Y
HN Search
Hacker News Search
new
|
comments
|
top
|
jobs
davmre
searching PlanetScale…
1.
▲
2.
▲
3.
▲
4.
▲
5.
▲
6.
▲
9 ms
·
31.
▲
Theano, TensorFlow and the Future of PyMC
(medium.com)
4 points
by
davmre
8y ago
|
0 comments
32.
▲
by
davmre
8y ago
What do you think about government-financed prizes for private-sector drug development, e.g., the government pledges $X billion if you develop an Alzheimer's cure and place it in the public domain? Bernie Sanders has a proposal like th
33.
▲
by
davmre
9y ago
'Textbook' communism involves the government directly managing the means of production. Owning a minority stake in many corporations that are run for profit by incentivized managers in a market economy has, perhaps, some similarit
34.
▲
by
davmre
9y ago
One possible mechanism is a social wealth fund that would gradually come to own (and distribute the proceeds from) a nontrivial portion of the national capital. Matt Bruenig wrote a NYTimes oped about this idea: https://www.nytim
35.
▲
by
davmre
9y ago
I haven't seen anyone here argue that it's difficult to invest foreign money in the US. Apple's only complaint is that they couldn't funnel the profits from such investment back to their US shareholders, without paying t
36.
▲
by
davmre
9y ago
Tensorflow's advantage is that once you build your model, you can run on everything from a massive cluster to a mobile GPU without significant modification. Because you're just writing a description of a computation graph, it'
37.
▲
Why Ruth Porat Should Be Uber’s Next CEO
(fortune.com)
3 points
by
davmre
9y ago
|
0 comments
38.
▲
by
davmre
10y ago
There are always going to be more important causes, under any particular person's view of "more important", which depends very strongly on both subjective values and (in the case of AI alignment work) on precise probabilities
39.
▲
Opening a new chapter of my work in AI
(medium.com)
247 points
by
davmre
10y ago
|
74 comments
40.
▲
by
davmre
10y ago
I don't see where this argument is coming from. There's no reason why self-driving cars can't be as or more capable at interacting with cyclists than human-driven cars. Even if issues with self-driving cars did require new re
41.
▲
by
davmre
10y ago
You're welcome! FWIW I mostly agree with argonaut's point elsewhere in this thread - very few people successfully self-teach ML from a textbook alone. So whichever book(s) you choose, it might also be worth working through some co
42.
▲
by
davmre
10y ago
I haven't read it in detail, but my impression is that it is mathy, like Bishop, but focuses more on 'classical' frequentist analysis, whereas Bishop takes a more open-ended Bayesian perspective and covers important machinery
43.
▲
by
davmre
10y ago
You can order the standard machine learning texts from most to least math-y, and least to most modern: - Pattern Recognition and Machine Learning (Bishop 2007) - Machine Learning: A Probabilistic Perspective (Murphy 2012) - Deep Learning (G
44.
▲
by
davmre
10y ago
There are plenty of domains where the objective is quite well defined. Video games are excellent examples. An agent that achieved superhuman performance on, say, Starcraft, would have a very impressive suite of capabilities involving some s
45.
▲
by
davmre
10y ago
If the inferential question you're interested in is, "given data X, what do I conclude about underlying cause/variable/parameter T?", then you are a Bayesian, like it or not.* Sure you can define a likelihood functi
46.
▲
by
davmre
10y ago
It's almost as though the Times publishes pieces by multiple contributors representing a diverse range of perspectives.
47.
▲
by
davmre
10y ago
I think it's important to view this piece in the context of the recent election, where the median Trump supporter had a higher income than the median Clinton supporter. Yes, rural poverty is a thing and we should provide compassion and
48.
▲
by
davmre
10y ago
Assuming that AI-generated wealth is accruing to some person or people, you can just tax it as ordinary income. The main difference is quantitative: in a world where all wealth is generated by AIs controlled by a small number of people, you
49.
▲
by
davmre
10y ago
What predictions does IIT make that could be 'de facto' falsified in the sense you're imagining?
50.
▲
by
davmre
10y ago
Have you read Scott Aaronson's "Why I Am Not An Integrated Information Theorist" ( http://www.scottaaronson.com/blog/?p=1799 )? He commends IIT for making falsifiable predictions, but argues that some of t
51.
▲
by
davmre
10y ago
It's not clear that Knuth would be a particularly knowledgeable leader on these issues either, whatever his other strengths.
52.
▲
by
davmre
10y ago
Pieter Abbeel has been working towards household robots for quite a while; given his involvement and several of his students at OpenAI it's not surprising they'd be thinking along similar lines. The issues you mention are real, bu
53.
▲
by
davmre
10y ago
Sounds like a sweet deal. You should found a company!
54.
▲
by
davmre
10y ago
Because those are groups to whom the startup ecosystem has not historically been open and available. The path towards "open for all" is to identify the current failure cases, and direct focused effort towards fixing them.
55.
▲
by
davmre
10y ago
You've described why it's almost impossible to do anything about housing policy at the local level. That's why state-level action such as this bill is so important. The state legislature represents a much broader constituency
56.
▲
by
davmre
10y ago
Not shut out completely, but the government is pushing back: http://www.nytimes.com/2016/04/22/technology/apple-no-longer...
57.
▲
by
davmre
10y ago
> Firstly, the capital cost of the older schools may well have been paid off. In the long term nothing is "paid off"; you have to keep up a regular schedule of renovations and renewal projects, or things fall apart. Which is wh
58.
▲
by
davmre
10y ago
Again that's a wash, though, as long as interest rates on new bonds are not significantly higher than those the city is already paying on its existing infrastructure bonds (or on future bonds that will be sold when necessary to repair&
59.
▲
by
davmre
10y ago
> Some of those costs can maybe be amortized via debt issue (bonds), but the cost of servicing those would still be borne by the community overall and not necessarily the newcomers' taxes alone. This seems like another way of saying
60.
▲
by
davmre
10y ago
Why would you need to raise taxes to pay off bonds? Why isn't the increased revenue from the new residents' taxes sufficient? If your population doubles, sure you need to sell enough bonds to double your infrastructure, so the ann
More ›