Bizarre Hey Lucid Interaction

[hmp10]

I'm shocked and can't believe it is happening in real life. I think it's very unlikely this "feature" will be deployed in Europe (it will be very likely against EU regulations). Still very unlikely I'll consider a car with this kind of homemade "assistant" that listens to all conversations all the time.

Yes, it's one thing for a computer program to listen to sounds in order to trigger when it hears a certain sound sequence. It's another thing for the program to analyze the content of a conversation in order to tee up suggestions relating to that conversation.

 

[xponents]

In a sense, yes. We’ve often had Alexa ask if it should order more xyz based on our ordering history. We find this annoying and turned it off.

Doesn't it say that after you are about to buy some other things? For example, would it give you that ad in the middle of dinner? I'm genuinely curious, as we are a Google Home family, but nevertheless, I think it can be agreed that (if true), Lucid doing this is a terrible step.

 

[tonybot]

I agree, too. I have had my fair share of experiences where some CS folks were clue-less or guessing. As someone in cyber security, if this is true, it concerns me. I am not sure where anyone consented to this level of surveillance. If they do this, huge privacy and legal risks are involved, including the risk of litigation. I hope this is a clueless CS associate and the Lucid leadership will clarify.

 

[Lucken]

I do agree, but isn't the Lucid's microphone that weird flux capacitor thingy on the roof (which isn't pointed at the driver)? For families, voice commands (by passengers) are often used, and I could imagine many other scenarios where a passenger would need to talk.

I still find the microphone to be very directional where even my wife in the front passenger seat is usually misunderstood…and she speaks very clearly.

 

[hmp10]

I do agree, but isn't the Lucid's microphone that weird flux capacitor thingy on the roof (which isn't pointed at the driver)? For families, voice commands (by passengers) are often used, and I could imagine many other scenarios where a passenger would need to talk.

Correct. It's that three-pronged slit in the crossbar between the front and back seats.

 

[levsha]

Question tho, does anyone think those Amazona and Alexa devices aren't listening to you?

Amazon, Google, Apple spent a lot of the time tuning their assistant to properly detect a magic command at a comparably low level, and to only start listening to you more carefully after this magic command. If they will be caught that they intentionally listen to all your conversations in your car they will very likely get a huge fine, at least in EU.
I can't understand how people can advocate such bad behavior using rumors like "Amazon is spying on you".

 

[xponents]

Amazon, Google, Apple spent a lot of the time tuning their assistant to properly detect a magic command at a comparably low level, and to only start listening to you more carefully after this magic command. If they will be caught that they intentionally listen to all your conversations in your car they will very likely get a huge fine, at least in EU.
I can't understand how people can advocate such bad behavior using rumors like "Amazon is spying on you".

Very well said. This articulates my feelings perfectly.

 

[hmp10]

I still find the microphone to be very directional where even my wife in the front passenger seat is usually misunderstood…and she speaks very clearly.

My usual front-seat passenger has a distinct Polish accent that even I sometimes misunderstand, but both "Alexa" and "Hey, Lucid" understand him from the passenger seat about as well as they do me from the driver's seat -- which means usually but not always.

 

[Lucken]

Doesn't it say that after you are about to buy some other things? For example, would it give you that ad in the middle of dinner? I'm genuinely curious, as we are a Google Home family, but nevertheless, I think it can be agreed that (if true), Lucid doing this is a terrible step.

Nope, it will do it anytime.

 

[Cuestix]

I had the "Hey Lucid" assistant pick up conversations a couple of times and I assumed that it was just too sensitive to anything that sounded remotely like "Hey Lucid." At least one of the times we were talking about the car, so the word "Lucid" probably was said before it started parsing our conversation. I just turned off the voice activation in settings and press the button on the wheel if I need the assistant. Seems easy enough to work around the issue. If I can't reach the button with my thumb, I'm probably in more trouble than "Hey Lucid" can offer for help.

 

[hmp10]

I agree, too. I have had my fair share of experiences where some CS folks were clue-less or guessing. As someone in cyber security, if this is true, it concerns me. I am not sure where anyone consented to this level of surveillance. If they do this, huge privacy and legal risks are involved, including the risk of litigation. I hope this is a clueless CS associate and the Lucid leadership will clarify.

I sent an email to Marqie Carter at Lucid to inquire when I first heard about this. I haven't heard back yet.

 

[Babyrocko1908]

Amazon, Google, Apple spent a lot of the time tuning their assistant to properly detect a magic command at a comparably low level, and to only start listening to you more carefully after this magic command. If they will be caught that they intentionally listen to all your conversations in your car they will very likely get a huge fine, at least in EU.
I can't understand how people can advocate such bad behavior using rumors like "Amazon is spying on you".

I didn't say "spying," I said listening to you I'm certain for algos and such. I admit I didn't read the terms and conditions when I plugged in my Echo. I'm pretty sure it was disclosed in there somewhere. I simply decided to discontinue using it.

 

[xponents]

Nope, it will do it anytime.

Well, I stand corrected. Sorry about that. Nevertheless, I think we can all agree this is a step in the wrong direction. Just because Alexa does it, Lucid doing the same is not justified.

 

[Lucken]

My usual front-seat passenger has a distinct Polish accent that even I sometimes misunderstand, but both "Alexa" and "Hey, Lucid" understand him from the passenger seat about as well as they do me from the driver's seat -- which means usually but not always.

Interesting, that hasn’t been our experience. It will recognize me with nearly 100% accuracy, but my wife not so much. Either way I much prefer the steering wheel button launch for any of these ‘voice assistants’. I just feel like a dork with ‘hey Siri’, ‘hey Lucid’. After hearing some of the experiences here, it’s no great loss for me to turn off ‘hey Lucid’.

 

[fisheral]

One of the exact phrases my wife remembers (among others) was, "Relationships can be difficult. Would you like to talk more about this?"

 

[hmp10]

I just turned off the voice activation in settings and press the button on the wheel if I need the assistant. Seems easy enough to work around the issue.

Why should we have to work around a convenience that is becoming standard in luxury cars -- i. e., passengers' ability to initiate voice command -- in order to avoid having unwanted tee-ups for ads thrust upon us?

Again, it is mystifying how readily people are willing to tumble to yet another advertising intrusion into our lives. Aren't constant spam calls and texts on our land lines and cell phones and constant ads on streaming services for which we already pay a fee enough to tire us of this?

I remember years ago when CNN first got started, one of their big selling features was that, by paying a fee for their news, you would be spared advertisements. Now, 80% of the time I switch to CNN on my paid-for cable feed, I find they're at commercial break and I move on to another channel.

Do we really want our cars to be next?

 

[Lucken]

One of the exact phrases my wife remembers (among others) was, "Relationships can be difficult. Would you like to talk more about this?"

The first question you need to ask is the advice that LA is giving any good?

 

[AirQuality]

I have a decent amount of experience with AI assistants in my line of work. OP’s interaction can best be summed up as a combination of two factors:
1. Extreme sensitivity on the car’s part to the “wake word”, ie “Hey Lucid” — not helped by the fact that there are a lot more things that sound similar to “Hey Lucid” than to “Hey Siri” or “Alexa”.
2. The general tendency of LLMs to try and be helpful if they believe they are being asked questions.

So if we assume that Hey Lucid is connected to a large language model from one of the big providers (OpenAI, Google, Anthropic, Meta, etc), the typical response from such a model today is to answer the question that it thought it heard, and offer follow-up questions should you need anything else.

I do not believe that there is any intent to sell anything at least in OP’s case. That’s just LLMs doing their thing. It feels intrusive because Hey Lucid thought it was being addressed, which is because Hey Lucid is way too sensitive.

Also AFAIK the mics are only in front of the mirror, not in the center of the roof.

 

[AirQuality]

One of the exact phrases my wife remembers (among others) was, "Relationships can be difficult. Would you like to talk more about this?"

This is a classic LLM response.

 

[AirPlur]

The first question you need to ask is the advice that LA is giving any good?

I wasn't going to say it at first but that was my initial curiosity. If it is then the savings in therapy alone could justify the payments. Then again, if its LLM is only trained by other Air owners...