mike_mg

Predicting what people want to consume is hard, especially if # of impressions is your success measure. More broadly, I have been endlessly surprised by how users use my products and what in particular they liked. You do tend to get better at feeling this out from seeing people interacting with your product, but you stand to be bewildered, forever. If only you saw in what circumstances people read your posts!

At the same time, some might say it's about the area under the curve. If 10 folks get their mind blown by in-depth treatment of some curious topic, it's roughly same amount of utils as 1e5 impressions on some silly quip if you ask me.

I, for one, am perpetually grateful to lcamtuf. I have been looking up to him for like 15 years, and he has shaped me profoundly by showing what level of focus, productivity and insight is possible. You wouldn't think that someone's life trajectory can get changed by super detailed CNC lore write up, but here we are, years later (: Thanks!!

Also, if you're reading this lcamtuf, I would like to put one vote in favour of re-instating the American essays. Pls don't pull a Kafka on us! I did read your "choosing how to be remembered" post, but still, the fact that you took the American essays down feeds right into the topic of this current post. I found them positively entertaining and insightful.

wait, aren't "they" going to gather evidence on user of "my account" and associate the downloads from this domain to "me" and come after me and ask hard questions?

This is probably just caffeine addiction :)

One way that works wonders in terms of fixing bad coffee for me is diluting the drink with lukewarm, warm or hot treated water (mineral, brita, etc).

Thanks! Removed. What I meant by that is "maybe we will add donate button in the future"

I find the inverse surprising: that many algorithms that work on real-life robots _do_ _provide_ error bounds and their optimality / convergence properties are proven in the papers that introduce them.

A great example of this is motion planning, where papers both on sample-based methods (such as SST), and on search based (descendants of the A* family) argue at length the theoretical optimality and convergence properties.

On another note, I think requiring more theoretical analysis as a guarantee of safety could partially be an AI-winter meme rather than practical solution. Point in case: do people run a quick check of aerodynamics maths before boarding a flight? No - they rely mostly on the engineering and regulatory process that gradually made passenger flights safer.

+1 on that, that's a great project.

PyTorch with its explicit state that can be easily examined by hand in PyCharm debugger will be way easier for people coming into the field.

Could you please explain me how? I watched the lecture, but can't make connection. I understand how neural networks work and how they are optimized. I'm stupid. Thanks

traffic issues aside, great job. Quality resource, will recomend to cs undergrad friends