Hacker News new | past | comments | ask | show | jobs | submit | mrdrozdov's comments login

This paper has been online at least since 2018.

It has since been published: https://proceedings.mlr.press/v98/stelmakh19a.html


No.


This might provide some guidance: http://gltr.io/


This and related techniques are trivially foolable by fine-tuning the model.

They're also trivially foolable by using sampling techniques or settings which encourage the model to generate rare words a lot.

Also foolable with filter-assisted decoding: https://paperswithcode.com/paper/most-language-models-can-be...


This is such a click-bait title and has nothing to do with scalability.


I don't believe this (yet) and here's why.

The claim is that Google search is producing worse results than in the past. The analysis is mostly anecdotal, and similar claims have been made before in a more concrete way. A prime example is "time to cook onions" giving incorrect results, covered in this slate article: https://slate.com/human-interest/2012/05/how-to-cook-onions-...

What we need is to see is specific queries, the results returned, why they're wrong, and what they should be instead.


Some folks should post comparisons between Google and other sites such as:

* https://bing.com

* https://duckduckgo.com

* https://you.com

Maybe use the categories mentioned in the twitter thread such as health, travel, recipes, product reviews, etc.


This paper the article refers to is fantastic! I think it's a work most in ML research should become familiar with. And if you believe in the power of benchmarks and data, then this holds even more true. Investing in diversity in datasets is likely an impactful way to make progress in AI/ML.

Minor typo in this article...

ARTICLE: Among their findings – based on core data from the Facebook-led community project Papers With Code (PWC) – the authors contend that ‘widely-used datasets are introduced by only a handful of elite institutions’, and that this ‘consolidation’ has increased to 80% in recent years.

...but right after, they quote the paper and clearly it is 50% not 80%. See the quote from the paper:

PAPER: ‘[We] find that there is increasing inequality in dataset usage globally, and that more than 50% of all dataset usages in our sample of 43,140 corresponded to datasets introduced by twelve elite, primarily Western, institutions.’

...and the article is leaving out this relevant quote from the paper:

PAPER: Moreover, this concentration on elite institutions as measured through Gini has increased to over 0.80 in recent years (Figure 3 right red). This trend is also observed in Gini concentration on datasets in PWC more generally (Figure 3 right black).

...and in general the article is right that inequality is increasing over time, but Gini is a specific metric to measure inequality, and 0.80 is not the same as 80% inequality.


They’re more written than you know. If you put similar effort into improving social skills, you will improve. Start with this book: https://www.goodreads.com/book/show/4865.How_to_Win_Friends_...


Ocean's Eleven bitcoin edition...


You can't mine Bitcoin on these cards.


Challenge accepted --- BTC mined on a gameboy: https://www.laptopmag.com/news/you-can-mine-bitcoin-on-a-nin...

Also, you're assuming whoever stole these cards are rational beings and not characters out of Ocean's Eleven.


or how about mining on a xerox alto? http://www.righto.com/2017/07/bitcoin-mining-on-vintage-xero...

They are cheating by putting it near to a known hash to verify that it works, you can't earn anything off it.


Why?


Because Bitcoin is mined on ASICs. It's not a graphics card, instead a single chip optimized for one function only: mining (by generating hashes).

GPU based mining is mostly done to mine Ether. Ether will switch to Proof of Stake somewhere in the next 1-2 years, after which the competition between gamers and miners for GPUs will end.


That's the optimistic outcome, the pessimistic outcome is the miners will fork "eth-classic" and keep going. It's up to the reader to decide which is most likely, or both.


Or they will move to some other coin that is rising in popularity and doesn't have GPU resistance built in. The latter is something that feels like a negative these days if you want a lot of buzz about some new crypto tokens...


My understanding is that miners are on board with the new Proof of Stake system of ETH 2. Holders even more, as it will reward ETH to ETH holders.


You could mine bitcoin on them, but it would be so inefficient compared to ASICs which can hash many magnitudes faster, you would have an astronomically small chance of mining a block.


I don’t know. I watch a similar amount of movies each year, and I still enjoy it. If anyone is looking for some more obscure recommendations, can check out The Dreamers, and Stilyagi.

EDIT: I have to add that I reference movies a lot in conversation. Often, I’ll watch a movie then immediately call a family or friend to discuss some finer point. This happens frequently, sometimes for a fairly mundane movie detail.

EDIT2: Now I really want to make a list of movies just from this year, since my number has definitely gone up since COVID. I think I’d easily break 100 in 2021 alone.

EDIT3: Here’s a list from my Netflix history since June 1. Mix of TV and movies. I added Justice League Extended Edition and Replica even though they’re HBO because I watched them recently (within the last week). This isn’t really a representative list of my watching, plus I tend to watch a bunch of similar movies/shows, then switch to a new cluster. This group is particularly action heavy because I was playing a lot in the background recently while doing other work. All of these were fun! Even if I don’t think they are the best ever :))

  Movies 2021 June-July

  Zack Snyder’s Justice League
  Replica
  The Take
  Darc
  American Assassin
  S.W.A.T.
  Sniper Legacy
  The Interpreter
  Redemption
  Extraction
  Spenser Confidential

  TV

  Biohackers
  Shooter
  Quantico
  Sweet Tooth
  Record of Ragnorak
  Bodyguard
  Hollywood


I absolutely love SWAT. It feels like a movie they had fun making, and doesn't take itself very seriously.


Adding a few more from HBO and Amazon...

  Greenland
  The Little Things
  The Conjuring
  Killerman
  Ghost in the Shell (live action)
  Redline
  Sputnik


I think the core argument has much more to do about plagiarism than learning.

Sure, if I use some code as inspiration for solving a problem at work, that seems fine.

But if I copy verbatim some licensed code then put it in my commercial product, that's the issue.

It's a lot easier to imagine for other applications like generating music. If I trained a music model on publicly available Youtube music videos, then my model generates music identical to Interstellar Love by The Avalanches and I use the "generated" music in my product, that's clearly a use that is against the intent of the law.


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

Search: