1/ White Box Autonomy β Global AV Event Observatory is now live. π
2,923 real-world autonomous-driving events, brought together in a common evidence framework β with structured exploration and an interactive map.
π https://t.co/ya7WygPEjs
In evidence, not hype, we trust.
Replace human drivers with robots, and traffic congestion will magically disappear. Right? π€
After all, humans are selfish, impatient and terrible at cooperating. Robots should be much better at communicating and working together.
That's the theory, anyway.
Now watch this clip for a glimpse of our robot-filled roads of the future.
Apparently, robots can be just as uncooperative as humans. And sometimes, even dumber! π
The lesson for AV companies? Instead of each trying to dominate the market, perhaps it's time to make their robots talk to each other.
A perfect recipe for a rear-end collision. Tesla FSD's merge strategy: commit, panic, then slam the brakes mid-merge. Anyone following us would have had a very bad day.. This has been a big headache for me for quite some time.
#FSD#Tesla#SelfDriving
What a start! First day with Tesla FSD V14.3.12, and already a potential head-on collision avoided! π
A blind bend. Parked cars on both sides. An oncoming ute suddenly pulling out around a parked car, straight into our lane.
FSD spotted the danger and braked to a complete stop. All on its own.
Hope the impressive performance continues @Tesla@elonmusk
A nice surprise from Tesla this morning! π
I've been buried in deadlines and barely paid attention to a message Tesla sent me a couple of days ago.
This morning, I discovered that FSD had been updated from V14.3.5 to V14.3.12!
Given how long I waited for V14.3.5, I honestly didn't expect this update until the end of the year.
Great to see the FSD version gap between Australia and the US shrinking so quickly! Heartening to know that our lives are finally being valued a little more equally! π
Well done, @Tesla and @elonmusk !
And the first improvement? That sudden stop in the middle of merging I posted about yesterday didn't happen this morning. https://t.co/6tSdJjxTqc
One successful merge isn't proof of a fix, but I'll happily take it! π
A perfect recipe for a rear-end collision. Tesla FSD's merge strategy: commit, panic, then slam the brakes mid-merge. Anyone following us would have had a very bad day.. This has been a big headache for me for quite some time.
#FSD#Tesla#SelfDriving
It's surprising to see Tesla FSD still struggling with such basic road rules.
FSD V14.3.5 attempted to directly cross busy traffic to reach its destination on the other side of the road, despite the clear road marking.
No obvious ambiguity. No unusual edge case. Just a straightforward road rule that most human drivers would have little trouble understanding.
Sometimes FSD handles incredibly complex situations beautifully, yet still struggles with something seemingly simple.
And that's a reminder of how much work remains to make autonomous driving truly reliable.
Did I celebrate a little too early? π
Tesla FSD V14.3.5 did an amazing job reversing out of a dead end and executing a beautiful three-point turn in a rainy morning.
I was genuinely impressed. What a manoeuvre!
And just as I was about to give FSD a standing ovation...
BANG! It ran straight over the kerb!
That certainly shook me awake!
From brilliant to embarrassing in three seconds.
Apparently, with FSD, it's never over until it's over.
Unpopular opinion: Could some of Tesla FSD's sign-recognition problems have a surprisingly simple, low-cost solution? π€
After encountering numerous speed-limit issues with FSD, I've developed a theory.
At higher speeds, FSD often misses a speed-limit sign if it appears only once. But when the same sign appears again, FSD can quickly recognise it and correct its mistake, as shown in the video below.
So here's an idea: redundancy.
Why not install speed-limit signs more frequently? Or repeat bicycle symbols along bicycle lanes?
Miss one sign? Here's another chance. And another.
Of course, too many signs could create clutter. But surely there's a sensible balance.
Could a little investment in road infrastructure help solve problems we're spending billions trying to fix with AI?
And here's the million-dollar question:
If companies like Tesla and Waymo stand to benefit from AV-friendly infrastructure, shouldn't they help foot the bill? π€
Should Australia put the brakes on robotaxis before they even take off? π€
The Transport Workers' Union warns that robotaxis could threaten up to 85,600 jobs, drain billions from Australia, and undermine road safety.
It's calling for restrictions on deployment and stronger protections for workers.
But here's the bigger question: Should Australia slow down autonomous-driving innovation to protect existing jobs, or embrace the technology while managing its consequences?
A debate Australia can't afford to avoid.
More detail can be found here: https://t.co/t9hM3UIIxq
I selected Chill mode. Apparently, @elonmusk defines βChillβ relative to his Mad mode for flying cars. π
Tesla FSD V14.3.5 took on FOUR buses and a swarm of cars in an epic battle for road space on a rainy morning.
A cautious driver would have waited, especially with one bus approaching from behind and another signalling to turn.
But FSD? Challenge accepted!
Halfway through, it hesitated briefly, perhaps having second thoughts.
Then apparently decided: Too late to chicken out now!
If this is Chill mode, I'm not sure I want to experience Mad mode. π¬
Is Tesla FSD better at changing lanes than most human drivers?
Well, definitely better than me! π
Lane changing is one of the few driving tasks I'm quite happy to leave to FSD.
Watch FSD V14.3.5 perform two consecutive lane changes on a packed main road during morning rush hour.
The second one is particularly impressive. FSD patiently communicates, negotiates, and eventually creates a safe gap to complete the manoeuvre.
My only complaint?
I still had to wave to the driver behind to say thank you!π
Perhaps someone close to @elonmusk could suggest a feature that lets FSD express appreciation (or occasional frustration!) through sounds or a thumbs-up (or thumbs-down!) displayed on the side mirror?
Now THAT would make FSD truly human-like! π
Having a bad day? Don't beat yourself up. Just look at these two photos. π
Even Tesla FSD occasionally messes up something it's incredibly good at!
Since getting FSD, parking is one of the few driving tasks I've happily delegated almost entirely to it. And most of the time, it does a fantastic job.
But then came these two masterpieces. π€¦
Photo 1: Something I might have done in my first year of driving.
Photo 2: Something I wouldn't have done even in my first year of driving!
So, next time you make a silly mistake, remember:
Even AI has its embarrassing moments. And unlike us, it doesn't even have a bad day to blame! π
Misrepresenting Tesla FSD as fully autonomous isn't just misleading. It can be dangerous.
FSD remains an SAE Level 2 system, requiring constant driver supervision.
Yet some posts on X portray it as if drivers can simply switch off and let the car handle everything.
That kind of messaging risks encouraging dangerous overconfidence.
The recent crash involving a Tesla and a California Highway Patrol vehicle is a sobering reminder of what's at stake.
Promoting FSD's achievements is fine. But exaggerating its capabilities while downplaying its limitations? That's irresponsible.
Accuracy matters. Lives are at stake.
A self-driving Tesla crashed into a California highway patrol vehicle, seriously injuring two officers inside.
According to the officers, the driver said he was asleep at the time and using the carβs driver-assistance technology.
Two near misses in less than ONE minute on a dark night. Tesla FSD's lightning-fast reactions β impressive or concerning? π³
Watch both incidents and judge for yourself: Is FSD getting better at avoiding danger, or simply reacting faster to danger it should have anticipated?
First, at the beginning of the video: FSD started pulling out of a parking lot just as another vehicle approached rapidly.
FSD stopped almost instantly. Impressive reaction time!
But here's the catch: Would a vigilant human driver have started moving in the first place? π€
Second, towards the end: While turning left alongside a passing vehicle, FSD suddenly seemed to realise there wasn't enough space.
Its response? SLAM ON THE BRAKES! π±
The hard braking automatically disengaged FSD.
Two potential collisions avoided. But were they situations FSD should have got itself into in the first place?
Reacting quickly to danger is impressive. Anticipating danger early enough to avoid the need for emergency reactions is even better.