Hardware 3 will far exceed the safety of the average human driver, though not be as good as Hardware 4.
Mostly false False in substance, right in a detail.
Prediction: Statements about events outside the person's control, checked once they fell due: did they come true?
Said for an organisation they lead or work for; it counts on both records.
What would count as true or false
The claim has two parts: (a) Tesla vehicles with Hardware 3 (HW3) will drive far more safely than the average human driver, and (b) HW3 will be less capable than Hardware 4 (HW4). Part (a) is the main test. 'So far exceed' means a large margin, roughly at least twice as safe, for example half or less of the human crash rate per mile on a comparable basis. TRUE: credible crash-rate data specific to HW3 shows its self-driving or Autopilot system (and in the end the self-driving system the claim is about) at about 2x or more safer than average human drivers on a comparable basis (similar roads and crash definitions), and HW4 shows better performance or capability. Tesla's own data counts only if it is broken down by hardware version and its method is disclosed. Independent or regulator analysis (for example NHTSA or an insurer) carries more weight. MOSTLY TRUE: HW3 is clearly safer than the average human, but by less than 'far' (about 1.3x to 2x), or the HW4 comparison is not documented. MIXED: the evidence is mixed. For example, HW3 beats humans only with supervision, or only on Tesla's own figures, which independent analysis disputes. MOSTLY FALSE: HW3 is not shown to beat human safety without a human supervising it. Examples: Tesla says HW3 cannot reach unsupervised self-driving without a hardware upgrade, or HW3 is left out of driverless or unsupervised programmes, while the only favourable data is for supervised Autopilot. FALSE: credible data shows HW3 driving at or below average human safety.
Set before the research began.
Evidence (4)
-
E1 Elon Musk admits that millions of Tesla vehicles won’t get unsupervised FSD The Verge · 22 April 2026 Own record
“I wish it were otherwise, but Hardware 3 simply does not have the capability to achieve unsupervised FSD.”
-
E2 Elon Musk admits that millions of Tesla vehicles won’t get unsupervised FSD The Verge · 22 April 2026 Own record
“relative to Hardware 4, it has only one-eighth of the memory bandwidth of Hardware 4.”
-
E3 Additional Information Regarding EA22002 Investigation National Highway Traffic Safety Administration (NHTSA) · 25 April 2024
“system controls may be insufficient for a driver assistance system that requires constant supervision by a human driver.”
-
E4 Tesla Autopilot investigation closed after feds find 13 fatal crashes related to misuse TechCrunch · 26 April 2024
“Gaps in Tesla’s telematic data create uncertainty regarding the actual rate at which vehicles operating with Autopilot engaged are involved in crashes.”
Reasoning
In E1, Musk is reported as saying that Hardware 3 "simply does not have the capability to achieve unsupervised FSD." The criteria list this exact situation as an example of mostly false. E2 supports part (b): HW3 has one-eighth of HW4's memory bandwidth. Independent sources give no HW3 crash rate showing it is safer than human drivers. NHTSA (E3) describes the system as requiring constant human supervision, and (E4) says gaps in Tesla's data leave the real Autopilot crash rate uncertain. So the main part of the claim, that HW3 would far exceed average human safety, is not shown by this evidence, and Musk has said HW3 cannot achieve unsupervised FSD [E1].
Cite this claim
Elon Musk, “Hardware 3 will far exceed the safety of the average human driver, though not be as good as Hardware 4.” (said 26 January 2023). Verdict: mostly false. Reliability API, claim c_e18822b455c3, method 0.1.0+c876030c, dataset release v1.1.2. https://epistacklabs.com/claims/c_e18822b455c3
This claim is in dataset release v1.1.2, with the same verdict.
Reply to this claim
More prediction claims by Elon Musk
-
Mostly false
“AfD is going to win an epic victory!”
The AfD party will win a big victory in Germany's upcoming election.
-
Mostly false
“for 2023, Cybertruck will not be a significant contributor to the bottom line, but it will be next year.”
Cybertruck will not be a significant contributor to Tesla's bottom line in 2023 but will be in 2024.
-
Mostly true
“if it's a smooth year, actually, without some big supply chain interruption or massive problem, we actually have the potential to do 2 million cars this year.”
-
Mostly true
“So we think demand will be good despite probably a contraction in the automotive market as a whole.”
-
Mixed
“Batteries will probably not be the limiting factor in Cybertruck production.”
Batteries will probably not be the limiting factor in Cybertruck production.
Verdicts are made automatically; how verdicts are made.