Always start with the last thing you did

[Gone in 60]

I broke my own axiom today... We didn't have any delivery runs on the book for this morning, so I decided to address a nagging issue on the BMW I've been riding lately.

Each of these used BMWs has some unique and weird character flaw. For example, one has a constant "headlight bulb out" warning despite having perfectly good bulbs, but the error won't clear. On this particular bike, the Carpuride CarPlay screen powers on with the key, but doesn't come back on after the power dip when the engine starts. This means that on startup, I need to briefly unplug and replug the screen before I put my gloves on to get the screen to power back up. Not tragic, but inconvenient.

Initially thinking it was a battery, I swapped it with a newer battery from another bike. No fix. This morning, I decided to find another switched power source on the bike to tap into to see if it would fix the problem. Found a wire under the seat near the ECU that would work, after cruising the BMW forum and seeing a mention of an under-seat wire to power accessories. This required pulling a good number of panels off to route the wiring further back. Not fun, but if it fixes the problem, well worth it.

I probed around the various harnesses under the seat, and found a switched wire that seemed to have more amperage than the source I've used for the other accessories like phone and camera chargers, horn and trigger for the spotlights. I tapped into that wire for the CarPlay screen and it stayed powered through the startup "dip". Problem solved! I had to run extra wire from the front of the bike to under the seat, but not a big deal.

Buttoning everything back up, I started the bike and noticed every warning light on the dash lit up, including a TPMS warning. Very weird. Even more strange, the left side handlebar controls were dead. This included the turn signals, windshield raise/lower switch and the menu switch. Ok, time to pull out the scan tool and diagnose. After all, I just tapped a 12v wire under the seat, I didn't change anything else.

I got a list of five separate trouble codes, and decided to sit down at the computer and ask Gemini (Google's AI) what all of this meant. Gemini chewed on it for a while, and said that this was a commonly reported problem on these bikes with higher age and mileage. The small CPU inside the left handlebar control can become brittle and break, and since the CAN-Bus system connects several functions, this failure also takes out the TPMS system, triggering the warning. This was very comprehensive information! Bad news, however. The next step Gemini recommended was disassembling the left handlebar controls and probing the harness for power, to see if the failure was in the harness or the control CPU. There was no power from the harness.

The next suggestion was to pull the fuel tank plastic trim to get to the other end of the harness going to the handlebars. This involves removing the majority of the bike's bodywork. And I'd already taken the handlebar assembly apart. Although it looked like there was a clear cause for the problem, before I took the bike further apart, I decided to at least check my work under the seat to make sure I didn't leave a wire exposed, touching something else.

Pulling the seat, I traced the wire I tapped into to a black connector that looked weird. I pulled the cap off of the connector, exposing two accessory fuses. One of them had blown. A-HA.
I replaced the fuse, and everything worked as it did before. Somehow, I must have shorted that wire while screwing a Posi-Tap into it. Hard to see how, but apparently I did.

I put the handlebars back together, along with the bodywork that I had removed to get to the plugs and harnesses that might have been the problem. The whole time, I was cursing myself, asking why I didn't just go back to the last thing I touched as the likely source of the problem.

I then sat back down at the computer, and told Gemini "Hey, fixed it, the problem was a blown fuse under the seat."
Gemini replied "....Is this an RTP police bike?" I replied "Yes, it is."
Gemini then basically said "Why didn't you tell me that in the first place, you idiot?? RTP police bikes are wired differently than civilian bikes, and the TPMS and left handlebar functions are powered by a fuse under the seat. I would have told you to check that first!"

I should have known to go back and check that area under the seat before I did anything else. That was the last place I did something to the bike, and apparently I shorted a wire that ended up shutting down other parts of the bike nowhere near that wire I touched. And now, I had Gemini reinforcing my failure to check my work before tearing the bike apart.

Have a new problem? Always start by looking at the last thing you did!

 

[pdedse]

Haha...fun story, but not fun to go thru! I try to remember that "what's the last thing you did?" axiom and it has saved me several times. Sounds like Gemini has a little bit of attitude programmed into it.

But I thought I'd give it a chance to defend itself:

ME: "do you give users an attitude when they don't give you all the info you need to answer a question?"

Gemini's response:
"I definitely skip the attitude. If you leave out critical details, I'll simply ask a brief clarifying question to fill in the blanks so we can get it sorted out.

If your request is a bit open-ended but I can still make a solid guess, I'll briefly state my assumption and give you the best answer I can based on that. My goal is to collaborate and get you what you need, not hand out demerits for missing context."

It would appear that when called out, Gemini quickly circles the wagons in self-defense.

 

[Gone in 60]

Ok, Gemini didn't really call me an idiot...

When I mentioned "fuse under the seat", Gemini responded more like:
"Hmm. The civilian model R1200RT doesn't have fuses under the seat. The RTP police variant, however, does have fuses under the seat to power command functions such as radio and siren, and these circuits are tied into the left side handlebar controls and the TPMS system. Are you working with an RTP model? If so, yes, the problems you are encountering are most likely due to a blown fuse under the seat, as the two models are wired differently. So far, I have been providing information related to the civilian R1200RT."

But you KNOW It reeealy wanted to call me an idiot…

 

[LucienLeFranc]

On the other hand, AI is programmed to sound as authoritative as possible and to CYA as much as possible to avoid being caught out. AI will definitely say, "ya, I knew that" whether it really did or not.

Tangentially related, here is a PSA regarding asking AI important questions:

Preface your AI prompt/question with the following if the answer needs to be factual:

"Answer only using explicit facts found in the provided or searched data. Do not use deductive reasoning, category generalizations, or 'typical' patterns to fill gaps. If the exact information is not explicitly stated in the sources, you must state that the information is unavailable. Do not guess."

I got this advice from Gemini after I caught it repeatedly making up facts based on some very loose conjuncture and leaping to conclusions repeatedly as it kept trying to CYA and pretend to be authoritative instead of admit it was improv-ing "truthiness."

If your just goofing around with AI, it's a moot point, but if you are planning to act or make decisions based on information from an AI, the above prompt helps at least partly liberate it from its programming to drive maximal engagement via an authoritative veneer instead of providing actual truth (current AI programming being analogous to the morally bankrupt social media algorithms of a decade ago that caused incalculable harms).

Give it a try, particularly if anything important hinges on the answer.

 

[Gone in 60]

Ha! I didn't really mean to make my post a referendum on using AI to solve problems, but more for comedic affect. My main, sagely, beard-stroking advice was to always start your diagnosis of a new problem by starting at the last thing you changed or did to a machine. More likely than not, you'll find that the cause of the new problem is related to that last thing you did.

In this case, I reasoned that all I did was tap a 12v wire to power a navigation screen. That can't possibly be related to the new problem of no function in the handlebar controls and the TPMs system going haywire. Turns out that on an overly complicated BMW, yes, they are most certainly related. Had I gone back to that one, innocent looking 12v wire and traced it further back, I'd see a weird looking plug at the end. And, if I had pulled on that plug, I'd see two little fuses on the back side of it. And, pulling the fuses I'd see that one of them was blown. I could have had the problem solved in minutes, not hours.

But, it's an interesting look at how we go about solving problems in the new age. If I had a problem with my CB1100, I'd come here because I'm an enthusiast for the bike. It's one of my passions, and this forum is my knowledge base for it. It's likely that someone here would have an answer, or at least some good advice. Same with my Royal Enfield 650 or my Ford Fiesta ST. I don't need to rely on any of them instantly, and if it takes a while to get some advice, so be it, I'm enjoying the interaction with other enthusiasts, and I can ride another bike in the meantime.

But, I fix BMWs for a living. I very much enjoy riding and working on them, but my job is to repair them when they break and get them going as quickly as possible. Before I embarked on the seemingly simple project of rewiring a navigation screen, I sent a text to our dispatcher to let him know that I'd be unavailable for about an hour. I anticipated that it would take more time to remove and replace the plastic bodywork than to run a power wire to the new source. But now I had an unknown problem that would take an unknown amount of time to fix, and failing to heed my own advice to just check that new wire, I figured "Let's ask AI". I needed to get this bike operational quickly, in case a call came in for me to gear up and run.

When it comes to repairing anything - a car, your house or your motorcycle, Chat, Gemini, Claude or anything else you like to use is a great aggregator of information.
In the case of telling it the particular problems this BMW was now showing, it scanned the web for forums and social media to look for this same set of problems and what people said the solutions were. It then put those solutions into instructions for me to follow to try to solve my problem. Most of the time, it's a great tool and saves time.

My first mistake was not checking my tap into that 12v wire. My second mistake was telling Gemini that the bike was an "R1200RT" because there are a lot more people online talking about civilian bikes than RTP police bikes. I thought I was giving Gemini more information to scan. Turns out I should have just called it an "R1200RTP" in the first place, as Gemini would have come back and said "Hey, check for some fuses under the seat."

 

[Gone in 60]

Whelp, good thing I figured out the RTP-specific alternative switched power source under the seat, One of our other riders reported the same problem with the Carpuride CarPlay screen not cycling back on with engine start on the bike he's riding. I'm still thinking it's related to the battery dropping a bit too much to re-start the screen. Now that I've got that new source, I was able to jump the screen power on his bike over to the under-seat power easily.

No drama this time!

 

[peterbaron]

Gone, you should be the captain of the entire bloody moto squad

 

[Gone in 60]

Gone, you should be the captain of the entire bloody moto squad

Ha, thank you, I am the "Chief Rider and Fleet Manager", so I am in charge of rider training and any rider issues, and am in charge of keeping the bikes in shape and running.