jedharris

12 ms·

91.
by jedharris 8y ago
Google, Apple, etc. tried to make symbolic AI work for their problems but found it had intractable problems. Google fairly publicly transitioned to a deep learning approach for translation which greatly improved their results. Apple is st
92.
by jedharris 8y ago
There are plenty of domains where symbolic AI has had every opportunity to use unlimited computing power, plus decades of design and experimentation, but has lost to deep learning. a good area for examples is human games (e.g. chess, Go, At
93.
by jedharris 8y ago
Great model! Simple enough to understand, fits a lot of recent phenomena. Scary implications, but we should be able to respond effectively. Unexpected but good to see Schneier doing this kind of work.
94.
by jedharris 8y ago
Is it easier to break a tank than operate one? Then why a robot?
95.
by jedharris 8y ago
Agree about type inference being a great tool along the lines we need. Notably it extracts information from the code and lifts it to the type level, needing only occasional guidance when it can't figure enough out by itself. We don&#x
96.
by jedharris 8y ago
See related HN thread < https://news.ycombinator.com/item?id=18050706>
97.
by jedharris 8y ago
Thanks for this! Hoare (partially) answers a question I asked in the comments on "The Little Typer" (a good treatment of programming with dependent types). That approach to dependent types brings our code closer to being verifiab
98.
by jedharris 8y ago
Again lots of interesting points. Only time for one response. You highlight networking but I also mentioned databases. Schematizing your comment "Our... systems "mostly work" because... the difference between a perfectly coo
99.
by jedharris 8y ago
Several interesting points but maybe they show something different from what you intend. Network stacks, databases etc. that correctly handle trillions of interactions, some of them adversarial, have indeed been "refined around [their
100.
by jedharris 8y ago
Several comments elaborate on the big gap between normal programming practice (e.g. structurally recursive algorithms) and the great difficulty of dependent typing those practices. Some of these comment on how dependent types are "just
101.
by jedharris 8y ago
Excellent point. What mathematicians write to each other are arguments not proofs. Machine checkable proofs are really proofs. This should not be embarrassing to mathematicians. It opens up potential for a lot of progress in several dire
102.
by jedharris 8y ago
If you are right then calling them "proofs" is misleading. The results aren't proven, just made convincing. So I guess they are "arguments" -- only detailed enough to convince others who already know enough about
103.
by jedharris 8y ago
Maybe an independent group could manage the schedule and have lab tests done so the drug details could be determined. Then the academic researchers could track the subjects, analyze the results and publish.
104.
by jedharris 8y ago
Yes but... the level you are at is not just or mainly a function of IQ. I cannot evaluate the quality of a structural engineer who is much more skilled than me. Their cool new structural ideas could be brilliant or just terrible; I can&#x
105.
by jedharris 8y ago
You definitely can't trust people's self-evaluation either way. Lots of really skilled people are self-critical. "Reading people" of course helps but it is nowhere near enough. But you don't have to become an exper
106.
by jedharris 8y ago
I think people can recognize those at level N+1, but not (reliably) level N+2 and beyond. In fact they are likely to think that people much above level N+1 are "crazy". Current political discourse contains lots of examples of peo
107.
by jedharris 8y ago
Your last phrase is key: They really didn't want the experts, they just wanted the results. No one can hire good people if they think they already know better. Good people got good by picking good learning situations, and avoiding ta
108.
by jedharris 8y ago
To hire someone with expertise to solve a problem in a domain where you aren't an expert: You find people who claim to understand the domain, and then discuss the problem with them in depth. Your goal is not to evaluate them in gener
109.
by jedharris 8y ago
This is theory, empirically the data says otherwise. in 2007 there was a huge amount of evidence in plain sight that the housing market was a bubble, but the market didn't adjust to a rational level. See "The Big Short". Th
110.
by jedharris 9y ago
Yes indeed. CRISPR is a big shiny red button. However I don't think there's any better option than pursuing enlightenment values to address this problem. Complaining that we can't always get that right (as Gray does) is not
111.
by jedharris 9y ago
Gray in his review doesn't establish that enlightenment thinkers had no common values. He only points out that some of them had wrong conclusions. The leap from that to implying that they lacked any common values is entirely illegiti
112.
by jedharris 9y ago
Thanks for that way of putting it. It articulates my feeling when I couldn't do it for myself. ALSO note that what you propose is an "enlightenment compliant" way of proceeding. Unfortunately much argument (especially in the
113.
by jedharris 9y ago
I'd like to see evidence that Pinker thinks it is about "this special group of people". What I have read in the book doesn't support this. He of course gives examples of the most remarkable people but goes on to say th
114.
by jedharris 9y ago
IANA historian, but do read contemporary scholarly history. Chris Wickham whom everyone seems to respect thinks that population in most regions of the Western Empire fell by 50% or more after 300 AD. Setting aside arguments about what the
115.
by jedharris 9y ago
What do you recommend? Without a criterion for trusting some previous work this would make science impossible.
116.
by jedharris 9y ago
PG&E here in Northern California does offer net metering -- sell to them when the sun is up and buy back at night, guaranteed a wash in $$. Good deal for us.
117.
by jedharris 9y ago
Grounding common sense and philosophical ideas of space and time in current physical theory. Most of the discussion makes sense across a broad range of theories, it is not limited to just loop quantum gravity. Impressively readable.
118.
Space and Time in Loop Quantum Gravity(philsci-archive.pitt.edu)
2 points by jedharris 9y ago|1 comments
119.
by jedharris 9y ago
Amazon (and Walmart before them) do a lot to create this situation. Walmart notoriously kills local businesses when they move into a town -- not by targeting them directly but by competing them out of business. If the alternative employme
120.
by jedharris 9y ago
The comparison with the brother's job at the Stater Brothers warehouse across the street is pretty clear about this.

More ›