Hacker Newsnew | past | comments | ask | show | jobs | submit | more danblick's commentslogin

I'm skeptical. Machine learning algorithms are only as good as the training data you can provide them. There are lots of tools in genetic analysis you can use to try to understand function of some sequence (e.g. looking at conservation or homology). I don't see where deep learning would provide new value on top of hand-built statistical models that are already in use. (Geneticists have known about machine learning for a long time...)

Obligatory reference: https://xkcd.com/1831


What is LIME?


LIME: Local Interpretable Model-Agnostic Explanations; https://www.oreilly.com/learning/introduction-to-local-inter...

"“Why Should I Trust You?” Explaining the Predictions of Any Classifier": https://arxiv.org/pdf/1602.04938.pdf

https://homes.cs.washington.edu/~marcotcr/blog/lime/

https://github.com/marcotcr/lime

Anytime anyone makes snide HN comments like "oh you can't understand why neural networks make predictions" the correct response should always be "why doesn't LIME work in your specific case".

LIME is being used within the EU to explain credit decisions and fraud detection flagging on neural network based models, which is quite a high bar to regulatory oversight to pass.


Hadn't heard of that before, thanks!

In this case, I understood the question to be "will deep learning do a good job predicting the function of/phenotype emerging from individual SNPs" and I don't think model interpretation would help (for starters, the model is trained to predict linkage and doesn't deal with data related to phenotypes).


I believe that comment was edited.

Of course the NN won't interprete the results, it will just provide better results.


You're right.

If I had a game like this, I would want people to think they could make a lot of money by playing.

An easy way to create that impression and to generate hype would be to make some high-value self-trades.

If the game founders or early adopters didn't do this, they probably should have.

It's not proof they did but there's means/motive/opportunity.


I'm kind of imagining a future waterworld scenario where Earth's ice caps have melted completely and humanity looks back at electronic cat breeding as one of the gross excesses of our civilization that ultimately led to its destruction.

They're pretty cute though.


Okay. I'm not a zealot about these things, but I think the author underappreciates the nicer aspects of agile. My favorite book on agile looks at it from a mathematical / risk management / process model standpoint.

https://books.google.com/books/about/The_Principles_of_Produ...

The concepts used in the book - batch size, uncertainty, feedback, control, queue size and latency - these really are useful things to think about when scheduling work.


That said, it seems the author is making a larger point than the efficacy of any individual methodology: Convincing people to co-operate within any reasonable methodology/framework has the larger effect than effects of different methodologies. That having a religion/philosophy is more important than which religion/philosophy in terms of national co-operation, etcetera.


He freely admits he has never read anything about any of this or made a serious study of it, or had to implement any of it. It is more of a perspective or list of observations from life in the trenches than a serious study or even a very informed position. The whole list of methodologies even mentioned in the article is just three: waterfall, hacking, agile.


I second your recommendation. Any product manager or engineering manager who hasn't read that book is probably working on the basis of many flawed assumptions and optimizing for the wrong metrics.


I used to think in terms of stories, but then I heard Tyler Cowen tell me a story about the dangers of stories, and now I don't think in terms of stories so much. (My story of personal growth.)

https://www.ted.com/talks/tyler_cowen_be_suspicious_of_stori...

(PS: my comment is actually just a joke from the talk, but in any case I really enjoyed and talk and would recommend it.)


Obligatory cite: https://xkcd.com/552/


Yeah, "moving the web forward" sounds a lot like building barriers to entry. On the other hand, these browsers are open source software, so I'm not sure that's a great argument.


Scheme is a wonderful and flexible language. There's a reason it was used as a teaching language and in books like SICP. I don't use it very much, but you really have to admire some of the power in the language, like the way variable binding is handled to make macros so powerful. In short: scheme is probably one of the best base languages for building DSLs. If I had to pick "most tasteful" programming languages I would put scheme in the top 3. (Above Python, which is the language I use the most, and above C++ and Java which are other languages I use at work.)


Check out their investor presentation which compares their idea of AV ride-sharing with existing ride sharing services:

https://www.gm.com/content/dam/gm/events/docs/5265893-685163...


My understanding is that "driverless testing" here means either (1) moving the safety driver to the back seat or (2) following the vehicle with a second vehicle with a remote kill switch.

Thanks for posting that link though, I've been curious to read more about this.

"""At first, most of Waymo's driverless cars will have an employee in the back observing the vehicle's behavior. If something goes really wrong, they'll be able to push the "pull over" button to stop the car."""


Yes, driverless testing involves having a passenger function that where the passenger can request that the car pull over at the next available spot.

I'm not sure why you're splitting hairs on that though, when there's a definitive difference between having testing with someone in the driver's seat and not having someone in the driver's seat.


Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: