Hacker News

Top stories

Live mirror
30 storiesupdated just nowView source snapshot
  1. Claude Code reads AGENTS.md only when telemetry is on(szypowi.cz)
    132comments
  2. I Don't Want the Details(michaelheap.com)
    68comments
  3. Stripe built its internal AI platform(stripe.dev)
    11comments
  4. Jev in 25 Lines of Python(nobodywho.ai)
    132comments
  5. Z80 REPL(abagames.github.io)
    10comments
  6. GPT-6 Sol and Luna(openai.com)
    793comments
  7. Claude Opus 5.5(anthropic.com)
    1015comments
  8. Jev in practice: typed decisions, scoped authority(tenuo.ai)
    3comments
  9. Tokens Too Cheap to Meter(jyn.dev)
    39comments
  10. Two Git ignore files nobody told me about(dinculescu.dev)
    32comments
  11. Web-based IBM 1620 emulator and IPL-V from 1963(github.com/pkimpel)
    2comments
  12. QuestDB (YC S20) Is Hiring a Sales Engineer(questdb.com)
    discuss
  13. Samsung accidentally freezes its smart fridges with a software update(androidauthority.com)
    114comments
  14. The Price of Intelligence Is Falling Rapidly(marginalrevolution.com)
    27comments
  15. The GitHub wiki is an anti-pattern(michaelheap.com)
    49comments
  16. Montreal adopts bylaw banning insults against police, municipal employees(cbc.ca)
    8comments
  17. Transit rewards(waymo.com)
    249comments
  18. OpenAI GPT–6 Astra breaks Enigma message that has resisted solution since 2005(cryptocellar.org)
    411comments
  19. Microsoft killed FoxPro in 2007. Anyway, here's FoxPro revived(foxscript.org)
    223comments
  20. Show HN: Ive Sent It – online courier for files, with signed proof of delivery(ivesentit.com)
    12comments
  21. What California is learning from solar panels built over irrigation canals(kqed.org)
    531comments
  22. How did AMD Ryzen get 50% faster in two years?(lemire.me)
    167comments
  23. ReBarUEFI: Resizable BAR for almost any UEFI system(github.com/xcuri0)
    61comments
  24. 'We hacked the FBI:' Hackers say they have data on all FBI employees(404media.co)
    519comments
  25. SAML: A fractal of bad design(trailofbits.com)
    153comments
  26. Data-only attacks are easier than you think (2024)(usenix.org)
    31comments
  27. WordPress: Unauthenticated path traversal leading to conditional RCE(github.com/wordpress)
    112comments
  28. Pentagon says overreliance on AI contributed to missile strike on Iran school(bloomberg.com)
    396comments
  29. Claude Opus 5.5 Intelligence, Performance and Price Analysis (Max)(artificialanalysis.ai)
    99comments
  30. Show HN: RxFilm Studio–Create and edit your product videos with AI agent(rxlab.app)
    7comments

Comma's hands-off driving tech under investigation after 2 fatal crashes

24 pointsby 1h agotechcrunch.com
14 comments
34m agoHN ↗

Completely predictable. They’ll likely dodge a lawsuit in this case (since the car that was hit was an unoccupied police car) but it will inevitably happen.

32m agoHN ↗

This goes to show that any car, be it internal combustion, electric, manually operated or self-driving is still a car, and thus is still capable of inflicting death (among other negatives). The difference is still just in the frequency of deaths (and with self-driving this is still under debate).

27m agoHN ↗

This goes to show that any car, be it internal combustion, electric, manually operated or self-driving is still a car, and thus is still capable of inflicting death (among other negatives).

No shit.

The difference is still just in the frequency of deaths (and with self-driving this is still under debate).

No, the difference is not just frequency. It is, what is the probability I would have made the same mistake. Some of these cases would very unlikely have happened by a (sober) person. Which demonstrates, that we are still dealing with incredibly jagged intelligence.

10m agoHN ↗

Regardless, self-driving software can't make mistakes. Not one.

Saying that it's better than a human driver is not adequate. I think as a a society we (begrudgingly) make allowances for human error. There is zero solace for a death at the hands of an algorithm.

22m agoHN ↗

Some of the crashes, including at least one of the fatal incidents, may have involved the use of “forked” or modified versions of Comma.ai’s software, according to the ODI.

i am still hesitant of tesla and similarly large players, it seems absolutely wild to me that people are using a fork of a small company's software to drive their car.

18m agoHN ↗

Should it drive someone else's car instead?

17m agoHN ↗

Pretty clearly, it shouldn't drive anyone's car.

14m agoHN ↗

"…may have involved the use of “forked” or modified versions of Comma.ai’s software…"

Or, it could just be scapegoating. Like blaming AI for blowing up a girl's school.

4m agoHN ↗

i am still hesitant of tesla and similarly large players, it seems absolutely wild to me that people are using a fork of a small company's software to drive their car.

Of course they are. This is exactly why we have laws and regulations.

But there are always people who think they know better. George Hotz, the founder of comma.ai, is a very prominent one (despite falling flat on his face more than once). Who do you think is attracted to comma.ai?

“We’ve always been the North Korea of self-driving companies; we are driven by nobody else’s agenda. That’s not going to change.” -George Hotz

"Hotz is clearly a very capable engineer and has accomplished some impressive feats in his young life. However, listening to him speak at Techcrunch Disrupt he also comes off as incredibly arrogant even by Silicon Valley standards. The fundamental problem that I personally have with Hotz and many others in the valley is that they seem to have the idea that just because they can do something, there is no reason not to, regardless of the consequences for others."

https://www.forbes.com/sites/samabuelsamid/2016/11/01/though...

19m agoHN ↗

Is it legal to operate a motor vehicle modified to such an extent on a public highway? Retrofitted self driving using, as far as I can tell, completely unregulated software seems absolutely crazy to me. Free software is a good thing, but don't run your buggy shit in a way that can hurt others.

2m agoHN ↗

According to the article: Comma.ai’s own website states that the “performance” of openpilot Adaptive Cruise Control can be affected by “vehicles in the same lane that are not moving.”

Seems the AI view of the road is kind of like the game Cruis'n where the road just moves forward and you can hit things in the road and keep going. https://www.youtube.com/results?search_query=Cruis%27n

To be honest though, I have a Ford with soke form of Blue Cruise. On the highway in certain areas it can pretty much drive itself. It is terrible in traffic, cant see more than one car ahead and slams on the brakes to stop - it freaks me out, I don't drive like that. Systems like these and seemingly Comma's do not appear to have enough information to know what to do in certain situations, it is why the Ford system is limited to certain situations (and even Ford has issues as in this article).

Maybe I'm old, I haven't been super impressed with these self-driving systems and much prefer to be in control myself.

19m agoHN ↗

AFAIK their driver monitoring is very good and checks if you're looking at the road, so I wonder how this happened.