Thanks for that link, I read through that and absolutely love it! I already downranked sites I found to be AI-written on a personal basis, but this could be much more powerful. From what I understand, there’ll be human review of every AI report, so the potential to abuse the system is also relatively low (if the Kagi team does their due diligence)
Like most search now Kagi has chosen to include Instant Answers that are AI generated, which means they’re often wrong
You briefly mentioned in your user-experience-list that the AI answers are only there when you want them to be, but I just want to emphasise it, since to me it makes a world of difference in comparison with other Search Engines like Google. You only receive an AI answer if you press a specific “Gimme AI answer”-button (which is very unobtrusive) or add a question mark at the end of your search query!
I rarely jump to the defense of some company, but I only know of this one lori-person who tried to lay out reasons why Kagi is bad and, as you showed very well, @[email protected], most of their reasoning/arguments aren’t really all that good when you look at them in more detail. And when they just plain refused to take an interview with the lead of Kagi, by burying their head in the sand and going “I don’t care if you want to clear up misunderstandings, I don’t want to talk to you!”, it kind of sealed this person as not being a trustworthy source of criticism and more of “I’m mad that a new company is doing something different than other companies in the same sector”
I’ve been using Kagi for about 3/4 of a year now and I will certainly renew my annual payment to them. Of course it’s not a magic bullet without any flaws at all, but currently it does the things it offers much better than any competitor I could find and all they want is around 10€ per month. They won’t spam you with advertisement nor will they suck up your (arguably infinitely more valuable) private info to sell to the highest bidder. For now, Kagi has been doing and still is doing more good than most other tech companies.
I would love for it to be different, but they’re mostly right.
The hardcore shareholders, who probably have shares in more than one company and for sure only see these companies for their monetary value and nothing more, would not care if the company’s creative work featured AI giveaways like twelve-fingered people occasionally and inconsistent storylines, if it would mean they could save on all their artists salary by paying only for one AI subscription.
Yes, you can still tell (mostly) when something’s made by AI, but the fact is that we already do see creatives being replaced with AI, leaving them free to do dishes and laundry instead of the other way around. The Coca Cola AI ads are one prominent example. Executives and shareholders don’t care about their product being inferior if it means it saves them even 20% in expenses. And we both know that replacing all your creative team (often even just one or two) with AI is a bigger saving on “Creative expenses” than just 20%. We know that because we can literally look up salaries vs subscription price for stuff like Sora and Veo3.
Yet, contrary to what I perceive as your main argument here, we don’t see widespread adoption of AI in all kinds of companies to do the tedious labor. That seems to still be done often either by traditional methods, because LLMs and generative AI is just not good at repairing a leak in toilets or checking for damages in a factory or welding or even just pushing a button to announce break-time.
Edit: spellings
I think it’s more simple than you assume. From my limited experience (many stranger’s anecdotes and my team recently being fired literally because “the other (very different) production location is able to do it without a dedicated Quality Management team”) most employers / company chiefs just want to make more money or, at least, increase the perceived value so that being bought out becomes realistic and leaves them with more money. They don’t actually care if their product works well or efficient, as long as number go up. Maybe the original company founder does but how many companies are still there that have the founder for long-term in key decision making and without shareholders who kinda hold the real power and couldn’t care less if the company cleaned up oceans or burned children because to them it’s just one combination of letters that make them money?
As @[email protected] suggested, the top management might not even understand that AI won’t help, so they think it will make a short- (savings due to firings) and long-term (increased efficiency or otherwise better product) profit. And those that are very informed about AI understand, at the very least, that they can increase short-term profits by firing employees (thus saving on needing to pay salaries to pesky humans) under the guise of increasing efficiency.
So to top management it’s just a decision of “do I want more money now and in the future?” or “do I want more money now and maybe also trick idiots into buying us out before it goes belly-up?”
Lastly, I think you might ascribe more self-reflection ability to middle management than they have. I want to believe that most of them truly think they are a crucial part of making the company work, so they don’t even see that replacing humans with AI would make them obsolete and thus prone for firing.
Unfortunately the main accusation - that Meta systematically downloaded porn to use for AI training data - might actually be false if they aren’t lying about only having downloaded around 22 vids per year. Seems way too little data for a training set to me. Nonetheless they shouldn’t get away with downloading files they didn’t have permission for.
You’re right. With all the shit being brought into the spotlight, I easily forget to see all the good the internet is still being used for.
Amazing how in less than the span of my 31 year lifetime the internet turned from absolutely world-changing great and probably directly or indirectly the most fun and useful tool humans can play around with to not-yet-unusable-but-definitely-closing-in slop where you can’t spend an hour without getting mad at some system working against us.
Edit: I hate ending on a dismal note, so let me remind everyone that the “small web” (human-created blogs made with nothing but passion for their craft) exists, even if it’s harder to find. Similarly, for most enshittified websites like YouTube, there’s an alternative to the enshittification (even if it’s just getting something like Grayjay to circumvent the crap on YT)
As @[email protected] said, this is about much more than software. You can’t pirate a gym (excluding the Venn Diagram of probably 0.0000001% of people who both want to go to a gym and know how to hack themselves into said gym’s database).
Click-to-cancel hurts every consumer in America and only benefits the providers of any subscription service.
Yeah! I used DDG for quite a while and it’s pretty okay. Kagi definitely isn’t without fault but for me it’s the best true alternative to Google and I happily pay for it. Allows me to save so much time (cumulatively) by just guiding me to the actual result in most cases (instead of sponsored and ad-infested garbage sites)
Alright. I had to read up again on why this is newsworthy in the first place. Because of the language in their new ToS regarding usage of user data. The article I read, asked why they would only now update their terms despite the California Privacy Act having been in effect for a while now.
I’m very sure, optimistically assuming they are honest and really didn’t change the way they handle user data, that an auditor found the previous wording of their ToS just not clear enough. Working in Quality Management and having attended quite a number of audits, this happens all the time. Company has a process for years, sometimes decades, but then needs to change the wording in a document because a new and overly by-the-books auditor will demand such to have it not only be “correct in spirit” but also “technically correct”. Nothing in the actual process needs to change.
Again, this is me assuming that they really havent done something different in the way they handle data. Isn’t Firefox open-source? Could some savvy code-reader go through it to see if something about the data collection has changed?
Look, man, it’s trying its best. And frankly, I think it’s about as ready as it could ever get to replace every single billionaire in the world, considering the sanity of many billionaire’s choices we hear about lately.
True, true. We literally omly have a wood furnace, so are absolutely not affected by this, but I’ll see how reporting potential GDPR violations in the name of someone else works.
Sorry that it didn’t land as an obvious joke. With the NSDAP NPD AfD on such a steep rise, I think I have transcended gallows humour and arrived at necromancer humour levels to be able to cope with this reality.
Thank you a lot for the load of information! I just now got to reading it all. I was very skeptical about the fact that it is fed by the output of other LLMs but the way you explain it makes sense to me that it might not be that much of a problem. I guess a super blunt analogy could be “It’s only incest if it’s your children” lol
Thanks for the explanation. I don’t understand enough about large language models to give a valuable judgement on this whole Deepseek happening from a technical standpoint. I think it’s excellent to have competition on the market and it feels that the US’ whole “But they’re spying on you and being a national security risk” is a hypocritical outcry when Facebook, OpenAI and the like still exist.
What do you think about Deepseek? If I understood correctly, it’s being trained on the output of other LLMs, which makes it much more cheap but, to me it seems, also even less trustworthy because now all the actual human training data is missing and instead it’s a bunch of hallucinations, lies and (hopefully more often than not) correctly guessed answers to questions made by humans.
Does open sourcing require you to give out the training data? I thought it only means allowing access to the source code so that you could build it yourself and feed it your own training data.
Wouldn’t be the worst thing. I don’t mind the EU cooperating with USA but we’ve been piggybacking too comfortably off many US “services” when it comes to consumer tech and the military. That way we forgot to build something on our own in case of “Europe’s Saviour” turning into a political (if not, yet, military) enemy.
Watched this video earlier today and I definitely hope German city planners in my area don’t embrace this required car-centric approach to infrastructure more than they did for cities like Munich .
Smart! If my company ever digitises enough to make this a possibility, I’ll implement (or at least suggest) it!
If you’re tech-savvy enough for it, please consider setting up an Immich library for your photos. That way, they won’t ever need to leave your home while still providing surprisingly powerful face recognition and photo search by context.