1 karma · joined December 4, 2019
Google just recently refused to release salary numbers across their workforce because it "would be too expensive" which is really another way of saying the truth would paint them in a very harsh light. Going further back the big three were colluding to keep engineer salaries low with non-poaching agreements.
The industry as a whole is unhealthy and it is important to continue calling out all the bullshit and not becoming complacent with the status quo. For every Susan Fowler there are a few thousand that suffer in silence or worse yet they speak up and are punished.
I don't think optimizing a rats nest of functions with SGD is that impressive. If a model lacks explanatory power what use is it to people?
In this case when the solution from the NN is incorrect what recourse do you have? The actual solution might be an arbitrary permutation of what the NN gave you and there is no way to tell which rows and columns will have to be reshuffled to get the actual solution or even if there is a solution. The constraints might be inconsistent and you will never know.
Maybe an NN for Sudoku should use AlphaGo type of architecture.
People think this is a tool problem but it's not. Well, the tools kinda suck but fundamentally it comes down to culture and mind share. My background is in pure math but my programming knowledge is steeped in the hacker tradition. It is very hard to overcome all that momentum.
There is no killer application for verified programming. Whatever you can do in Coq in 10 weeks you can do in C in 2 weeks with a few extra buffer overflows. We'd be further ahead if there were more tools for validating C code than trying to rewrite the world with dependent types.
False. There is a famous quote by Ford about markets and faster horses. Point being the market doesn't know what it needs or wants until someone actually makes it because typically people just want more of the same but faster and cheaper.
Most walled gardens are built for convenience of consumption whereas most federated networks seem to assume a more active and informed participant. The kinds of features you'd build for one group are at odds with what you'd build for the other one.
Then again Brave seems to be tackling the problem from the right angle. I hope their model takes off and people start incorporating similar ideas into other open networks that respects the network participants instead of just treating them as passive consumers.
There is no logical or information theoretic perspective that justifies or explains conformance behavior. It's all pure biology.
Really the problem is writers and journalists who are not domain experts trying to explain this stuff to laypeople. There is no requirement for story writers to have been mathematicians or biologists when it comes to popularizing so you end up with half-truths and outright lies.
I've written integer programs, ray tracing simulation software for fields of heliostats, custom orchestration tools and deployment pipelines, etc. and not once has anything said or done during a whiteboard interview been relevant to any of it. The actual job is pretty much orthogonal to whatever the interviewer thinks they are figuring out during a whiteboard interview and as above the stuff I've worked on goes over the heads of 90% of the programmer population so 9/10 they can't and are not qualified to judge things properly but somehow they still think through the magic of the whiteboard they'll divine the candidate's true capabilities.
When working on hard problems I go to the library and don't stand in front of a whiteboard to see if I'm smart enough to reinvent some wheel. Chances are someone else has already figured it out and I just need to read some paper and then implement it. If they haven't figured it out then that means the problem is harder than I thought and will either need to rethink my approach entirely or spend my time on something else.
Unless of course the job is to stand in front of a whiteboard and wave furiously, in which case you should be using a whiteboard to interview.
The first principle is that you must not fool yourself – and you are the easiest person to fool.
So I don't get why he's using Feynman as an example here. Feynman was a very inventive and precise thinker. I think the post starts with a false dichotomy so whatever conclusions are drawn are resting on pretty shaky grounds.