equestria

5 ms·

31.
by equestria 2y ago
Mozilla has ~700 employees just to keep an ailing browser on life support. Brave has ~250 employees, but they're building largely on Google's core engine, so they're getting a ton of engineering for free. Browsers are massive
32.
by equestria 2y ago
Sort of? Mozilla is not doing well. Further, the only reason Google is paying Mozilla is to keep a notional third-party competitor alive; search traffic from a sub-3% browser is not worth that much. If the Chrome deal goes through, there&#x
33.
by equestria 2y ago
Because you need to pay something like 1,000 engineers - and not just any engineers, but engineers used to Alphabet's SF Bay Area salaries and equity packages. This quickly adds up to billions of dollars. You have the option to massive
34.
by equestria 2y ago
For folks who are curious about what it actually does, check out the garak/data/ subdirectory. For the most part, it just seems to have an array of static prompts, e.g.: https://github.com/NVIDIA/garak/bl
35.
by equestria 2y ago
It's basically the same thing, except the initial charges in each cell are created through the photoelectric effect when the sensor is exposed to light. After that, it's the same conveyor-belt-style readout proces as for the BBD.
36.
by equestria 2y ago
> As opposed to the DoD, which strictly fights for freedom, liberty, and democracy? Yes, the whataboutism is unwarranted here. The US government is no angel, but is far more constrained in this regard. The bar to become "the enemy o
37.
by equestria 2y ago
> especially with their strict control of information. Most people there will have difficulty independently learning about cybersecurity. I'm puzzled by this assertion. I know quite a few self-taught infosec folks who grew up there.
38.
by equestria 2y ago
I'm not sure this is a useful test. You can most certainly get an LLM to infinitely "correct" or "improve" its own output. But take the "The work uses graphs..." paragraph and plop it into an AI text detec
39.
by equestria 2y ago
Paste it into any AI detector (e.g., https://quillbot.com/ai-content-detector ). They're not perfect, but they're pretty good in the aggregate. This text is almost certainly generated by an LLM.
40.
by equestria 2y ago
There are definite proofs in mathematics. In natural sciences, I think you're sort of restating what I said: there are various levels of confidence based on experimental data. Because it's not binary, the hypothesis / theory
41.
by equestria 2y ago
It wasn't a terribly useful distinction to begin with - especially in natural sciences where there is usually no definite proof, just various degrees of confidence. We had quite a few linguistic purity battles in tech, too. You never w
42.
by equestria 2y ago
But... there's nothing to learn. The publishers measure this, do A/B testing, and most assuredly, they wouldn't be making more money with fewer ads. Most visitors are not blocking ads, and there's money to be made on e
43.
by equestria 2y ago
Market size matters. There's a whopping total of 71 bidders on that.
44.
by equestria 2y ago
> Maybe license people to buy meds and take it away if they abuse it? That's better than the draconian system we have now. I sympathize with your broader point, but... how is that better? "Sorry, you were buying too much nasal
45.
by equestria 2y ago
Yeah, and it's pretty simple. You have essentially three composition layers: original text; a gaussian-blurred zero-offset "shadow" to add glow (CSS text-shadow); and then a pattern of darker horizontal stripes on top, produc
46.
by equestria 2y ago
This is actually an interesting rabbit hole. Their other constraint is that they are trying to be transparent and fair, for some definition of that word. They need that for Section 230 protections, they need it to fend off antitrust lawsuit
47.
by equestria 2y ago
Not really? The vast majority of helpful websites monetizes the same way the spammers do. Plus, even without affiliate links, purchase attribution is already scary good. Did you know that credit card issuers and ad tech companies collaborat
48.
by equestria 2y ago
I'm not convinced this is the explanation. It's true that for some product-centric queries, you get mostly paid results. But for information-seeking queries, Google tries to give you organic results. The problem is simply that the