tensor

40 ms·

751.
by tensor 3y ago
I think there is a big difference in terms of data recovery though. You can't take a compression algorithm, for example, and claim that its "just some statistical analysis" when it can reproduce the original perfectly. Heck,
752.
by tensor 3y ago
This is a bit strange as presumably interest rates were raised in the first place to prevent inflation. Turning around and scolding the banks for doing what the government wanted in the first place seems counter productive. Wouldn't re
753.
by tensor 3y ago
Unfortunately the old American disease of not ensuring competition and instead passing regulation ensuring monopolies stay in power means the "problem" is not actual solved, and is in fact made worse.
754.
by tensor 3y ago
Despite the language, I have a very hard time believing that Zoom is actually trying to train a generative AI on private video calls. If I had to guess this is there so that they can train adjacent services such as malware scanners for shar
755.
by tensor 3y ago
Then choose a license like AGPL and no corporation will touch your code. If you choose an Apache 2 license or similar, don't be shocked when people use it as the terms intend. Personally, I will continue to release any code I open sour
756.
by tensor 3y ago
People who have lots of children, and people who are prosperous, currently do not care about the future. In fact, it seems the wealthier people are the LESS they care about the future for some reason. If you ask me, people need to be more e
757.
by tensor 3y ago
I don't understand how pull requests would help with reviews. Peer review is generally looking at the methodology and analysis, not simple typos or wording that might be corrected by a pull request.
758.
by tensor 3y ago
I very very strongly disagree. It's important that researchers get access to diverse data and industry collaboration is often crucial for this. If companies are required to make all their data public they will be far less willing to co
759.
by tensor 3y ago
Hire a baby sitter and go to an office or co-working place. This site is an echo chamber of anti-office sentiment but I can tell you from my experience running a company many many parents actually preferred going into the office exactly bec
760.
by tensor 3y ago
I don't think anyone actually doing NLP research has thought that BERT is always better than simpler methods. Linear classifiers with ngrams, or even better, large margin linear classifiers, are well known to be competitive with things
761.
by tensor 3y ago
FastText isn't a LLM, it's a token embedding model with a simple classifier on top.
762.
by tensor 3y ago
I honestly don't know why anyone would use this gzip approach in production. If you want to do text classification, really the two options you should consider are a best in class linear model like confidence weighted linear classificat
763.
by tensor 3y ago
An artists work is copyrightable, a writers work is copyrightable, an a personal model could reproduce those and also produce new works in the same style. Also, data can be intellectual property without being copyrightable.
764.
by tensor 3y ago
A great many podcasts now have ads of various sorts.
765.
by tensor 3y ago
Those of us who remember when Google first started might even remember that part of it's early success was precisely that it didn't have tons of banner ads like the search engines that came before it. In no way was it the first
766.
by tensor 3y ago
If an individual trains a model on their own data to embody their own skills and behaviour, so that they can then sell/rent that model out to work on their behalf, well in that scenario not being able to treat the weights as intellectu
767.
by tensor 3y ago
There is actually an old argument about whether Computer Science is math or science. The lack of hypotheses and physicality drives the "it's math" camp. I tend to fall into that camp. Most computer science is either math or a
768.
by tensor 3y ago
There is no reason to say "every story has two sides." Some may, and you can call that out, but no, not every story has two sides. For example, a report about a new science result, the science is "one side." The "ot
769.
by tensor 3y ago
Looking at the article talking about this proposal, it appears that the requirement is actually that it's as easy to cancel as it is to sign up, not that it is "one click." A confirmation box is pretty standard practice in UX
770.
by tensor 3y ago
I absolutely despise this "every story has two sides" quip. No, every story doesn't have two sides. Yes, every story has multiple versions full of complete bullshit, but when we talk about a "side" we mean a reputab
771.
by tensor 3y ago
Yeah, I'm really not a fan of this sort of rate limited invite system. I think it will hurt the eventual growth of these services.
772.
by tensor 3y ago
And then they release an updated GPT that breaks all your tuned prompts.
773.
by tensor 3y ago
Absolutely, 100% agree. I just wouldn't touch the original LLaMA weights. There are many amazing open source models being built that should be used instead.
774.
by tensor 3y ago
It has nothing to do with any sort of feeling. Perhaps you should check your own mental state. It is the same as any confidential data. Logs, readings from sensors, etc etc. If it's confidential and given to a 3rd party through a contr
775.
by tensor 3y ago
It doesn't have to be copyrightable to be intellectual property.
776.
by tensor 3y ago
A reminder that llama isn't legal for the vast majority of use cases. Unless you signed their contract and then you can use it only for research purposes.
777.
by tensor 3y ago
I'll try debian again when they move to a yearly release. Using nearly two year old software in the worst case really sucks for some use cases.
778.
by tensor 3y ago
Correct. Essentially the root of the issue is that the news sites believe that the value is in their article, but in reality for the majority of users the value is in the headline, which the news sites give away for free along with the link
779.
by tensor 3y ago
Exactly, this is not human in the loop. The plugin was created without guard rails. A human in the loop guard rail would be "here is an issue template, please confirm to post this". It's really a simple change and this is the
780.
by tensor 3y ago
If you posted your data into a service where the TOS allows this use then yes.

More ›