ma2rten

42 ms·

391.
by ma2rten 6y ago
That would create the wrong incentives.
392.
by ma2rten 6y ago
This is true, but they can also splinter off or be online versions of of real world communities. For example Facebook started from collages.
393.
by ma2rten 6y ago
There is also a large NSFW scene on reddit.
394.
by ma2rten 6y ago
The resulting model is shared, so everyone who gives their compute time will benefit from it. Neural networks automatically error correct so they will be robust to some amount of corruption.
395.
by ma2rten 6y ago
Lots of them: https://github.com/elyase/awesome-gpt3
396.
by ma2rten 6y ago
I wish you could hide stories by keyword or by domain.
397.
by ma2rten 6y ago
Engineers tend to overestimate how difficult machine learning is. That is exactly how a good data scientist would solve this problem. If (and only if) this initial solution is not sufficient then you can iterate on it (maybe we should also
398.
by ma2rten 6y ago
Mercurial is much more user friendly. Also big companies such as Facebook and Google use it because it's more extensible.
399.
by ma2rten 6y ago
Do you really? I've been using as my primary editor for last 10 years or so, I rarely make use of the grammar. I sounds fancy and powerful in theory but I don't end up using it that often (even though I'm aware of it). That s
400.
by ma2rten 6y ago
There is a bit a grey area, which is why I was careful say "the model learned/understood/knowns something" in my earlier comments. However, You seem to be making the Chinese Room argument. If you define meaning such that
401.
by ma2rten 6y ago
I'm using "understood" as "can apply in an unseen context". For example if you a human student understood a math concept it means they were able to use it in a new context and have not just memorized how to solve a
402.
by ma2rten 6y ago
You can look at OP's profile.
403.
by ma2rten 6y ago
Don't use Electon. Your users WILL notice and complain about performance. How will people notice if you are using Electon, but not if you are doing this? Is Electron doing something that makes it unnecessarily slow or are you saying
404.
by ma2rten 6y ago
I'm using "understood" metaphorically similarly to how people say machine learning models "learn" something when they really just minimize a loss.
405.
by ma2rten 6y ago
Maybe not, but the fact that it actually made a sentence with a word in it that it could not have possibly seen in the training data tells us that it understood something about the meaning of the instructions.
406.
by ma2rten 6y ago
Right, in this context I define meaning as the relationship between symbols. Form would be how to structure a sentence (but that sentence wouldn't have to make sense). If the parent comment was making some kind of Chinese Room argument
407.
by ma2rten 6y ago
The GPT-3 paper goes into this. They consulted experts: The assessment was that language models may not be worth investing significant resources in [by state actors] because there has been no convincing demonstration that current language
408.
by ma2rten 6y ago
I'm not sure what your definition of meaning is. Take this example from the GPT3 paper: To do a "farduddle" means to jump up and down really fast. An example of a sentence that uses the word farduddle is: One day when I was
409.
by ma2rten 6y ago
GPT-3 doesn't care about anything except predicting the next token. It learned something about structure and meaning in the process.
410.
by ma2rten 6y ago
HN is not a typical user base.
411.
by ma2rten 6y ago
Didn't Spain try to do the same thing with Google News, which just ended up with them shutting down the product in Spain? https://support.google.com/news/publisher-center/answer/9609...
412.
by ma2rten 6y ago
Google requires its employees to use a security key for access to all internal systems including admin tools, source code and email. Every since google started enforcing this policy the number of successful phishing attacks has gone down to
413.
by ma2rten 6y ago
Their liberal persona is actually fairly mainstream, probably to the right of the average Bernie Sanders supporter.
414.
by ma2rten 6y ago
I wouldn't worry about it. If we can automate programmers, we can also automate most other jobs.
415.
by ma2rten 6y ago
I am actually wondering if search engines really need such a large index. The vast majority of sites in Google's index are crap. If you were able to better select which sites to index it might help search quality and improve efficiency
416.
by ma2rten 6y ago
Maybe if enough people do the same, reddit will start attracting the same kind of SEO spam. It's almost like Google is the victim of its own success. If everyone try to game your system some will success.
417.
by ma2rten 6y ago
Yeah, I take that back. DeepL is better (again), although it is highly dependent on which languages you translate and which context. Quite impressive, especially that DeepL (and every other machine translation application) makes use of rese
418.
by ma2rten 6y ago
Are you sure that set out to proof one-shot works? Maybe they found fine-tuning performance disappointing and decided to publish this instead.
419.
by ma2rten 6y ago
Are you sure about that? I think DeepL was ahead for a brief time after it was released, but Google Translate has caught up.
420.
by ma2rten 6y ago
Previous Discussion: https://news.ycombinator.com/item?id=20896156

More ›