Autonomous Vehicles Powered By End-To-End Deep Learning with Alex Kendall, Wayve.ai

[shrineofchance]

Autonomous Vehicles Powered By End-To-End Deep Learning with Alex Kendall, Wayve.ai - FYI - For Your Innovation

The FYI - For Your Innovation Podcast offers an intellectual discussion on recent developments across disruptive innovation—driven by research, news, controversies, companies, and technological breakthroughs. Hosted by ARK Invest, ARK and guests provide a unique perspective on how to best...

pca.st

 

[voyager]

[URL unfurl="true"]https://medium.com/@RebelScience/why-tesla-cannot-solve-full-self-driving-378a8214b4ff[/URL]

The cortex instantly accommodates existing perceptual structures to see the new pattern. No new structures are learned. A neural network, by contrast, must be trained with many samples of the new pattern... The AI community thinks they know better. They do not. They are lost in a lost world of their own making. Computer vision is completely unlike visual processing the brain.

 

[Khashoggi]

Tesla has changed their approach several times now for FSD. It is likely they will continue to do so until they find a solution.

Somebody saying for years that something can't be done, should be ignored. That's not a solution.

Neural nets are one piece of the puzzle, if you were to look at brain function, especially in the Cortex, there are areas called Cortical Columns which could be similar to functions that tie together the networks. We don't understand brain function yet, so are trying to represent what we can in silicon.

This is the problem of - there are things we know, there are things we don't know, and there are things we don't know we don't know.

Edison was asked if it was true if he failed 10,000 times to invent the lightbulb, his response was no, that he found 10,000 ways not to make one.

 

[voyager]

This is the problem of - there are things we know, there are things we don't know, and there are things we don't know we don't know.

Ah, the Donald Rumsfeld observation. Splice in: "there are things we know we don't know"...

 

[EVNow]

[URL unfurl="true"]https://medium.com/@RebelScience/why-tesla-cannot-solve-full-self-driving-378a8214b4ff[/URL]

The cortex instantly accommodates existing perceptual structures to see the new pattern. No new structures are learned. A neural network, by contrast, must be trained with many samples of the new pattern... The AI community thinks they know better. They do not. They are lost in a lost world of their own making. Computer vision is completely unlike visual processing the brain.

Actually, the differences between CNN and biological neural networks are well know and discussed in AI community.

There are even new NN proposed to make them more like the brain - but that has not taken. I’ve several posts on this in the past …