HNHacker News
TopNewBestAskShowJobs

anjc

1,414 karma · joined October 24, 2011

submissionscomments
anjc··on Survey of scientists sheds light on reproducibility crisis (2016)
Non-reproducibility doesn't imply shoddy work so it isn't fair to damage people's reputations for this reason. A paper may be non-reproducible because it isn't descriptive enough, because their data is not published yet, because a software library is deprecated, because a URL has changed etc, all of which are only discovered when somebody goes to reproduce it. Researchers can strive to make their work reproducible and still unwittingly fail.

If Github goes bust in 20 years you'll find a significant chunk of current work is non-reproducible despite author's best efforts.

anjc··on Machine learning is fundamentally conservative
> Search for a refrigerator or a pair of shoes and they will follow you around the web as machine learning systems “re-target” you while you move from place to place, even after you’ve bought the fridge or the shoes

> ...

> This is what makes machine learning so toxic.

I'm saddened to learn about our toxic contributions to society and can't wait to hear about alternative mind-reading approaches for fridge recommendation in the next article.

Words like 'conservative' and 'toxic' are misleading because they imply that there are better alternatives that are not being chosen. Far better are the terms by commenters here, 'descriptive' and so on. That the article is not written for machine learning practitioners makes it even more misleading.

anjc··on Categories of deep recommendation systems
At some point you presumably have to assess user-similarity in order to let similar users recommend items to each other? How's this different from e.g. user-user CF in terms of compiling everyone's info into a single model? I'm not criticising but rather just curious
anjc··on Categories of deep recommendation systems
What do you mean "human powered" rather than ML? How are the implicit/explicit judgements of relevance (e.g. likes, clicks, views, star ratings) that are used in rec algorithms any less human than, say, a recommendation from a single trusted expert?

Personally I think it's amazing that even basic recommendation algorithms can instantly select 10 mostly relevant items from corpora of millions of irrelevant items. I think people are too used to receiving excellent recommendations and do not realise what receiving bad recommendations would be like. Hybrid systems such as Spotify's are really amazing when you think about it.

anjc··on ACM signed letter opposing open access
I don't know what's meant to be A and B here. I already listed foreign aid as a counterexample in the post you're replying to.
anjc··on ACM signed letter opposing open access
Or by someone associated with them (in their lab, uni, co-authors as you mention etc). Essentially this means you must cite certain papers in some areas, as for larger conferences you're sure to be reviewed by such people.
anjc··on ACM signed letter opposing open access
I don't make this assumption. I simply think OPs argument that research should be open because tax payers have funded it is weak, for the reason I stated, i.e. it implies that non-tax payers should not have access, or conversely, implies that taxpayers in each country should get to decide how this research is distributed, possibly restricting access to non-nationals.

> The presumption you're making is that a country would never spend money for the public good, where that is defined to include the entire human race, rather than just nationals.

I think in general this is the case? Obviously governments budget for foreign aid, and also contribute to collective efforts e.g. climate change or (relevant to this discussion) EU funds such as Horizon 2020. But I don't think any governments do/should consider "the public" as meaning "the entire human race".

anjc··on ACM signed letter opposing open access
The argument seems to be that access to research should be unrestricted because "scientific publishers represent institutionalised theft of tax payer money", i.e. that theft of tax payer money is bad. In that case, why should people who live in tax havens, low-tax countries, or countries that don't adequately fund academia, have their access to research subsidised by people who live in high tax countries and who fund the research via tax? This seems like the same issue except at a higher level, i.e. sovereign theft instead of institutional?

I have no problem with open access by the way, I'm not advocating for restrictive publishing practices here.

anjc··on ACM signed letter opposing open access
> It would not be a problem if Arxiv was full of such articles, that would never cut it in an actual publication, but such articles are mixed with many that are eventually published and that just adds an amount of noise to the process that makes it hard to be sure what you're reading

I don't see this is a problem as, if you want to cite an arXiv paper, the responsibility on you increase, to ensure that the methodology etc is sound. This would be a better state of affairs than the "you must cite this highly cited paper just because it's a highly cited paper" situation we're in now.

anjc··on ACM signed letter opposing open access
If your issue is that tax payer money is being stolen then presumably open access should only be granted to citizens of a country, for research arising from academic institutions in that country? Otherwise countries will similarly be stealing tax payer money from one another. Actually, do public libraries get access to Elsevier etc?
anjc··on Fast-SRGAN: Deep learning model to convert low res pictures to high res
I looked through the GAN paper that this idea uses, and it seems that they generate training data by downsampling high res images. So you can presumably create large amounts of high quality training data for problems like this.

Do such methods introduce any problems in terms of e.g. overfitting (say, overfitting to the method that you use to degrade images)? Also, do fields tackling ML problems such as this where you can generate perfect training data tend to have faster improvements over problems where you can't (i.e. where you need to manually label training data, or where the target is ambiguous like sentiment analysis)?

anjc··on Undercover reporter reveals life in a Polish troll farm
> With all the propaganda going around, it is hard to find the truth

If bot/organized posts are properly classified as spam then surely the factual accuracy of someone's postings don't really matter anymore, as anything "inaccurate" falls into the realm of "opinion" or just human error. Presumably it's ok to be wrong.

So I don't see any problem with a person believing that e.g. Epstein was killed, and posting online about this belief. I don't see a problem with grassroots communities arising around this belief. It only seems to me to be an issue when such postings are sponsored/encouraged because of e.g. some political aim, and proper spam detection would mitigate this.

anjc··on I wasted 10 weeks of my life and $1900
I didn't add the "snark" to that snippet, and no, the rest is no better.

> If OP had done any research, he would have realised "Russian Reddit clone" was not a good idea with a real likelihood of success. Way too ambitious.

> If OP had planned, he wouldn't have stumbled around aimlessly for 10 weeks and would have understood success, especially when growing any type of "community" based project, takes time.

I've now pasted 60% of the comment, and none of it is useful or actionable and is all rather insulting..."stumbled around aimlessly", "not a good idea", "throwing shit at a wall", these are trash statements and no budding entrepreneur should listen to them.

I'll also add that OPs approach is a typical "move fast and break things" method of finding a product-market fit, and it seems to me that OP did stumble upon a product and a market in a discussion platform for cancer patients. They need a monetisation strategy, but for somebody to turn around and say this is "shit" and was "aimless" is useless and shouldn't be encouraged.

anjc··on I wasted 10 weeks of my life and $1900
This all goes without saying. However we're talking about developing and marketing a product in the first world. In this context, $1900 is not only nothing, it's a bargain given the experience OP gained.
anjc··on Tesla Q3 Financial Results
> I told my kids either the shares will pay for their college, or they may need some scholarships.

Probably not the best strategy to put essential money into a growth stock during an unprecedented bull run and expect the growth to continue for another 9 years. The likes of TSLA being at a higher price than today in 20-30 years are decent, but in 9 years? Very uncertain. See: every growth stock that hit a peak in 2000, and what happened in the following 9 years.

At the very least, it's probably not the best idea to tell your kids that this is what you're doing.

anjc··on I wasted 10 weeks of my life and $1900
> "I spent stupid money throwing shit at a non-stick wall, realized the wall was non-stick, so started spending money throwing shit at a different non-stick wall instead".

This is the worst possible interpretation of OP's venture, and must be written by a 12 year old if $1900 is "stupid money"

Edit: I'm not saying that $1900 is not a significant amount of money to many people in the world. But it is NOT a lot of money in the context of developing/marketing a tech product in the first world.

anjc··on I wasted 10 weeks of my life and $1900
> I wasted 10 weeks of my life and $1900

There are people spending more to learn less. Seems like good value to me. Best of luck with your next venture, if you're reading.

anjc··on Linear Algebra and Applications: An Inquiry-Based Approach
I haven't gone through this yet but I really like the idea of each new concept being described in the context of a useful application. Thanks OP
anjc··on An Inside Look at Microsoft’s Newest Flight Simulator
I find the touch interface in XPlane to be fine. I can hold one Oculus Touch controller and also use that hand to hold a hardware yoke and throttle. But I would be very happy if something could track my fingers and I had haptic pads on my finger tips.

Definitely agree on trying to use Touch controllers to move the yoke...frustrating.

anjc··on An Inside Look at Microsoft’s Newest Flight Simulator
I use a CV1 with XPlane and...OK the instruments are pixellated from a fully seated position, but lean forward a few inches and they're legible. Any such issues are fully compensated by the sense of height you gain and positional awareness. I could never go back to a monitor setup for flight sim.

If my CV1 is fine, I'd imagine that the current headsets are more than adequate for reading instruments.

anjc··on An Inside Look at Microsoft’s Newest Flight Simulator
Agree, but I suppose I meant more in terms of optimizing performance by making use of idle 2080TI RT cores. Volumetric cloud lighting as OP said, cabin reflections, water perhaps..

I'll take anything that would mean the community doesn't need to keep modifying .CFG parameters to get an extra few FPS

anjc··on An Inside Look at Microsoft’s Newest Flight Simulator
This looks amazing, and I'd love to switch back to MS after having to switch to XPlane. However: no VR no buy.

No matter how many improvements they've made, VR + 13 year old FSX will be much more immersive and useful. It's insane that any sim developer would not even have VR on their post-launch road map. If running their engine at a sufficient framerate is an issue for VR, then it's insane that they didn't consider this from the ground up. Even 25 years ago VR was a consideration for flight sim developers (anyone remember EF2000 and the VFX1?).

anjc··on An Inside Look at Microsoft’s Newest Flight Simulator
They said that they aren't using raytracing I believe. That'd be great if they did, I don't think there are any killer apps for a 2080ti yet, but this would certainly be one.
anjc··on An apology to our community, and next steps
> First of all, we hurt members of our LGBTQ+ community when they felt they couldn’t participate authentically

> thank you for acknowledging that. I will be able to sleep in peace tonight knowing that I have still a place here

If users were genuinely unable to sleep because one random faceless person asked if they could continue their practice of only using gender neutral pronouns, then this seems to be a situation in which you shouldn't apologise to them, because they'll never be satisfied. It must be a nightmare being assigned to a CoC project now.

anjc··on Joi Ito Resigns from M.I.T. Media Lab After Outcry over Jeffrey Epstein Ties
I'm with you. Posters here seem to be making several logical leaps due to hindsight bias. His previous convictions do not imply that he's a human trafficker, and if he was, he almost certainly would not be advertising it in this manner.
anjc··on She Wanted a Man with a Good Job Who Is Nice to Animals
Surprisingly accurate summary.
anjc··on Are We Making Much Progress? Analysis of Recent Neural Recommendation Approaches
There are fields (recommender systems in particular in my opinion) where you will not get published if you report a strong baseline and your approach does not outperform it, but you may get published if you use weak baselines and your approach outperforms it. This doesn't make sense to me as surely it's possible for promising new approaches to not yet outperform current state of the art methods.

These peer review aberrations incentivise these bad research practices I think.

anjc··on King – Man + Woman = King?
Yeah I can't disagree with this. I think it's typical to present the highlights of your work though. In this case, the issue seems to be "why do the authors consider 'black:killer' to be a highlight"? It might be racist but it's not academically fraudulent (a strong statement) in my view.
anjc··on King – Man + Woman = King?
They're reporting real findings which they allow everyone to verify. There's nothing academically fraudulent about this.
anjc··on Jony Ive is departing Apple, but he started leaving years ago
Well I appreciated the word play
← PreviousPage 7 of 33Next →