturing, the environment, etc.

Some of this we're just going to have to weather, but the message for AI users comes straight from IBM (once the world's leading computer manufacturer) back in 1979:

"A Computer Cannot Be Held Accountable. Therefore A Computer Should Never Make A Management Decision."

The owner of the engineering firm that designed the Hyatt Regency spent the rest of his life lecturing on the disaster, to serve as a warning to his fellow engineers about the real-life consequences of sloppy design.

I don't think Sam Altman or Mark Zuckerberg or Elon Musk will have the courage or the honor to do that when OpenAI / Meta / xAI are responsible for getting people killed.

I recently was seen by a nurse who provided medication. Gonna be as vague as possible here, you'll forgive me.

Based on my age she wasn't sure if I would require a repeat prescription.

I sat and watched her, a nurse with presumably decades of experience, read Google AI results and say she couldn't find the right answer.

She later texted me and told me she would need to see me again some months down the line.

I work in the field that produces and manages that medication. I read the government guidelines and guidelines laid out by my country's healthcare system.

Both sources with the most reliable information that my organisation offers in regards to this treatment.

She "looked" and texted me the wrong info. Five seconds of reading through the official channels would have shown exactly what I needed in plain and clear writing.

I believe she asked Chat GPT.

I have friends whose collegues work with people from privelleged categories and when trying to discuss how to work with their information all of the staff, save him, are advocating for AI to be used. He's the only one pushing back because of safety concerns about their patient's data. None of the other professionals had even considered that AI could be fallible or untrustworthy.

One of his collegues couldn't find an email for a doctor's surgery. She asked Chat GPT. It made one up. It wasn't a real email. A two minute search found the right one, alongside a phone number to double check with the surgery's reception to see if another number would be suitable.

These are two small mistakes, with healthcare. With people's lives.

People are going to die.

If I wasn't in my field I could have suffered through unecessary treatment. If that email Chat GPT imagined had been a real, unrelated email address, then that patient's information could have leaked had my friend not stopped their coworker.

People are going to suffer.

People are going to die.

Inspiroiduin