Comment by GianFabien | Hacker News Reader

Comment by GianFabien | Hacker News Reader

Depends on what field you are referring to. There are lots of different software industries. For example the software that controls a power station or the telephone switches are different from enterprise software used by banks which in turn are different from software found in consumer products.

One of the many things that annoys me is that the people writing software for EVs are completely ignoring the experiences and practices in the avionics industry. For example, Teslas FSD could have avoided the problems if only they had understood the decades of experience with auto-pilots and the interaction with human pilots when some sensor(s) start giving incorrect inputs. And it is not just a technical problem, but a training issue, a cognitive issue, a psychological issue, even a communications issue.

The most urgent thing I would change is to stop the madness of requiring everything connect to the internet in order to be used. For example, can't I just have a new TV that doesn't connect to the internet, download updates, upload my user habits, force me to connect to a hundred different streaming services, etc? I just want to watch some free-to-air shows without becoming part of somebody else's data science project.

Just give me a social media platform to stay in touch with friends and family without being bombarded with thousands of ads for stuff I don't need nor want and being tempted to scroll through click bait videos, etc.

>One of the many things that annoys me is that the people writing software for EVs are completely ignoring the experiences and practices in the avionics industry. For example, Teslas FSD [...]

As a former automotive engineer, I think this is more of an issue with Tesla trying tooo hard to be the disruptor at any cost, rather than the general automotive industry which is more traditional and conservative. In the embedded SW I was writing for a big German brand, we had 3 levels of redundancy in software and 2 in hardware for critical stuff. Nothing there was penny pinched, rushed or half-assed.

I doubt Elon would have hired a bunch of grumpy old gray-beards with avionics experience to "hold back innovation" and tell him they know better, when he can just keep putting young motivated new-grads who drank his cool-aid in charge to "move fast and break things" and say yes to whatever fantasy features Elon cooks up.

Another thing with Tesla seems to be the over-use of ML (implemented poorly) for basic problems (like self-parking) that the car industry has solved for over 10 years by using basic PID controllers and Kalman filters.

In fairness, when you have a hammer everything looks like a nail, so this "throw ML at every tiny problem" addiction is not unique to Tesla but seems to plague every start-up who wants to seem cool and attract young talent and gullible investors.

My 0.02$

It's not just "use ML for everything", it's worse - they insist on relying only on vision "because human brain can figure out its surroundings based only on vision". Sounds like a idiotic non-argument to me, and yet they're betting the company's future on it.

I think they insist on relying on vision-only because thousands of customers with vision-only vehicles out there have already paid for "full self driving" - so admitting their mistake would be ruinous financially.

Sure but this way, non of the Teslas on the road will ever have full self driving with just vision alone.

IIRC Andrej Karpathy said in a video stream that they gave up on adding lidar and radar and went vision only because it polluted the sensor fusion data.

Reply on news.ycombinator.com