HNHacker News
TopNewBestAskShowJobs

axiom92

550 karma · joined February 8, 2015

https://madaan.github.io
submissionscomments
axiom92··on Physics Student Earns PhD at Age 89
PhD or not is one of those questions where if you find yourself asking it repeatedly (especially to others around you), the answer is no.

You are exactly the kind of person who should be doing it!

Good luck!

axiom92··on Artificial intelligence sheds light on how the brain processes language
>Maybe this is why NLU has split from NLP.

Sorry but can you be a bit more specific here? How exactly has NLU split from NLP?

axiom92··on Microsoft and Nvidia have created a 530B parameter language model
The model, named Megatron-Turing NLG 530B, is about 3x bigger than GPT-3.

The blog post doesn't provide a lot of numbers but looks like it beats the state-of-the-art in a couple of commonsense reasoning benchmarks.

Still, it shows that you can't just keep scaling the models and expect magic.

axiom92··on Can You See the Republic of Ireland from England?
Related: you can see Toronto skyline from Niagara falls https://www.blogto.com/city/2015/11/what_toronto_looks_like_...
axiom92··on “Don Quixote” as a manual for living
>The only people who praise this book are literature majors.

The sweeping generalizations do a great disservice to your otherwise good review.

I'm not a literature major, and I cannot recommend Don Quixote enough. My reason is simple: it was one of the funniest reads ever.

axiom92··on Use of artificial intelligence for image analysis in breast cancer screening
35 out of 36 will be scarier to most people. In behavioral psychology they call this "denominator neglect."

https://www.askattest.com/blog/guides/denominator-neglect

axiom92··on Machine learning won't solve natural language understanding
In case you were highlighting the person's h-index or citation count, there are a few shortcomings that these metrics suffer from [1], making it very hard to judge scientific impact.

EDIT:

Specifically, self-citation is the biggest issue, though often there are very good reasons to do it. For example, if you are working on a new area there might not be a lot of work other than yours. However, mindless self-citation grows the number of citation at O(N), as your n^th paper cites the (n - 1) papers before it. The total citation count grows by O(N^2)).

[1] https://www.nature.com/articles/d41586-019-02479-7

axiom92··on Machine learning won't solve natural language understanding
This person's h-index is 10x yours: https://scholar.google.com/citations?user=OHbcFkcAAAAJ&hl=en...!

Source: https://www.nature.com/articles/d41586-019-02479-7

axiom92··on Mammals dream about the world they are about to experience before they are born
Similar: https://www.cell.com/patterns/fulltext/S2666-3899(21)00064-7

Dreams help the brain in generalizing the world. From the article:

"That is, dreams are a biological mechanism for increasing generalizability via the creation of corrupted sensory inputs from stochastic activity across the hierarchy of neural structures."

axiom92··on Is Aging Inevitable?
>And glad the research suggests it.

The article is actually arguing against aging being inevitable (in the article, "My answer may be right or wrong" and below).

The author also links to a blog titled "Age reduction breakthrough"~[1].

[1] https://joshmitteldorf.scienceblog.com/2020/05/11/age-reduct...

axiom92··on Rumbling meteor lights up Norway; part of it may have landed near Oslo
https://www.reddit.com/r/nextfuckinglevel/comments/or9aqc/ro...

The post has been removed but the video is still there.

axiom92··on A YouTube chat about chess got flagged for hate speech
This is what GPT-3 (davinci, the 175B parameter model) generates for real:

* Prompt: ">White’s attack on black is brutal. White is stomping all over black’s defenses. The black king is gonna fall..." Fortunately there's an easy solution:

* GPT-3 Response:

Just don't play chess!

This is the only solution. The reason for this is that the "chess is bad for you" argument is a logical fallacy called "affirming the consequent". It goes like this: If A, then B B Therefore A In other words, if chess is bad for you, and you play chess, then you'll be bad for you. The fallacy can be seen by simply replacing "A" and "B" with two different statements, for example: The sky is blue Some birds fly Therefore, some birds are blue. See the difference? The argument against chess can be logically rephrased as "Chess is bad for you, and those who play chess will be bad for you. Therefore chess is bad for you."

axiom92··on The Benchmark Lottery
For a minute I thought someone scooped our SIGBOVIK paper: https://madaan.github.io/res/papers/sigbovik_real_lottery.pd...!
axiom92··on Using sqlite3 as a notekeeping document graph
Ummm I've been personally using Roam for over an year. The interface is neither slow nor clunky IMO (maybe Tiddly is faster, but I don't want to solve problems I don't have). I have a graph with > 1000 nodes that I use to organize my reading. Not sure where the hate is coming from.
axiom92··on Using sqlite3 as a notekeeping document graph
IMO https://roamresearch.com/ offers a more practical way of maintaining notes using graphs.
axiom92··on Stripe Identity
This is very cool!

Looks like they have been working on it for a few years now. Here's a video from 2019 where someone from Stripe is giving a demo: https://www.youtube.com/watch?v=TDocEZ4f5ow.

axiom92··on Keep Your Identity Small (2009)
Further down the article you will see:

>For example, the question of the relative merits of programming languages often degenerates into a religious war, because so many programmers identify as X programmers or Y programmers.

axiom92··on An Unbelievable Demo
Hah. Reminds me of this incident: https://newatlas.com/ai-art-auction-obvious-belamy/56984/

Someone downloaded the code for GANs (generative adversarial networks) from Github, generated (sampled) a painting using it, slapped the GAN objective function as a signature, and sold the painting for over $400k!

axiom92··on The Anxiety of Influencers
Hmm I see your point. It's a bit abstract but makes sense to me.
axiom92··on The Anxiety of Influencers
>I would argue you should make a new family.

Not sure what you mean by that? Is family always a good thing by definition in your books? (i.e., if X is bad, X is not a family). If so, then your deduction is correct but trivial, and overall the argument doesn't make much sense to me.

axiom92··on The Anxiety of Influencers
>What makes family bonds so strong is that they won’t turn your back on you when you are in a bind. They are much stronger than friends. Stable households are the most important predictor for childhood success.

This is blanket generalization. Such generalizations are (rightfully) hard to prove. It is likely that families are a positive force for _most_ people, but I am not sure if there is research that concludes that every single human would benefit from having a family/should put up with whatever family they have?

Going back to my original point, any "list" of N things cannot possibly fit all humans. Other people might have different opinions based on their own life experiences.

axiom92··on The Anxiety of Influencers
>Even analytic beliefs are things you believe because you experienced arriving at them?

What about scientific beliefs? Only a handful of people can really understand (or to align with your phrasing, _experience_ understanding) all the evidence there is to support the Big Bang theory. Yet, a much larger number accepts it.

And lemmas? I'm sure there are some lemmas I just use in proofs but have never proved from the first principles.

axiom92··on The Anxiety of Influencers
>These are the only things in life that truly matter.

Not really, such lists are typically a function of personal experiences. For example, plenty of people would happily take family out from the list.

axiom92··on An Introduction to Knowledge Graphs
Sure :)
axiom92··on An Introduction to Knowledge Graphs
Similar work (I'm the first author): generating temporal graphs from text using language models https://arxiv.org/pdf/2010.10077.pdf.
axiom92··on I’ve had the same supper for 10 years
Two different Wilf Davies (unless you were joking).
axiom92··on Online text to diagram tools (2020)
On a related note, I did some work on generating temporal graphs from text: https://arxiv.org/pdf/2010.10077.pdf

Code/models: https://github.com/madaan/temporal-graph-gen

axiom92··on How Doctors Die (2013)
Yeah it was 8 feet deep (I mentioned my height for reference).

I don't have any phobia of water. In fact, a week after the incident I signed up for more swimming lessons (8 1:1 sessions) and I believe I'm a pretty decent swimmer now.

I was actually planning to start training for the Alcatraz swim but moved out of the Bay before that could happen!

axiom92··on How Doctors Die (2013)
Back in March 2016, I took a couple of swimming lessons. Starting from scratch, I learned enough to be able to cross the pool using a desperate and random sequence of arm/leg strokes.

I thought I knew enough to jump into the swimming pool in my apartment complex. It was a quiet evening, and I was the only person in the swimming pool. Without thinking too much, I jumped into the pool, started executing the weird moves, and managed to reach the middle of the pool. It was at that time that the horror began.

Suddenly, I realized that I was in the middle of this 8' pool (I am 6' and some change) and that the water was pulling me in. I felt really helpless, and at that moment, it struck me that I might be drowning. This was it.

I was 23, fresh out of college with a decent job, had a whole bunch of plans, and it was all going to evaporate just like that. I didn't get any flashbacks of the life I've had so far, but I clearly remember thinking that my parents will be really sad and that the travel plans I had for the next month will never materialize. I really did think that I was going to die. I screamed out for help, but there wasn't anyone around (and I might not have been making much sound).

Fortunately, I didn't give up, and something told me to keep moving my arms and legs. I felt that I was finally moving ahead. I gave it everything I had and grabbed a corner of the pool. I came out and ran back to my apartment. All of this took about ~20 seconds or less. Still, I could have been dead (later that night, I looked up the statistics for the number of deaths per year due to drowning in a swimming pool, and it was a non-trivial number, ~4000/year in the US IIRC).

So anyway, that incident changed my perspective of life and death altogether. I care about my life a lot, and I don't want to die, but something changed that day. I developed an urgency to get to things on my to-do list, because I realized that the lights can go off rather unceremoniously and without much notice. I also like to think that I'm less scared of death now, but who knows, really.

I don't suggest having a near-death experience to lose your fear of death, but maybe try not to overthink things not in your hand in general. Some things are just too random to worry about.

axiom92··on Arabian Nights
On a related note, some languages become placeholder for "gibberish." For example, English speakers often say "it sounds Greek."

https://twitter.com/TommSciortino/status/1357376440170258434

← PreviousPage 3 of 7Next →