151–160 of 172 posts
The demo docs site says the feature is only available to 'Ultimate' members, while the linked page is ambiguous. Is it the case that Sidekick is only available for such Kagi members?
Can you share a link to the page that says so (not correctly).
Earlier quoted context omitted.
Watermarking is not at all like a digital signature and a lot like steganography. I only have a surface level understanding of the process, but it works by biasing token selection to encode information into the resulting text in a way that's resistant to later modifications and rephrasing. I have my doubts about the effectiveness of this method and realistically, it won't make any difference because the bad actors wi…
The only way to make that stenography robust is to have the encoded message be generated with some secret key that can be verified. Otherwise anyone could manually fake the stenography into human typed messages assisted by some encoder and you'd have no way of telling if it was really typed by an LLM. That line of thinking is what makes it have to be like a signature to work like you said for "any sentence". I also t…
Earlier quoted context omitted.
It's impossible to make. You cannot prove any sentence was created by an LLM and you can't prove it wasn't.
Unless you design the LLM yourself and purposefully watermark the output. https://arxiv.org/pdf/2306.04634.pdf
Earlier quoted context omitted.
Thanks for linking to the docs. I really like this feature, but as you said, it felt like it showed up at random
Since last week, it will trigger whenever your search query ends with a ? or starts with How or Why or other interrogative words https://kagi.com/changelog#3179
Earlier quoted context omitted.
Is there any reliable AI generated content detector today ? I've tried many free and paying ones online, but they're aren't reliable
It's impossible to make. You cannot prove any sentence was created by an LLM and you can't prove it wasn't.
The problem is that people keep consuming the samey low quality content instead of skipping it (think superhero movies and Netflix series that are all indistinguishable from each other). As long as they're satisfied with that, they'll fall for fake product reviews too.
Earlier quoted context omitted.
Problem is that many webites used to hire writers which wrote tangentially related posts to get their main product higher ranked. Like LogRocket and Partition Minitool do. Combine that with that guy who boasted about his 'SEO heist', I think it's a very valid concern.
Isn't this exactly what 37signals and even joelonsoftware were? Isn't HN essentially a free conduit to YCombinator awareness? I don't see the problem with what you're describing. It seems like one of the most contributory ways to market well.
Another decent one would be linux sysadmin info from Digital Ocean and the likes.
But for every joelonsoftware there are 99999 sites that have all copy/pasted the same tutorial about something basic and try to push some random product or just ads.
Earlier quoted context omitted.
Ultimate/normal customer, and i agree. Personally i think AI _is_ Search, and while we don't need to force them together in some massive behemoth now - laying foundation for being familiar, comfortable and well integrated in the future seems foundational to today. In my eyes at least.
Strongly held belief that I think is backed up by academic consensus: we should not treat LLMs as stores of knowledge. As my mantra goes, “language models, not knowledge models”. So the future of search might involve LLMs at a very fundamental level (and I think it will!), but they’ll never be the central component. Humans will never ever ever invent a better knowledge system than a database / a piece of paper, I gua…
Earlier quoted context omitted.
Is there any reliable AI generated content detector today ? I've tried many free and paying ones online, but they're aren't reliable
It's impossible to make. You cannot prove any sentence was created by an LLM and you can't prove it wasn't.
Someone did something like that to identify HN authors (as in correlating similar writing styles between pseudonyms) a few years back, for example: https://news.ycombinator.com/item?id=33755016
Or a study applying similar analysis to LLMs: https://arxiv.org/pdf/2308.07305
Of course, LLM output can be tweaked to evade these, just like humans can alter their writing style or handwriting to better evade detection. But it's one approach.
Earlier quoted context omitted.
I have personally started to avoid Google’s and others immediate answers, because they are often very wrong because they are picked out of the contex. They try to show what you want to see, but it often means something else.
This is so true! While planning a trip abroad last year we were unsure about whether $thing is legal in $country. Google proclaimed in bold letters that, yes, $thing is legal in $country, but this line was taken from a site with the title "common misconceptions about traveling in $country" an in fact $thing was not legal. Such a basic mistake, I haven't trusted the instant results ever since.
Earlier quoted context omitted.
I have used Kagi for several months now, and I find that the ability to decide which searches I want quick answers for to be useful. There are certain searches where I feel comfortable with accepting a quick answer, and others where I don't think it would be useful. Being able to avoid the clutter unless I want it is nice.
Sports scores and stocks. If kagi inlined those results I would be very happy to not need to use the Google bang. Additionally, the google bang destroys suggested terms and with no kagi history I'm typing the whole query.