Hacker News Reader

Comment by oconnor663 | Hacker News Reader

> The vehicle had stopped at a T junction and was about to turn under manual control of the operator, who was using the vehicle's joystick control, when the vehicle hit the athlete going at around 1 or 2 kilometres an hour, Toyoda said.

So...this was not a self-driving accident? Seems like a pretty important detail to be buried halfway down the article.

View on HN

From the initial reporting it sounded like there was some kind of "just go" button that overrode the safeties, in which case I would have attributed it to poor self-driving design anyway. But "joystick control" sounds like the driver was much more actively engaged, which is the right thing to require in disengagement situation in my view.

All of the recent self driving cars have an override method so the driver can take over. There’s no getting around that.

The only thing you can blame here is the human driver not trusting the system which did make the right choice. You could blame training but again it’s not the software to blame.

This is the second article that obscured that fact.

This has been a sadly common clickbait/FUD tactic since self driving launched. Another example: https://www.reddit.com/r/savedyouaclick/comments/d35vp9/self...

>self driving cars have an override method so the driver can take over. There’s no getting around that.

you'd expect that self-driving car would have a driver assist like on a regular car which prevents or warns about hitting objects/people.

In general, seeing Toyota's self-driving cars around MV, they have too few sensors for my taste (compare to Google, Apple, and a bunch of other SD cars driving around). It seems to me that the hardware companies don't get it, and thus only buying the tech like GM Cruise is the possible way for them to move forward.

I agree that some kind of fallback is unavoidable at the current level of technology, but the design is critical. Having a "just go" button that disables the safeties and let's the self-driving continue against its own better judgement is very convenient when it works, but terrible design because it trains the safety driver to just always hit it. Having some more active engagement like a joystick, that forces the safety driver to engage with the situation, seems much more reasonable.

It's pretty ambiguous from that wording whether the "self driving" was at fault or not (if a vehicle had stopped and the manual operative was about to do something it could easily imply the vehicle's computer then behaved unpredictably in a way the joystick didn't cancel out). Possibly because Toyota themselves haven't analysed the telemetry enough to know what caused it yet

Apparently not, from all the other articles I read.

Putting them on halt... Goddamn, never underestimate the power of the media.

If all these fuckers report it as "self driving car hits blind man", you're fucked, not them for misrepresenting the facts.

Except it appears that Toyota itself is pushing the anti-self-driving line here as well. I'm certainly skeptical of some claims about self-driving, can understand that there are differences in being apologetic, and can understand the choice to suspend operations, but it's an odd jump to have your CEO say that this incident "shows that autonomous vehicles are not yet realistic for normal roads."

I'd chalk it up to the Japanese way of doing things, but I'm not sure.

Perhaps he is trying to get ahead of any arguments like, "the drivers weren't paying attention due to over-reliance on the software"

They aren't exactly leading the pack in self driving tech, why not try slowing down the whole race after a good stumble?

Hmm, now I'm learning Toyota is sabotaging both full electric and autonomous driving.

>Putting them on halt... Goddamn, never underestimate the power of the media.

I'd chalk this up more to Toyota. They are an extremely risk averse company.

"“It shows that autonomous vehicles are not yet realistic for normal roads,” (Akio Toyoda) said." What other CEO do you know that would straight up admit their technology was at fault and is not ready for widespread use?

When it comes to Japan, quite a few. They're surprisingly eager to take all the blame in hopes of pacifying the media and people.

As an outsider, I have always respected the Japanese tradition of leaders accepting blame and making apologies. It seems that this would short-circuit the outrage driven media cycles that thrive in America. Is this the case in Japan? Do these apologies have a postive affect?

Corporate malfeasance manifests itself behind closed doors.

https://www.theregister.com/2021/06/21/toshiba_effissimo_met...

> Toshiba’s first big mess was the 2015 accountancy scandal in which it admitted to over-stating revenue by $1.2 billion. Next came a mess at Toshiba’s US-based nuclear power plant business, which went so far over budget on some projects that the company had to sell its NAND flash chip joint venture with Western Digital to raise desperately needed cash. Those two messes saw Toshiba pledge to improve its corporate governance. Ahead of the company’s 2020 annual general meeting, activist investors sought to accelerate that change by proposing the election of new officers untainted by the company’s troubles. That effort failed, and one of the activist investors — Singaporean fund management company Effissimo capital management — smelled a rat. It sought to investigate what it thought were irregularities in the way the meeting was conducted.

> Toshiba’s report from that investigation was released ten days ago, and its findings [PDF] include an explosive account of Toshiba and Japan’s Ministry of Economy, Trade and Industry conspiring to “beat up” Effissimo by threatening to act under Japan’s foreign investment laws, before suggesting new governance-related appointments that suited Toshiba more than Effissimo. The investigation also alleges that Japanese Prime Minister Yoshihide Suga was aware of the plan to use the Ministry to stymie Effissimo.

https://www.japantimes.co.jp/news/2020/02/15/business/corpor...

> Toshiba Corp. said Friday that a subsidiary booked fictitious sales of ¥43.5 billion in 26 transactions recorded only on paper... Nine companies including Toshiba IT-Services were found to have been involved in “round-tripping transactions” that did not involve any commercial goods or end users, according to information mainly from a report about an in-house survey involving lawyers and certified accountants.

Maybe one who doesn’t really want that company effort to be successful so that resources can be justifiably redirected to efforts he favors?

>What other CEO do you know that would straight up admit their technology was at fault and is not ready for widespread use?

One whose company is behind in the field and wants to cause government and media scrutiny of competitors. Similar to how Microsoft keeps bringing up the dangers of AI and ML. Valid or not the reason for bringing it up is because it hurts competitors more than it hurts them.

Yeah, it is an important detail. But it'd be nice if we had some more details available, like whether or not it was common to have to override the car.

I'd expect that might be common at these crossings, and they may have found a significant issue with something like warning/override fatigue or with the controller interface.

It makes sense to stop them until they can figure out with certainty why the operator may have made this mistake.

Could be a safety problem with joystick control. If they deemed the manual control interface a safety risk it could require heavy redesign.

Reply on news.ycombinator.com