@sodazerosugar@jasminedarling Blast freezer is the most practical. But laser cooling https://t.co/Rl4fpSOepS is the most similar to a microwave but with the opposite effect. Injecting EM energy to create movement via a standing wave = microwave. Injecting EM energy to cancel movement = laser cooling.
@tscottme@TFloodeen And if you keep reading a few more verses in Gen 3:22: "And the Lord God said, Behold, the man is become as one of us, to know good and evil..."
@DirtyTesLa So all the work that Tesla has put into FSD to train our cars to switch lanes and avoid being behind semi trucks at stop lights is now useless? =)
Yeah, it will exit cleanly an exit early about 5% of the times, and it will get the wobbles like the videos above maybe 1% of left turn on small round a bout attempts.
I've use FSD for 90% of my driving and love it! It manages busses, police, heavy pedestrian areas, construction, and many other edge cases with ease. It is amazing technology!
Hoping 14.3.1+ will fix the round-a-bout wobbles when it rolls out wider.
I have similar issues with the indecision around right turns. Typically happen at round-a-bouts where navigation wants to go left, but FSD on occasion will oscillate between staying in the circle, or exiting one exit early. I keep vigilant at round-a-bouts to not climb the curb...
I would say that calculation requires intelligence.
At its heart calculation is comprised of memory and operators. 2+2=4 requires 2 memory spots and the ability to operate the steps to add. Even basic adding requires managing carry, overflow, rounding decimal places, number type (decimal, imaginary, negatives, etc...) More complex operations will be made of multiple levels of operations and larger memory needs.
I think it takes intelligence to hold the memory and correctly apply the operation steps in our neurons, or it takes intelligence to build a machine capable of calculation in a mechanical/electrical way.
Theoretically/mathematically this is true, you could build a game tree with all the potential moves where each leaf node is a win/loss/draw. Then all you have to do is know what branch the game is currently on and steer it with your moves to the branches where you win. No evaluation (intelligence) function needed.
But this is false in the real world. Best estimate is that there are 4.82 x 10^44 different board positions. To hold one game of chess in memory it takes about 399 bits. To hold all of the positions in a linked chess game tree it would take 4 x 10 ^122 bits. This is more bits than the estimated number of atoms in the universe, or in other words, if you could turn every atom in the universe into a perfect 1 bit memory cell, you couldn't hold the whole chess tree. This is theoretically finite, but It is practically infinite.
There are "endgame tablebases" were we have solved chess but only for end games where there are 7 pieces on the board left. These are useful as you approach the end game. A 7 piece end game tree is ~17 TB of data.
Chess will always require an evaluation function/intelligence to drive to the most winning outcomes.
@teslaownersSV Also today, I was headed towards a round about with navigation wanting to turn left (3rd exit), but the car decided to leave early and take the 2nd exit. This one was pretty smooth, but sometimes it is very indecisive and jitters back and forth.
I have noticed since 14.2.2.2 my car will sometimes take right turns when I am not expecting it. Here is a clip from today where I was going straight down the main road, and FSD decided to take a right turn, go through a Canes parking lot, and then get back onto the main road going straight again.
@YangTrader@Tesla I have seen the same three times now. It occurs on small roundabouts for me. It seems to be trying to decide to exit (one earlier than navigation says) or continue to the correct exit
@Marco_Piani@DavidMoss Sentence #2 from the article:
“A Washington state Tesla owner rode his way into the record books after completing a fully autonomous coast-to-coast drive across the United States, which he detailed in a viral X thread.”
If you want to buy new, the cheapest would be the Model 3 Standard at $38,630. You then need to buy Full Self Driving (FSD) Supervised at $8000 one time purchase or $99 a month.
Currently the technology requires a driver to be in the drivers seat supervising and ready to take over if needed. How this might work legally or practically for a bilateral amputee, I don’t know.
In the very near future Tesla plans on releasing an unsupervised version that will be used in their Robotaxi service. Sometime after that they will release it for individuals, but timelines aren’t yet well known for FSD Unsupervised. I hope it is sooner than later for you and the many others who would benefit from greater mobility/autonomy.
I've been doing a bit of testing on this as I am getting some similar results.
It feels like it struggles to set the navigation when there is auto navigation to home or work based on your typical pattern. It also seems to struggle when you have a multi-point route and the tesla shows a button that says "Continue Route."
It also struggles to set the route if I try to use any of my saved favorites. When asked specifically about favorites, Grok says it doesn't have access to the saved favorites.
Single point routes seem to work very well, multi-point routes struggle more.
This is a fun integration to play with, and look forward to the increased utility. Much better than typing on the screen while driving.