miej

I agree of course, but I dont think that merely stating ones dislike on an internet forum will be an effective way to effect change.

So I pose the question: What actions may we take in order to help ensure a future of non-implementation of "the user is the product" systems, with a specific focus on those actions that are most likely to be the most efficacious at accomplishing the stated goal in an expedient and lasting manner?

i cancelled my prime subscription yesterday after finally getting frustrated enough by so much of what i've bought/wanted to buy on amazon being clearly fraudulent in one way or another. forget free shipping, i want products that work, and which actually are what they are advertised as. i've wasted too much time munging through reviews to try to figure out if a good product was stealthily replaced with a dud, if manufacturing standards dropped significantly, if the seller was paying for positive reviews, not to mention time/effort wasted on returns and discovering that the products i got were in fact blatantly counterfeit/defective.

yall probably know rule 34, but I'm pretty sure there should be another rule of the internet: "if a product exists, it will be counterfeited on amazon". just try taking a look at how many people on the amazon sellers forum have their businesses/brands seriously harmed by the myriad of counterfeiters on amazon.

i also recommend this write-up for more perspective on the extent and nature of fraud on e-commerce platforms: https://www.dhs.gov/sites/default/files/publications/20_0124...

i believe it would be quite difficult. if you actually look at the structure of the human brain (eg the homonculus model of the cortex) you will note that a very large portion of the brain is dedicated to just the hands/manual dexterity. Many of the current neural prosthetics seem to act by attempting to leverage the brain's use of already-existing neural pathways being used for other functions. Sort of a "turn on my microwave when i open my refrigerator"-type dependence. I think a good sign of truly independent supernumerary prosthetics would be if some users reported phantom limb sensations in their absence

oh, and perfect blue.

magnetic rose by otomo. one of my faves.

it occurs to me that one nice example of a systematic disadvantages being used in order to induce mastery in tech is in the Project Euler problems, where the tasks are generally to solve mathematical problems via programming, but under the constraints that generally, the brute-force solutions have wildly infeasible computation time, but when approached with appropriate algorithm design and mathematical foundation, can be (and in order to consider as 'solved' are required to be) made to run in under one minute of computation time on standard consumer hardware.

personal anecdote time:

I was recently playing Hollow Knight on the PS4 console. Mostly it was ok, but some portions of boss battles just felt downright unfair.

My original setup was to route the HDMI from the console to my AV receiver, then to an HDMI splitter (to strip hdcp), then to my 'gaming, very-low-input-lag' projector (rated at somewhere between 5-15ms input lag). It occurred to me that the difficulty may have been due to extra input lag caused by the receiver, so I tried an experiment, changing the connection order with the exact same components.

So this time I went from console, then to HDMI splitter, with its two outputs going to projector (for video) and AV receiver (for audio).

The difference was like night and day. Suddenly those battles that were 'almost unfair' went to 'almost trivially easy'. My best estimates for the amount of input lag the receiver was introducing was around 125-175 ms range. Pretty wild how much of a difference that made.

On a related, but different note - it reminded me of a concept I've purposely tried to implement it similar domains, that is: training with systematic disadvantages. In this case, the extra input lag acted as a systematic disadvantage that forced me to become hyper-responsive in order to stand a chance at succeeding. Additionally, it caused the viable response window to enemy actions to become short enough that specifically telegraphed moves were such that I needed to treat them as almost completely random, and adjust my response accordingly. In other cases, (especially in soulsborne games,) I specifically never invest skill points into raising my health in order to force myself to become better at the combat dynamics, and simply never getting hit, rather than allowing for health as a 'mess up meter'. Which generally allows for 'taking off the training weights' if ever needed,once a suitable level of mastery has been attained.

Does anyone else here have other examples of utilizing purposeful systematic disadvantages in order to improve one's mastery of a thing? Would especially love to hear any examples outside of gaming too, perhaps in career things/tech/etc.

i wonder what % of the hn crowd cant work if google auth is down

no. and in fact i'd propose that models like gpt-3 can potentially exhibit catastrophically self-defeating effects. consider that nlp models are trained on a corpus of text, much of which is found online. but in the early stages, it is presumed that much of the text content is in fact human-generated, thus providing the model with an essentially human prior to use as its 'ground truth' for natual language parsing and generation.

ok now fast forward to a point in time when the amount of text content online that is generated by gpt-3-like bots is non-trivial, or perhaps even majority. now when the 'ground truth' corpus on which the model is trained is collected, it can no longer be assumed to still maintain a 'human' prior. rather, it will have something more akin to a 'self' prior, with the original 'human' portion dwindling away over time (assuming the rate of generation of bot content out-paces human content in a monotonically increasing way).

so what happens next to the models? well either humans develop better methods, or my best guess is that eventually the models' ability to generate content deteriorates asymptotically towards a very non-diverse language. in a similar process to hebbian learning, such bots released in the wild and used to generate content that they (future generations of) themselves use to train on, generate a large-scale positive feedback loop. and tiny random fluctuations in the models weights eventually grow larger and more pronounced over time, resulting in incrementially more uniform content. potentially to the point where eventually the models only spit out "lorem ipsum dolor sit amet" on continuous repeat. (obviously not actually, but i think you get the point. it would probably be more like "the cat ran and the cat ran and the cat ran...")

could try to combat this by working to identify and only train on 'human content', and that may or may not be effective, insofar as model develpers will generally aim to have their bot content be indistinguishable from human content. a bit like GANs. at that point though, it becomes a little harder for me to reason on/make predictions about

Here, in order to expedite this process, i've set up a google form to collect info from anyone interested:

https://forms.gle/ueJrw9A7CNomB3yH6

also, I set up a discord server to help facilitate live chat: https://discord.gg/vmg2HSD

im down

so is the fbi just reporting on the cia now? I mean, Project Weeping Angel is pretty well-known by now, right?

ref: https://wikileaks.org/ciav7p1/cms/page_12353643.html

moved to oakland for a job, commuted via bart. 6mo later, tinnitus. gotta love the good old banshee scream of bart.

ah yes. I personally like to refer to this as a stargate theory. for example, if the gov actually had a piece of alien technology capable of opening a wormhole to other planets, perhaps the best way to cover it up would indeed be to create a public tv show about it. That way, anyone who tried to argue for its veracity would immediately be considered a borderline lunatic.

ah right, thanks. been a few years since I did any qm work last

wait, but doesn't typical qm operate off of integrating over every possible quantum trajectory, including relativity-prohibited ones where particles take superluminal velocities? so if qtt is valid and otherwise compatible with vanilla qm, how does it address that discrepancy?

"one plus one equals" --> "one plus one equals 12"

guess it still needs to work on its math skills :P

I quite like pilot wave theory, myself. It's a pretty cool formulation.

Hey guys, I put together a little project to implement dynamic neural networks in tensorflow, and I figured someone might find it interesting, so here it is!

Except this is an actual known "evil capitalist" scenario, where pharma execs literally bribed doctors to over-prescribe opioids. RICO conspiracy charges are generally reserved for the likes of the mafia, and even then are rare due to the overwhelming burden of proof required. I'm honestly not sure who I consider worse in this case: the pharma execs handing out bribes, or the doctors who accepted them.

Reference: https://www.justice.gov/usao-ma/pr/pharmaceutical-executives...

So, my question is - just how reliable is this? black-hat hackers are one concern of course, but given the revelations of snowden/prism/vault7, is it actually just the case that we are, for all intensive purposes, living in a post-privacy world? Of course, I've taken some steps to try to help my own internet security, but even for the straight-forward threats, its difficult to know what helps what, and the magnitude of the benefit conferred.

Are there perhaps any websites out there that provide comprehensive guides to building up internet security/privacy, with regular updates about what courses of action should be taken each each new database breach that becomes public knowledge?

This. I specifically used MMOs during periods of my life when money was short. You'd be surprised how expensive food is in comparison to a 25$ monthly subscription fee, and how fulfilling having actions cause their intended results is.

But this is also in light of the recent MA RICO conspiracy charges, which from my understanding is not really a...small deal, right? From what I've read, RICO is usually reserved for the likes of the mafia and even still is rarely used due to the overwhelming burden of proof required. Definitely despicable behavior though, with doctors taking bribes in exchange for breaking their Hippocratic oath

Source: https://www.justice.gov/usao-ma/pr/pharmaceutical-executives...

I just want to point out the notably inconsistent formatting of the presented data in this article. It's like a tour de force of how to create misleading representations from "accurate" data. Specifically, keep an eye out for what is used as the min/max values on the various axes, as well as interval sizing.