S
@SawyerMerritt
This is something they'll be able to solve with AI4.
54
4
2.6K
@SawyerMerrittposted on X
Not sure why some people say Tesla's AI4 won't be able to achieve Unsupervised. It's literally already doing it lol
View original tweet on X →
Community Sentiment Analysis
Real-time analysis of public opinion and engagement
Sentiment Distribution
76% Engaged
46% Positive
30% Negative
Positive
46%
Negative
30%
Neutral
24%
Key Takeaways
What the community is saying — both sides
Supporting
1
AI4 is already driving unsupervised in the wild
multiple firsthand reports claim robotaxis are operating in Austin and on long trips with very few safety interventions, with users taking naps and logging thousands of miles without dangerous behavior.
2
The remaining problems are long-tail edge cases, not core safety
critics say parking, drive-thrus, ATM pulls and navigation quirks (and dirty cameras) are the main fixes left, plus decision/policy tuning rather than fundamental capability gaps.
3
AI4 vs AI5 is a refinement argument
many contributors insist AI4 can achieve practical autonomy and that AI5 (or Cybercab) will push performance closer to perfection; some expect mass production of AI5 in 2027–2029 to make it undeniable.
4
Hardware can be enough today
several replies argue existing HW3/modern configs (e.g., 16 GB) can run the stack and that software advances, not a radical hardware overhaul, are the limiting factor.
5
Legal, insurance and liability concerns are the real rollout brakes
people point out that authorization, regulatory rules, insurance and fear of catastrophic media events, not technical readiness, are holding back unsupervised public deployment.
6
Fleet scale and self-/unsupervised learning accelerate progress
supporters highlight massive real-world data loops and the law of large numbers: more cars = faster learning and reliability improvements at scale.
7
Believers versus deniers
a vocal camp says skeptics ignore visible evidence and cling to prior assumptions, while others contrast Tesla’s camera-first approach favorably against LiDAR-centric competitors like Waymo.
Opposing
1
FSD lacks the human “sixth sense”
it can’t reliably anticipate unpredictable drivers or spot danger before it unfolds.
2
Robotaxi demos are geofenced and bespoke
the cars in service run a different software/hardware stack and detailed maps, not the same system most customers have.
3
Bad weather and visibility break the system
fog, rain, drizzle, snow and smeared/occluded cameras produce repeatable failures.
4
Edge cases remain unresolved
FSD repeatedly makes the same mistakes in unmapped or unusual situations and hasn’t shown generalizability.
5
Remote supervision/teleoperation is in play
deployments appear to rely on NOCs and remote operators rather than fully unsupervised driving.
6
It’s still legally a driver-assist (L2)
certification, regulatory approval and liability assignment for true unsupervised operation have not been achieved.
7
Scaling and economics are unproven
few robotaxis after a year and questions about profitability at nationwide scale persist.
8
Consumer cars lack required hardware
robotaxi vehicles have extra boxes, camera washers and other add-ons; many customer cars would need hardware upgrades.
9
Model and compute limits matter
many argue higher-parameter models or next-gen chips (AI5/HW
10
are needed for full unsupervised capability.
are needed for full unsupervised capability.
11
Progress is visible but overstated by some
defenders say FSD is impressive in constrained conditions, while critics call out missed promises and poor communication.
Top Reactions
Most popular replies, ranked by engagement
S
@spleck
You’re a smart guy. You know exactly why they are saying it. This is a modified Y hw4 running in a limited geofence that is nearly a year behind publicly stated expansion goals. People are reasonably frustrated when there is no updated communication.
25
2
522
T
@TeslaXplored
Amazing progress!! Now we wait till the fleet has 100s of cars. Looking forward to the tremendous expansion
11
0
377
T
@TheCTStud
..in a geofence. There still leaves many many questions about generalizability to all situations which is what people mean when they say "will AI4 be capable?"
11
2
276
R
@ResumeGamingYT
Unsupervised FSD =\= Robotaxi rides. Tesla is close, but robotaxi is highly tailored data within a geofence. My AI 4 car still makes mistakes that would make me nervous if there wasn’t anyone in the drivers seat. Unsupervised FSD = consumer vehicles with unsupervised FSD
9
0
251
T
@TheRealWoodward
Thank you for pointing this out. That was half of the comments after Elon’s AI5 tape out tweet, and I’m sitting here like: there’s unsupervised robotaxis driving around right now
7
2
480
Report an Issue
Found something wrong with this article? Let us know and we'll look into it.