(the best) Local LLMs are FOSS though, if bias is introduced it can be detected and the user base can shift away to another version, unlike centralized cloud LLMs that are private silos.
I also don’t think LLMs of any kind will fully replace search engines, but I do think they will be one of a suite of ML tools that will enable running efficient local (or distributed) indexing and search of the web.
When OpenWrite says “publish to the open web, Gemini, or Mastodon” what does it mean by Gemini?
Indieweb is for you! Give Micro.Blog a try, it has native ActivityPub integration. https://indieweb.org/Micro.blog https://micro.blog/
You’re joking right? “making up answers” in the case of search results just means a dead link. If you get a good link 99% of the time and don’t have to use an enshitified service, that’s good enough for 99% of people. Try again is the worst case scenario.
Also, you know what would make this all even worse? Laws requiring that people prove their identity in order to consume content or pull videos… just like age verification laws now being passed in several countries. What a coincidence.
Not to mention that the scraped indexes can and should be shared. Unfortunately what OP is seeing may be a move to thwart this type of brute force scraping, and might resolve as dynamically assigned domain addresses, where the URL of a set object is temporarily assigned and streamed only to a single or group of IP addresses that request it within a given timeframe before being rotated out until found in search again and then reassigned a new URL, etc. This is a frankly stupid use of resources, but can effectively be used to prevent crowdsourced indexes from proliferating, and to punish IPs or even MAC addresses or browser fingerprints associated with downloading and reuploading videos which almost certainly have stegnographic fingerprinting embedded that associate with who the video was served up to at the time it was downloaded.
If only I were the king of the world!
I think what you are arguing for is hardcoding requitement for signatures with an “age appropriateness” ranking into the OS. How does this change the current situation where adult sites and apps are legally required to have an age verification popup/warning? Whether signature based or graphically based, what is at issue here is age verification which means referring to some “repository of truth” outside the will of the user. The problem is that the effect of this is to link government ID directly to web traffick, as to truly verify age requires verifying identity meaning abolishing anonymity on the web and enabling complete tracking of dissent.
I could see a version of what you are describing akin to the way physical cryptographic keys are used to manage DRM on high end enterprise software, where identity/age verification would need to be done by the hardware vendor and not the software/site, the problem with that however is the aftermarket and multiple-user devices. You could say that the “age key” would be a hardware device sold to adults using physical ID akin to spirits or tobacco, something like a SIM Card but preferably with NFC rather than having to be installed in the device. “Adult Access” would then be enabled on sign-in by scanning the “age key”, enabling onboard software to serve software and sites that don’t have an “all ages signature”.
Honestly as I write this, it isn’t the worst solution, the main thing would be keeping the Age Key as an interchangeable, replaceable device that only interacts with the OS and isn’t referenced by other software, so it doesnt just become another Digital ID proxie.
Nice, shame the assistant isn’t on Windows.
Don’t go giving them ideas. That way leads to Digital ID at birth, which should be avoided at all costs.
Simple difference: spiders chemically synthesize long chain fibers not amorphous filament, the fibers are self supporting with tension, whereas hot printed plastics deform under gravity.
You could theoretically print this way with a printer that had a flow reactor nozzle that mixed the reagents for fiber formation on demand instead of a hot extruder, I have yet to see this but it seems likely the textiles industry is working on it somewhere.
Good to know.
Regarding polling efficiency, that makes sense. As I understand it ActivityPub uses a combination of push notifications at time of publishing and pull notifications at time of subscription/query for objects? I can see how offloading that to an instance for multiple users vs every user definitely increases efficiency for content discovery/inboxing. I know there are protocols like websub to add push notifications to blog publishing, but they are required to be done at the publisher/host side. I do see this as a big nudge in ActivityPub’s favor.
Duration of caching is set by the instance admin I take it?
At a minimum this is adding the number of instances that federate a given content streams to the multiple of storage needed to host the content, even if that storage is ephemeral. Not so big a problem at 100,000 users, but at 100,000,000 users this is a lot of storage cost we are talking about. Unless somehow the user/client doesnt cache the content they pull from an instance locally on their device when they view it?
Regarding Authorship, if there wasn’t an issue then ATProtocol devs wouldn’t have made it the cornerstone feature of their network. The ability to move accounts between instances and maintain content control permissions is currently one of the big focuses of development on ActivityPub as I understand it. Also as I understand it many Fediverse instances dont have edit functionaly enabled, meaning once the content goes out it is out of Author control. I’d like to know how delete requests propagate, when the “Object” is deleted does a request to clear cache go out to all federating instances?
My point was this isn’t an issue when all content is self-hosted, because the author as the host can edit, delete, or migrate all they want and maintain full direct control over the source of that content the client interacts with whenever a pull request comes in. Yes the user Caches the content when they read it, but there is no intermediary copy.
Actually there is a legal definition, a Public Benefit Corporation has statutes in its articles of incorporation which legally commit the company to pursue a set purpose which supersedes the fiduciary responsibility of the corporation to shareholders. This is important because it provides some degree of legal protection from activist shareholders suing the company for making spending or policy decisions which don’t directly maximize shareholder value. The body of law around this issue is still relatively murky, but some defense is better than none at all.
For example, shareholders could attempt to sue BlueSky into increasing advertising placement or data sales functionality intothe core platform to increase company revenues, but if that is at odds with their stated public benefit purpose the legal team for BluSky would have grounds to attempt to dismiss the suit on the grounds that the shareholders purchase the shares under the explicit understanding that these functions would be subordinated to the public benefit goals of the platform.
That’s why I said gait recognition, not facial recognition. Last year GaitNet hit over 99% recognition accuracy, given another year of training the error rate will have gone down and the recognition window will have gone up. https://pmc.ncbi.nlm.nih.gov/articles/PMC11323174/#%3A~%3Atext=Diverse+neural+network-based+gait%2Cresearch+direction+is+also+assessed.
Yes, and of course wifi is just one frequency band. That kind of technique could also be used with other frequency bands, I wonder what the resolution of 5G is for that kind of telemetry.
We are on the same page.
But the point wasn’t about the tracking, it was about sorting the data. The data acquisition has existed for years, but was never useful because it could never be processed in an accessible manner. Now AI can sort through the hours of traffick camera footage, identify your gate, and show every location you appeared on camera in the past 24 hours. It can also check cell tower records and record every cell phone that stood by the hotdog stand for more than 60 seconds, and crossreference the hardware IDs and Sim cards on those phones against carrier databases to match the walking gates of the people on camera to their name, address, SSN via credit check, and bank information.
My brother, how many examples do you need? Without mentioning the 30s and 40s, look at the Tuskegee Experiment where men were given syphilis intentionally and withheld treatment to study it’s spread in black populations. Or the San Francisco Operation Sea-Spray when dangerous bacterial were sprayed over the entire city to study it’s effects on the population. Ethical frameworks exist to prevent people from harming others in the pursuit of data, because if they aren’t prevented they will. I order to prevent this harm from being done, a clearly defined set of principles must be stated and written down so that everyone involved can have the same understanding when designing and approving studies.
huh, interesting, a private text based internet protocol where everything runs server side and all sends are encrypted by default. I see the appeal.