Y
HN Search
Hacker News Search
new
|
comments
|
top
|
jobs
GTP
searching PlanetScale…
1.
▲
2.
▲
3.
▲
4.
▲
5.
▲
6.
▲
48 ms
·
811.
▲
by
GTP
3y ago
Why should you be forced to buy a 20k€ Bosch rig? I don't see why there shouldn't be competing products offering the same functionality.
812.
▲
by
GTP
3y ago
Honest question: why do you use a VCS to store the documents? The kind of documents OP is talking about don't get updates, same for the photos.
813.
▲
by
GTP
3y ago
> Today I wanted to really knuckle down and make some real progress on my side project. I've updated my iPad, fixed the leg on a tripod, cut the grass, cleaned the mower, scrubbed the driveway. Guess how much progress I made on the
814.
▲
by
GTP
3y ago
The story about Ivermectin is much more complicated, I suggest you to read this: https://astralcodexten.substack.com/p/ivermectin-much-more-t... It also talks about why different people might develop different view in
815.
▲
by
GTP
3y ago
> Cult members say their protocol is built from the same digital DNA as the Tor browser and the chat app Signal Then just use those, as they are bettle-tested and received much more scrutiny than this new framework.
816.
▲
by
GTP
3y ago
I never heard about class generators before, could someone please point me to some resources to learn more about it?
817.
▲
by
GTP
3y ago
I'm not sure whether I agree or not, as without cryptography the only way to have a private discussion with someone, which I think is equally important to transparency, as people usually are "transparent" about personal stuff
818.
▲
by
GTP
3y ago
Maybe you're right. I was biased by my native language, which doesn't have the a/an problem that English has.
819.
▲
by
GTP
3y ago
For the spaces and for some (maybe most?) languages you don't even need a NN to add spaces: as words made of two or more words aren't that common, and when those occur you probably want to use the composite one, it boils down to s
820.
▲
by
GTP
3y ago
> If you use less powerful AI your products will most likely lose to competition who gamble and use a more powerful model (in theory) I think it really depends on many factor and there isn't a one size fits all answer. Depending on
821.
▲
by
GTP
3y ago
Yes, you said that those aren't directly comparable, but your comment seems to imply that you can account for the difference by having more artificial neurons than the number of biological neurons that we have. Sorry if this is not the
822.
▲
by
GTP
3y ago
> I saw mentions that SMS is insecure, but I never heard about a credible remote attack that didn't include provider cooperation or taking over your phone. Technically correct, but how likely those attacks are seems to depend on the
823.
▲
by
GTP
3y ago
Yes, KeePass (and KeepassX and KeePassXC) support generating 2FA codes, but it is generally not a good idea to store those alongside your passwords, as if you do those aren't a 2nd factor anymore. You can mitigate the issue by having a
824.
▲
by
GTP
3y ago
> the human brain averages having approximately 86 billion neurons, and GPT-4 has an estimated 1.8 trillion parameters. But that's not an apples-to-apples comparison, as the artificial neurons of neural networks are just a rough app
825.
▲
by
GTP
3y ago
For civil use, each country will decide its own regulations. For military use, the UN seems the most natural choice.
826.
▲
by
GTP
3y ago
Or, we try to set common rules before the thing goes out of hand. We already have UN rules about fire weapons, and your argument applies to those as well. But just as we have widely agreed upon rules for weapons, if there's the chance
827.
▲
by
GTP
3y ago
Wikipedia says: "When a process crashes, it neatly exits and sends a message to the controlling process which can then take action, such as starting a new process that takes over the old process's task" But this seems only a
828.
▲
by
GTP
3y ago
> The computer could easily solve the problems without displaying this, while it provided a back-up version of the system for us, so that we could continue to work with our tasks undisturbed" How could this work? Teorically , the O
829.
▲
by
GTP
3y ago
> "All IP (trademarks, copyright, patents) from one of the two merged companies gets released to the public" This sounds very extreme to me, sometimes acquisitions are done exactly because the buyer is interested in the other c
830.
▲
by
GTP
3y ago
> (hey, won’t be in office when it does, this is your problem next generation, haha~) Isn't global warming a problem we're trying to ignore and offload to the next generations too? Despite all you said is true, I still think we
831.
▲
by
GTP
3y ago
It was already known that some kinds of plastic release microplastic particles when heated. I assume this is why we have some plasticware that is labeled as being suitable for microwave oven usage, but at quick glance at the abstract doesn&
832.
▲
by
GTP
3y ago
What did Bitsavers do that was/is so interesting? A quick look on Wikipedia didn't turn out anything.
833.
▲
by
GTP
3y ago
But if the fundamental issue is that the JS code that runs in your browser is provided by the same entity providing the back-end, couldn't we have a different entity that -on a different domain- serves JS code that interacts with the b
834.
▲
by
GTP
3y ago
> The difference is that ambulances/police are allowed to break the rules in an emergency - famously, ambulances have a legal right to speed and run red lights in an emergency, and cops obviously have a right to trespass. I think th
835.
▲
by
GTP
3y ago
> If the test asked "should violator be punished?" I think it would be more meaningful, otherwise it's just synthetic and the controversy is just about semantics This is how I interpreted the test, but you're making a
836.
▲
by
GTP
3y ago
But this has nothing to do with the quality of a consultancy from university researchers. They know very well how the models work and the direction the research is taking, they not having the same amount of money available for training is i
837.
▲
by
GTP
3y ago
I think it's exactly the opposite: researchers are currently studying models that are better than ChatGPT. The reason why when we use one of the open source models those tend to feel worse than ChatGPT is twofold: 1. Those models are
838.
▲
by
GTP
3y ago
> ask people involved in the field, and risk conflicts of interests True, but in this case you can greatly mitigate the risk of conflict of interest by asking researchers that aren't directly involved with companies making money wit
839.
▲
by
GTP
3y ago
Not only that, but they also kept saying that governments around the world need to be quick at regulating AI usage. Many people pointed out how that could have been just a way to have some free advertising, this is the proof.
840.
▲
by
GTP
3y ago
But then how is this so different from running "docker-compose" and then do whatever you want withing the container? Is the difference just that they provide ready-made Docker images for certain environments so that you don't
More ›