onuralp

Low-hanging in the following sense: the premise - I tell you what I like, you tell me what else I might like - is compelling for the consumer, therefore, input is not the limiting factor. Moreover, recommender systems at such massive companies must be particularly amenable to A/B testing to help profile their customers more accurately. Finally, performance metrics are straightforward and easy to calculate.

Code phonology (vocalization of code) might be a good place to start: https://news.ycombinator.com/item?id=16554865

Two comments:

- Lay people seem to have a very good intuition about the actual heritability of intelligence (taken to assess the genetic and environmental determinants of intelligence). [0]

- If you'd like to read an up-to-date and thorough literature review on human intelligence, I highly recommend "Intelligence: All That Matters" by Stuart Ritchie.

[0] https://news.ycombinator.com/item?id=16468694

Better to link the final version - http://sci-hub.la/10.1126/science.aam9309

She is great. Check out MIA seminar series [0] featuring like-minded researchers.

[0] - https://news.ycombinator.com/item?id=14647070

They specifically cite that JAMA analysis among others, present the prevailing scientific theory in dietary science and heart disease epidemiology at the time, and challenge their narrative.

I suppose what I had in mind was akin to the following practical example, say a technical question asked on Stack Overflow. When the submitter can't reliably or efficiently anonymize the accompanying data [which would risk reveal the submitter's anonymity and compromise data], the submitter does not have any choice but anonymizes both the identity _and_ the data, which probably results in a sub-optimal exchange.

Imagine that you can effectively anonymize the dataset, I think one can think of circumstances where this could lead to a more effective communication as to provide further information (e.g., submitter's domain familiarity, language experience, resource constraints) and context (e.g., is this a standard practice) instead of fishing for more clarification in an iterative, and therefore less efficient, fashion.

Shared environment is probably not created equal though. While the family environment is constant, parents are likely to vary (deliberately and/or not) in their resource allocation (e.g., attention and care) among their kids. An interesting recent paper shows that the non-inherited genetic material of the parents may _still_ have an influence on the kid. [0] What exactly accounts for that influence remains to be elusive, and it would be quiet astonishing if the majority of the influence can be explained by well-established (or rather, exhaustively studied) behavior patterns in psychology. For example: birth order, number of same-sex siblings etc.

[0] https://www.biorxiv.org/content/early/2017/11/14/219261

Have you checked out scikit-allel? It is fairly comprehensive in terms of calculating basic population stats, and the developer is highly active.

scikit-allel: http://scikit-allel.readthedocs.io/en/latest/index.html

scikit-allel example: http://alimanfoo.github.io/2015/09/21/estimating-fst.html

zarr: https://github.com/zarr-developers/zarr

While I think the abstract alone is succinct, the authors present their theory a bit more in detail on their website: https://sites.google.com/site/hugomercier/theargumentativeth...

To be fair, I suspect that most of such comments are genuinely interested in pointing out the biases that might potentially (or allegedly) underlie what is perceived to be a fallacy. [0]

[0] https://plato.stanford.edu/entries/fallacies/#Bia

It's listed under 'bad reasons fallacy'.

Well, when I read this on the news I was fascinated by the extreme growth (>4 standard deviation compared to growth experienced by an average astronaut), however, it turns out that there was a measurement error.[0]

[0] http://www.bbc.com/news/world-asia-42630473

I recommend Peter Attia's highly engaging and informative talk [0] on the same topic, which I think does a better job at articulating the motivation behind possible approaches to attacking the issue of health- and life-span.

[0] https://www.youtube.com/watch?v=vDFxdkck354

You might also be interested to check out Automatic Statistician (developed by Z. Ghahramani, J. Tenenbaum and others) - https://www.automaticstatistician.com/index/

Awesome, I'll check it out - thanks!

Hi Craig,

This looks very interesting. You might be the right person to ask about something related that I am currently working on: do you know of any app that would extract keyword / name based parts of audio? For example, extract only the parts where Elon Musk speaks given audio input (podcast, YouTube etc.)? Alternatively, extract only the parts (-30 and +30 seconds) when a specific word is mentioned.

Thanks!

According to this study[0] on the diagnostic (and triage) accuracy of online symptom checkers (including WebMD among others), the reliability is a coin toss.

[0] http://www.bmj.com/content/351/bmj.h3480/

This reminds me of Ferenc's reminder that 'deep learning is easy' - 'http://www.inference.vc/deep-learning-is-easy/

These look fascinating, thanks.

I tried to crack this problem, and ended up (very naively) resorting to substituting division with modulo operation that is, a%(b+c) + b%(a+c) + c%(a+b) = 4 of which the min solution is a=1, b=2, and c=4. I am glad that the true solution involved EC which, if my understanding is correct, is basically modular operations in high-dimensional space.

Exploration versus Exploitation: Deciding When to Get Married - http://www.inference.vc/my_big_fat_data_driven_wedding_how_t...

Another self-quantified in a similar spirit - machine learning and ketosis: https://news.ycombinator.com/item?id=12279415

Not sure if the personal interview model can be scaled as to include an appealing variety of startups. You may want to take a look at these two sites[0:2] that I think do a pretty good job at aggregating the stories of the startups that did not quiet take off.

[0] Collapsed - https://collapsed.co/ [1] Autopsy - http://autopsy.io/

I think Mark Nelson had succinctly described the problem with OpenAI (more broadly, with the "AI safety" concerns): https://twitter.com/mjntendency/statuses/859189428748791810

Disclaimer: I have an engineering background.

I think this is a fair characterization of the two approaches.

I am currently taking a class by Strang co-taught with Alan Edelman (MIT/Julia) and Raj Rao (Michigan) that has a strong emphasis on applications and hands-on coding tasks (using Julia).[0] I am also making my way through CtM (thoroughly enjoying) and hope that they will release the video lectures soon as I think the lectures and CtM complement each other quiet nicely.

[0] Matrix Methods In Data Analysis, Signal Processing, And Machine Learning - https://stellar.mit.edu/S/course/18/sp17/18.065/

I think it would be quiet interesting to crowdsource a math-to-english interpreter (a la 'explain shell' [0]) for math equations.

[0] https://explainshell.com/

A Primer on Bézier Curves (with interactive animation) - https://news.ycombinator.com/item?id=14191577

You might be interested in this - "computational reproducibility using Continuous Integration to produce verifiable end-to-end runs of scientific analysis": https://greenelab.github.io/continuous_analysis/

I think it has to do with public key cryptography - in particular, elliptic curve cryptography[0], which was mentioned by Schneider to be one of the key topics on NSA's agenda[1].

[0] https://arstechnica.com/security/2013/10/a-relatively-easy-t... [1] https://www.schneier.com/blog/archives/2015/10/why_is_the_ns...

Bear App. http://www.bear-writer.com/