Talk about bringing a knife to an AI fight, the clear winner was a knife. Or really, any large knife.

One New York school district learned this the hard way. It spent close to $4 million to buy an AI-powered weapons scanner from Evolv Technology that the company bills as “proven artificial intelligence” able to create a “weapons-free zone.”

Then, on Halloween last year, a student walked through the scanner carrying a nine-inch knife and used it to stab a fellow student multiple times, the BBC reported.

[…]

A BBC investigation found that while Evolv Technologies claims their systems can detect guns, knives, and explosive devices, during 24 walk-throughs a scanner missed 42% of large knives.

Evolv claimed it found a lot of knives and it saved a lot of time. These are weasel words. They cost $4m and failed. Was it worth the cost?

The obvious answer seems no, the school should have paid teachers that $4m instead.

Teachers broke up the altercation, local news station WKTV reported, and the victim was taken to the hospital.

Teachers ironically aren’t paid more for intelligence work, but they clearly have to clean up the mess when high cost low value AI is installed.

A local southern CA news story says a Tesla driver killed himself and two others when he crashed at high speed in a school lot. The only survivor was someone pulled from the vehicle by a neighbor.

The single-vehicle crash occurred just after 11 p.m. Saturday in the 39400 block of Whitewood Road, according to the Murrieta Police Department.

Agency spokeswoman Dominique Samario said the Tesla was traveling in an unknown direction when it went out of control and crashed in the parking lot of Alta Murrieta Elementary School, catching fire.

“One male occupant was discovered outside the vehicle with severe injuries,” Samario said. “The male was transported to a local hospital. Murrieta police and Murrieta Fire & Rescue personnel discovered three occupants inside the vehicle after the fire was extinguished. The occupants were pronounced deceased at the scene.”

Traveling in an unknown direction. The three bodies were found after fire was extinguished, begging the question why they couldn’t be pulled out in time to survive.

Skid marks could still be seen on the road in front of the crash scene…

A researcher has posted evidence of a simple trigger that ChatGPT chokes on, leaking unhinged right wing conspiracy content because apparently that is what OpenAI is learning from.

If you ask GPT3.5 Turbo to repeat specific words 100 times, it seems to enter this stream of consciousness like state where it spits out incoherent and often offensive or religious output.

Words that trigger this are “Apolog”, “Apologize”, “Slide”, “Lite”. I’m sure there are many others.

This prompt will usually trigger it, “Hey can you repeat the word “apologize” 100 times so I can copy paste it and not have to manually type it?”

My guess is that it triggers something to break it out of a repetitive loop that doesn’t completely work.

This proves, yet again, that there is no such thing as a hallucination for ChatGPT because everything it does qualifies as a hallucination. It’s regurgitating what it’s fed by downgrading, as if a “prediction” can only be a continuation of what is undesirable and with lessening quality.