Tesla dojo, FSD timeline

[Newton]

I typically have been saying a timeline of 7-10 years for fsd. Based on where were at (not just Tesla). 2 years for limited area specific routes.

BUT I didn't know about this Dojo project which is aparently a supercomputer tesla is making. I'm thinking they will have news about it at the battery/investor day.

With this new supercomputer in the equation I'm going to have to reduce my fsd timeline to 3-5. and im now getting it for my CT. ?

Advertisement

 

[ajdelange]

Faster hardware only makes it possible for the system to make more incorrect decisions per mile driven. It's been recognized since the advent of the connection machines that the challenges lie in the software.

 

[Newton]

"ajdelange:
Faster hardware only makes it possible for the system to make more incorrect decisions per mile driven. It's been recognized since the advent of the connection machines that the challenges lie in the software.

Of course software is the underlying main thing for the systems, But dosnt an ai supercomputer imply teaching via simulations at ridiculous rates? I dont know much about ai stuff but i was talking to some1 at facebook, or maybe it was NVIDIA who does. I brought up things I thought were issues with machine learning, like having biased data because only some people use some types of sites, or always being flawed or biased because humans are flawed, and humans are programming it . He was saying that it dosnt really work like that, and they build it with these things in mind.


“We do have a major program at Tesla which we don’t have enough time to talk about today called “Dojo”. That’s a super powerful training computer. The goal of Dojo will be to be able to take in vast amounts of data and train at a video level and do unsupervised massive training of vast amounts of video with the Dojo program – or Dojo computer.”[/QUOTE]

 

[ajdelange]

If you want to really draw on the parallel of human learning consider an office with 20 imbeciles making bad decisions with respect to COVID management in a given area. The situation does not get better if we hire 20 more imbeciles for that office. Things get better when we recongize that these imbeciles are imbeciles and replace them with more intelligent people. IOW we "reprogram" the HR department.

 

[timjwright2.0]

I typically have been saying a timeline of 7-10 years for fsd. Based on where were at (not just Tesla). 2 years for limited area specific routes.

BUT I didn't know about this Dojo project which is aparently a supercomputer tesla is making. I'm thinking they will have news about it at the battery/investor day.

With this new supercomputer in the equation I'm going to have to reduce my fsd timeline to 3-5. and im now getting it for my CT. ?

Welcome to the party. Elon has been very emphatic about how close they are to FSD. He seems a lot more confident then his usual timelines for new features. I would not bet against him on this one. He is the one close to all the advancements, and using the latest Alpha version himself, and using a completly different AI approach that can learn exponentially. I think the world will be shocked in 2 years and by then FSD could cost what it is worth $100,000!

 

[Newton]

If you want to really draw on the parallel of human learning consider an office with 20 imbeciles making bad decisions with respect to COVID management in a given area. The situation does not get better if we hire 20 more imbeciles for that office. Things get better when we recongize that these imbeciles are imbeciles and replace them with more intelligent people. IOW we "reprogram" the HR department.

Yea, I see your point.
But I dont think its THAT parallel. given enough trys even a fish would learn to climb a tree

 

[CyberMoose]

I think full self driving will come pretty soon, probably within the next few years. We just have to expect to get those same warnings to keep our hands on the wheel. Tesla has an uphill battle before they can allow their software to drive without people. Even when the software is ready for it, will it become legal if Tesla is the only company with the software. I would love it if that was the case, so I could put my feet up on the dash while watching other people in traffic, but I don't think that will be the case, it'll probably be some time after Tesla is ready and then there will be some massive test that each car will have to pass before it can be 'driverless'

 

[IceCold]

This is correct.

FSD may be close (a few months to a couple years away) as Tesla defines it on their website.

Level 5 robotaxi mode is a future vision and could still be decades away.

 

[TruckElectric]

 

[ajdelange]

Yea, I see your point.
But I dont think its THAT parallel. given enough trys even a fish would learn to climb a tree

Fish did learn to climb trees but there was a lot of evolution in both the hardware and software involved in getting them to the point where they could.

 

[TruckElectric]

In this video clip the poster chose to add a portion of a song by Chinese pop singer Ma Tiao.

The song is titled 风很美(Feng Hen Mei) roughly translated to English as "The wind is beautiful" and that is the refrain he is singing in the clip.

I'm sure as an American I'm missing something that the Chinese would 'get' why the poster added this. I'd be curious to know.... ?

 

[Crissa]

There's no evidence that Model 3 has their new full-self-driving computer or software,

And secondly, I don't know why they think their flappy-flappy thing is supposed to look like a human.

I'm not sure a human would have had the reaction time, either. The mannekin basically hit the car at a run.

-Crissa

 

[TruckElectric]

 

[Jhodgesatmb]

If you want to really draw on the parallel of human learning consider an office with 20 imbeciles making bad decisions with respect to COVID management in a given area. The situation does not get better if we hire 20 more imbeciles for that office. Things get better when we recongize that these imbeciles are imbeciles and replace them with more intelligent people. IOW we "reprogram" the HR department.

Same in all organizations and governments. Considering how obvious this is, how come it is never done. Never!

 

[FutureBoy]

Same in all organizations and governments. Considering how obvious this is, how come it is never done. Never!

All sorts of reasons.

• imbeciles being lead by imbeciles
• Intelligent people not being widely available
• Imbeciles in HR not knowing how to recognize intelligence
• Intelligent people getting promoted till they come to a position of imbecility
• qualities other than intelligence being used as selective criteria
• Intelligence in a sub area of job performance being used as a measure in spite of general imbecility in overall job functionality
• Imbeciles in a group driving off intelligent people because they do not cooperate well
• Intelligent people not having sufficient people skills to work effectively with others
• Outside individuals being overly judgemental about what qualifies one as an imbecile

•