HN user

ballooney

1,341 karma
Posts0
Comments288
View on HN
No posts found.

I have rarely had such displeasure from hearing such an interesting interviewee having to cope with such a cretinous interviewer. I’m very surprised you recommended this.

I don’t like this site’s obsession with reducing everything to market opportunities, but… it’s extremely well documented that land mines, white truffles, cancer, diabetes, chemical weapons, etc can all be ‘sniffed’ by animals and it’s a mechanism that is almost always ‘better’ (cheaper, quicker, more deployable in the field) than human-engineered solutions. Surely there’s some vebture capital opportunity here for better sensors that would unarguably improve our lot more than AI, at least per dollar invested?

What he means and you're interpreting a bit too literally is that this [heatshield] is one subsystem where the risks are not well understood or quantified as, say, the propulsion system, for which we have a lot more experience and flight heritage.

Bayes rule has existed for nearly 300 years, there is no excuse for ‘only look[ing] at this data’ and that is NEVER a reasonable thing to do.

You’d need a load of additional propellant to insert yourself into the same orbit as the ISS on your return, which would have an exponential effect on the amount of propellant needed in the first place to get all this lot out to the moon. It would be a different vehicle.

Everyone in silicon valley would do well to remember why the web was built (by other people, elsewhere).

Hopelessly over-idealistic premise. Sama and pg have never been anything other than opportunistic muck. This will be my last ever comment on HN.

I do worry about seeing more of these posts, as a way of SV people - who bare a substantial burden of guilt for enabling the collective mess we’re in because the ad-tech/algorithm dollars were nice - collectively distancing themselves from facing said guilt.

No idea is this particular person is especially part of the problem, I’m just talking about general vibes.

The electrical efficiency of breaking water to hydrogen, then combusting hydrogen through a turbine to generate electricity, compares extremely disfavourably to most other forms of storage - it takes about 50kWh of electricity to produce 1kg of Hydrogen, and if you propagate that back through a turbine and make some conservative assumptions about electrolyser costs and so on and so on, you're sort of approaching 5x the cost of others forms of renewable electricity to make electricity from stored hydrogen.

Of course, if we're building enough renewable capacity that electricity is basically free when it's sunny or windy, that changes the eceonomics and maybe we should all be making hydrogen in that sort of [bumpily]-abundant future.

However, storing hydrogen is also a pain - the density is crap even as a liquid and very technically challenging, and the density is mind-bogglingly crap as a gas - you'd want to find some vast geological underground reservoir in which to store it economically.

None of these are insurmountable, it's just not an especially attractive option as things stand.

Apple M3 Ultra 1 year ago

I'm from the dark ages and am interested in this for non-AI things like CFD. What is the state of SDK support for these chips? Is there a nice rust or C++ library that abstracts the hardware and lets you just do very big Matrix multiplications?

Things already are that bad, you’re standing an inch beyond where the previous wave washed up to and saying “well if it gets here in the next few waves then suddenly we become in deep shit’ while all around people are shouting at you about the existence of a concept called ‘the tide’. This is happening right now. fight back.

Not in any intrinsic way, it’s just a mildly better way of representing attitude if your state vector includes attitude.

They’re both quite basic building blocks in state estimation and as SV’s focus shifts from web apps to drone warfare it will only increase in frequency.