First things first, Twitter. Small baby steps before you shoot yourself in the foot.
Focus first and foremost on getting rid of radicalizing terrorist content. That has been on your backlog for years now, while you prioritized more pressing matters, such as cutting off access to developers and apps. You kind of owe it to society and the parents of your teenage users to make this right.
Then, when it is 2023, go after the grieving trolls. Those who mock suicide victim families, DOX innocents, post indecent pictures of naked black people doing lewd things to watermelons. Some people may even help report such things manually for free.
Then bots. Kill 99% of all bots, keep only those with a 1000+ legit followers, or make bot owners verify them with a cute droid badge.
Then after one of Trump's tweets starts WWIII, and you are as significant as Myspace, maybe you can have a look at the health and civility of online conversation.
I'd start with a "My Safe Space" button, where you can automatically unsubscribe to all posts about a certain topic/containing a certain trigger word. If you can't wait that long, there is an easy fix to remove 50% of uncivil content by going back to a 140 character limit. I feel that was closer to the healthy and civilized discussion platform you envisioned Twitter to be when you started it.
1) Centralized solutions are costly to freedom of choice and privacy. Consider a solution where Dropbox, Google Drive, Facebook photos, S3, and YouTube is run through Filecoin. No more "You haven't used DropBox in a year, so we are closing your account". No more "This video has been demonetized for all advertisers in our network.". No more "we ran these nets over all your photos and use the information gathered to make a better advertisement profile for you". This is what happens when users make and own software: Limewire, Bittorent, Popcorn Time, DC++, Napster, Kazaa. Though most of these are now depricated/forced shut down, it was not for having a better alternative available.
2) Centralized solutions are currently much faster, but they may not be in the future. Also, other coins (that offer spare computing power) or commercial enterprises could act as a power nodes (sub-centralized solutions) in a decentralized network, to speed it up for a fee when needed. Decentralized peer2peer content delivery is currently owned by centralized players, like Akamai. I see no reason why it could not work when it is owned by its users.
3) Because the software has to operate in a trustless environment, a lot of attention has to be spend in making these systems more secure. Security through obfuscation is not a fall-back anymore. If we had known that LastFM stored password hashes as unsalted MD5, would we have trusted it with our accounts? We will have fuzzing tools and static program analysis to help automate checking the security of a contract.
4) Forks can be dealt with in a decentralized manner: voting power being assigned by a provable amount of tokens you own. You can run an entire election in this manner, cryptographically voiding any "vote rigging" argument.
5) You may have so much freedom as to shoot yourself in the foot. Therefore, Cryptocurrency banks may provide support for payment solutions, insurance, lending, for people who don't want to store all their money under their flammable mattress.
6) Smart contracts will fully replace the paper notarial system. Criminals will be booked into networked systems where their identity is biometrically verifiable for the duration of their "stay". Trading and lending circle contracts will be heavily vetted and proven standard smart contracts one can select from a library.
> As said earlier, it is lazy learning algorithm and therefore requires no training prior to making real time predictions. This makes the KNN algorithm much faster than other algorithms that require training e.g SVM, linear regression, etc.
Linear regression is way faster than KNN as the dataset grows beyond toy data. Both training and especially testing. In practical applications, test speed often trumps train speed.
> The KNN algorithm doesn't work well with high dimensional data because with large number of dimensions, it becomes difficult for the algorithm to calculate distance in each dimension.
KNN works fine on high-dimensional text. From something simple as Hamming distance on binary tokens, to euclidean distance on TFIDF, to cosine distance on 900-dimensional word vector aggregates.
> There are only two parameters required to implement KNN i.e. the value of K and the distance function (e.g. Euclidean or Manhattan etc.)
Also implement distance weighing (you probably want to weigh the 1-th nearest neighbor label higher than the 5-th nearest neighbor label).
> The KNN algorithm has a high prediction cost for large datasets. This is because in large datasets the cost of calculating distance between new point and each existing point becomes higher.
This is why you "fit" something like a K-D tree during training.
> Finally, the KNN algorithm doesn't work well with categorical features since it is difficult to find the distance between dimensions with categorical features.
Hamming distance works fine on one-hot encoded categorical features. If not, embed/reduce the dimensionality. If not, use feature selection first and do KNN on top 10%-20% features. Remember that you don't have to use the same distance measure for every feature column.
> If one of the features has a broad range of values, the distance will be governed by this particular feature. Therefore, the range of all features should be normalized so that each feature contributes approximately proportionately to the final distance.
You can skip this step (and feature selection) by learning an additional weight for each feature to multiply with before distance measure. It is rare for each feature to contribute proportionately to the target.
Anecdotal, but: I myself turned down a job offer in SV in part due to Trump. When asking 4 random co-workers, none seem keen to move to the US, 2 of them even citing safety concerns (they are minorities from India and Africa and heard horror stories of former colleagues still in the US).
Companies don't have constitutional rights nor obligations. If wrong, please correct me on this.
As it is private property, it is free to set its own rules, even if those rules impede on my free speech. If Facebook was a government, it could not make those rules as per first amendment.
You can still sue of course.
Look, I feel AI-powered surveillance is scary, and think China is going too far (perhaps showing us a glimpse of the future) in this. But where it is state-run surveillance in China, it is capitalist-run surveillance in the West. That Europe needs to step up and protect its inhabitants is a sign of what happens if you give companies free reign in handling (private) data: A big privacy mess.
You have no right to free speech or right to assemble on Facebook.
My face is recognized, tagged, and stored in Facebook servers, shared with intelligence agencies, and cross-referenced with all other photo's. And I don't even have a Facebook account (leading some employers to distrust me -- "what do I have to hide?" -- and lower their "social credit score" of me).
In contrast, I am relatively safe against China's spying apparatus.
Their vision of the future is fully inspired on, if not outright copied from, the US. China doesn't care about "keeping up appearances", but it is very much doing the same as the US, only more blatantly.
And where in China it may be the government, in the US it is EvilCorp or some other entity not bound by the constitution.
Focus first and foremost on getting rid of radicalizing terrorist content. That has been on your backlog for years now, while you prioritized more pressing matters, such as cutting off access to developers and apps. You kind of owe it to society and the parents of your teenage users to make this right.
Then, when it is 2023, go after the grieving trolls. Those who mock suicide victim families, DOX innocents, post indecent pictures of naked black people doing lewd things to watermelons. Some people may even help report such things manually for free.
Then bots. Kill 99% of all bots, keep only those with a 1000+ legit followers, or make bot owners verify them with a cute droid badge.
Then after one of Trump's tweets starts WWIII, and you are as significant as Myspace, maybe you can have a look at the health and civility of online conversation.
I'd start with a "My Safe Space" button, where you can automatically unsubscribe to all posts about a certain topic/containing a certain trigger word. If you can't wait that long, there is an easy fix to remove 50% of uncivil content by going back to a 140 character limit. I feel that was closer to the healthy and civilized discussion platform you envisioned Twitter to be when you started it.