Cybertruck Q1 2025 Deliveries (Lowest to Date) [⚠️ ADMIN WARNING: NO POLITICS]

[REM]

Tesla FSD is not AGI.

I guess this guy don't know what tf he's talking about then:

Sponsored

 

[SCTesla]

I guess this guy don't know what tf he's talking about then:

It's hype. "A different kind of AGI" is redefining what AGI is because FSD doesn't fit the definition. Again, Elon even said this on his most recent JRE podcast.

Tesla as a company says a ton of misinformation, but most large companies do.

 

[Gigahorse]

However, that is due to the battery pack size of ~75 kWh to ~123 kWh (1.6+ times bigger)

Yea at the end of the day efficiency is one thing but the other multiplier is just pure battery energy capacity.

 

[voxel]

I guess this guy don't know what tf he's talking about then:

If pathfinding is AI... then that's a pretty low bar. That Tesla was just meandering about looking for a connecting road. There wasn't much "planning" which is the core idea of AGI. LLMs are still pretty basic these days.

 

[dalton108]

If pathfinding is AI... then that's a pretty low bar. That Tesla was just meandering about looking for a connecting road. There wasn't much "planning" which is the core idea of AGI. LLMs are still pretty basic these days.

AND … if you listen to some very credible experts, LLMs are a dead end on the pathway to AGI.

 

[REM]

If pathfinding is AI... then that's a pretty low bar. That Tesla was just meandering about looking for a connecting road. There wasn't much "planning" which is the core idea of AGI. LLMs are still pretty basic these days.

You know what other autonomous systems would have done in that scenario? Come to a complete stop and put on the hazard lights; waiting for human intervention lol.

I think we have seen so much slow progression over the years that people are becoming unimpressed with further progress unless there is a major breakthrough. And I doubt there will be just one "earth shattering" update that acts as a flashpoint.

If the car isn't making decision in this scenario, then what do you define it as?

 

[REM]

AND … if you listen to some very credible experts, LLMs are a dead end on the pathway to AGI.

Not sure how that would make sense, as natural and conversational language is a core component of perceived sentience and intelligence.

 

[voxel]

AND … if you listen to some very credible experts, LLMs are a dead end on the pathway to AGI.

I believe it. LLMs are just one tool. AI is so vast and somehow we've just decided to brute force neural networks to "solve everything"

 

[voxel]

You know what other autonomous systems would have done in that scenario? Come to a complete stop and put on the hazard lights; waiting for human intervention lol.

I think we have seen so much slow progression over the years that people are becoming unimpressed with further progress unless there is a major breakthrough. And I doubt there will be just one "earth shattering" update that acts as a flashpoint.

If the car isn't making decision in this scenario, then what do you define it as?

It's just pathfinding. It's not AGI at all. The AI doesn't understand squat. It lacks the concept of the world.

I argue FSD and most autonomous systems can't handle exceptions very well. They lack a model of the world and how it works. That's why I don't believe FSD as designed - will work in the real world which is full of exceptions. The argument for LLMs is you can train the model with all these exceptional situations (like a disabled car in the middle of an intersection or u-turning car in a traffic circle lol) but that list of exceptions is massive... dare say nearly infinite.

 

[dalton108]

Not sure how that would make sense, as natural and conversational language is a core component of perceived sentience and intelligence.

Well, Pal-o-mine, as I understand that the argument, it’s that we will never achieve AGI from a simple “word calculator” because it needs to know/perceive many more things about the world in order to have general intelligence. I don’t wanna get metaphysical about it but as Alan Watts said, ‘the word water does not tell you what water is.’ Or put another way the word water isn’t water, it’s a weak symbolic approximation of it. Same for goes for LLMs’ understanding of reality.

But here’s the man himself explaining why he more “adventurous colleagues” are wrong in his opinion:

 

[REM]

It's just pathfinding. It's not AGI at all. The AI doesn't understand squat. It lacks the concept of the world.

I argue FSD and most autonomous systems can't handle exceptions very well. They lack a model of the world and how it works.

Lacking a model of the world?! What do you think this rendering is? Honest question.

Are you guys aware that your personal Tesla is uploading telemetry after every drive you take, and that dataset is being tossed into dojo for further training? FSD is learning on real world data.

Are you guys trolling me here? Not joking. I feel like some of you here are just messing with me.

Tell you what, I think I saved a clip from the dashcam where my CT performed real-time path tracing for a pedestrian that was crossing a multi-parted crosswalk. It clearly made a decision to stop and wait for them, even though it was technically clear to proceed.

 

[dalton108]

Lacking a model of the world?! What do you think this rendering is? Honest question.

Are you guys aware that your personal Tesla is uploading telemetry after every drive you take, and that dataset is being tossed into dojo for further training? FSD is learning on real world data.

Are you guys trolling me here? Not joking. I feel like some of you here are just messing with me.

Tell you what, I think I saved a clip from the dashcam where my CT performed real-time path tracing for a pedestrian that was crossing a multi-parted crosswalk. It clearly made a decision to stop and wait for them, even though it was technically clear to proceed.

Did you watch the video? Unless you’re doing semantics around the definition of AGI but I don’t know what you’re on about. FSD is definitely not AGI.

 

[REM]

Did you watch the video? Unless you’re doing semantics around the definition of AGI but I don’t know what you’re on about. FSD is definitely not AGI.

have you been following the whole thread here? I already addressed that at length:

I'll respond to the video when I get a chance to watch it. Though I'm already having a laugh at the immediate bias with the sensational title of "godfather of Ai".

 

[dalton108]

have you been following the whole thread here? I already addressed that at length:

I'll respond to the video when I get a chance to watch it. Though I'm already having a laugh at the immediate bias with the sensational title of "godfather of Ai".

actually no I haven’t been following this one that closely. And after you do you can tell me whether this guy is not considered the Godfather of AI. I don’t work in this field and I don’t presume to know anything but I have no reason to believe that the channel that it’s posted on is political and that the moniker isn’t well deserved.

But, again, you let me know your thoughts after.

 

[SCTesla]

actually no I haven’t been following this one that closely. And after you do you can tell me whether this guy is not considered the Godfather of AI. I don’t work in this field and I don’t presume to know anything but I have no reason to believe that the channel that it’s posted on is political and that the moniker isn’t well deserved.

But, again, you let me know your thoughts after.

REM blocked me for disagreeing with him about AGI.

Sponsored