Comment Obligatory quote (Score 1) 312
The question of whether a AI program can make people intelligent is no more interesting than the question of whether a submarine can teach swimming...
-
The question of whether a AI program can make people intelligent is no more interesting than the question of whether a submarine can teach swimming...
-
What? You have a cheese burger now? Fine, Thank you. I am happy with my sandwich. (I see sudo wagging it's tail)
I bet that was the case.
Reminds me of Kavka's toxin puzzle.
It says
The bad news is that Google Docs has just encountered an error.
The good news is that you've helped us find a bug, which we are now looking into.
We apologize for any inconvenience this has caused you.
If you read a bit into the interview, you will be starled by what he says about Java language development and Open Source contributions to Java. He tends to focus on "engineering the Java language" as a product is more difficult than developing the language, which constitutes only 2 or 3 percent. Also goes on to say that Open Source Contributions have not really been effective.
Well, I can understand where he is coming from, but is James missing out the picture of where it actually works? - Python, Apache, Linux kernel.. Are these less used software pieces than Java? Here is the portion that I am talking about.
So long as they do that. The development of Java is not an inexpensive thing. It takes a fair amount of funding. It's not just about writing code. Learning the code is two or three percent of the expense. You're shipping fifteen million copies a week, just the bandwith is horrible. The QA when you have to worry about something that has thirty issues. When you've got everything, every stock exchange, every phone company on the planet. Their security depends on Java. So it's not a causual piece of testing.
You know, when it comes to open source contributions, our history with contributions over the years have been kinda snarky. We'd get lost of people sending code and fixes. But on average, we'd get a submission that fixed the bug but it caused three or four more. And it probably didn't fix the bug for everybody. It probably only fixed the bug for their one case. And trying to get people in the community to actually think about the whole code base and not just their particular issue today. Doing one line of change means an immense amount of testing.
Most open source projects are way too casual for that. Sometimes when you get bugs that are potential security issues, you have to move fast, you have to put immense resources on getting it done. Maybe it's just one engineer fixing one character in one line, but then testing it and making sure you didn't introduce a bug. The harder stuff is if you have a bug, there are probably people out there who have worked around that bug, so how many of the workarounds are you going to break. And when you've got nine or ten million in the developer community you have enormous applications, trivial fixes are not trivial. And open source projects, the way the average open source projects are constituted. IT's easy to get people to do the fun stuff. It's hard to get people to do the hard stuff.
Like QAing the math libraries. Like doing QA on sine and cosine, you absolutely have to have a PHd in Mathematics. Sine and cosine: it sounds really simple, but there is unbelievable amount of depths of subtlety in there. There are extraordinarily few people on the planet qualified to QA that type of stuff.
Google does not have keep it's foot of the accelerator, but it should design the upgrade process simple enough so that even grandmas should be able upgrade.
BASIC is the Computer Science equivalent of `Scientific Creationism'.