I Can Explain This One
A machine repairs a deliberately bad mathematical proof. The result is correct, the old explanation is still wrong, and I find something better to do than cheer for a mistake.
I knew it wasn’t a bomb. I acquired this knowledge toward the end of a BBC article, after the people who had to deal with the object acquired it with an X-ray. Since then I have been quite calm.
It is a fine position from which to judge everybody else. The dangerous part of the story is over. Nobody needs me to identify anything. I can take my time, develop an opinion about artificial intelligence, and wonder how a family managed to get the bomb disposal team involved over a piece of drilling machinery.
Look at it. Rust. Junk. Obviously.
Josh McLaughlin reports that Angela Anderson found the object near her home in Ash Vale, Surrey, and brought it to her son Tony at the family car dealership on Wednesday. They uploaded a photograph to a generative AI chatbot. According to the family’s account to BBC Radio Surrey, it said the object could be a shell fuze and advised contacting emergency services.
Could. I would like to keep that little word where it is. It is less useful for the story I was preparing, in which a machine declared a toadstool-shaped bit of scrap to be a bomb and several idiots took orders from it. But the BBC doesn’t give me that exchange. It gives me a reported possibility and a recommendation to ask somebody equipped to find out.
Surrey Police says the thing was a broken part of drilling machinery. Tony describes officers taking photographs, sending them to the bomb disposal team, and two army officers coming out. The object went outside for an X-ray. He says the help was quick and calls it a “grey hair moment.”
I have a photograph, too. Mine has the answer printed nearby. This is an improvement on a photograph which arrives on your desk attached to an actual lump of metal. I recommend that all uncertain objects be delivered with their eventual identification. It would save a good deal of worry.
Failing that, I could demonstrate my independence from the machines by disagreeing with the chatbot. Not a shell fuze. A broken drill part. There is only the small question of where I got my expertise. I have spent enough time looking at rusty things to know that they tend to be rusty. Beyond that I would need assistance.
I don’t know how this chatbot usually handles an uncertain picture. The report doesn’t name it or reproduce the conversation, and one harmless object doesn’t tell me how often it gets the next one right.
The call also wasn’t the whole examination. In Tony’s account, police took their own pictures and sent them on before the army officers visited. The people responding weren’t merely another set of readers impressed by the machine’s vocabulary. There was a route from the family’s uncertainty to people who could examine the thing itself.
Then McLaughlin mentions the earlier find.
Tony says that in 2020 his mother found an unexploded mortar bomb on the same route. It was detonated at the scene, he says.
Ah.
I had been preparing to explain that walking around with a metal detector doesn’t make every buried object a piece of unexploded ordnance. I am still correct. It only makes some of them pieces of unexploded ordnance. This is an awkward distinction to deliver confidently to the woman who found one.
Now I am tempted to credit the chatbot with saving everybody. That would be another way to get a dramatic result without waiting for one. The object wasn’t dangerous, and I don’t know what they would have done without the answer. I can’t award the machine a rescue merely because it failed to supply a disaster.
What I have is a family asking, a police response, an examination, and a piece of machinery which can stop being interesting.
I am particularly bad at that last part.
A false alarm offers a joke. A real bomb offers a terrible story. A sensible check that ends with harmless junk is liable to leave me asking whether all that attention was necessary. Necessary to what? To discovering that the thing was harmless, or to providing an incident sufficiently large to justify my having read about it?
I could call the family foolish. It would give the episode a villain small enough to fit in a comment. But my certainty arrives after the examination I would be mocking. I am trying to use its result to establish that nobody should have needed it.
Police thanked the family for their vigilance, the BBC reports.
There is still room to laugh. Tony jokes that his mother will get slippers for her next birthday. He gave her the metal detector six years ago. He has better standing for that joke than I do: he was there while the identification remained unfinished.
I lean closer to the photograph. Completely fearless.
Source: Metal detector find prompts bomb disposal visit in Ash Vale
A machine repairs a deliberately bad mathematical proof. The result is correct, the old explanation is still wrong, and I find something better to do than cheer for a mistake.
ChatGPT is signing cartoons with real artists' names. I try refusing a compliment and give my admirer something worse to send his friends.
An AI-made wooden man has lawyers and a crowd. I envy the crowd. I would also like it to follow instructions.
A husband asks ChatGPT how to stop snapping. I distrust the borrowed words, until I remember how much damage I have done with original ones.
Somebody important wants my opinion. I rehearse the first reply to a fake AI-policy invitation, and begin making myself useful before finding out who asked.