Algorithms are not even close to substituting for human customer service, human moderation, and human judgment.
Maybe this means web-scale sites like Amazon's marketplace aren't viable. I'm fine with that implication. Stores like Target and Walmart seem to have the same prices with human-curated inventories, and I don't really miss the millions of extra products Amazon has.
I don’t think the answer is humans or algorithms, the answer is organizations incentivized to solve customer problems. Google doesn’t care about customers, if they did, they would fix this. Customers are locked in because where else will you go.
Just like Comcast doesn’t care because there’s only one cable company, Google doesn’t really have any AdWords or Adsense competition.
I think the solution is more competition so use ddg. Once customers have mobility, Google will work to retain them.
It's harder to step on a landmine with Google, but Google's landmine is far more devastating.
We've already seen lots of news stories about algorithms shutting down people's Gmail accounts, developer accounts, YouTube channels, etc.
I personally went through 6 months of hell when my brother's Gmail account was locked due to "inauthentic behavior". He had become disabled and unable to log in, and when I logged in from my machine, Google's algorithm flagged it.
I went through an infuriating loop of trying to unlock it and hitting the same algorithm. There was no number to call. I would've paid $10,000 to get back into that account, which would've required 10 min with a customer support rep.
I tweeted at Google and they told me I had to log in to Twitter with his account in order to get help (obviously impossible for me). It was the most angry I have ever been at any company. It locked me out of important medical documents, family messages, social media -- everything.
I finally fixed it by asking a friend who was a marketing exec at Google to help. He said there's an internal form you can use for this situation -- only available to Google employees.
Other people have had similar experiences and had livelihoods destroyed.
Whatever you think about Comcast, they will never ignore you and let your entire digital life get locked away without any kind of recourse.
I'm sure most people are generally happy to wait 2 days for something to arrive as well. Honestly, it almost feels like Amazon is trying to become an Alibaba and potentially look to take them on with the level of product quality that is now available there.
Half the results on any given product I search point straight back to Alibaba if I image search.
Sometimes I check if Aliexpress has it too, and if I'm not in a hurry I get it there for a huge savings.
Soon, your appeal will be replied to by an AI, too. That way, most people will have the impression that they were reconsidered by a human and found in fault, which will likely make a large percentage of them give up. We're stonewalling real humans by building fake humans :(
And that means GPT could possibly reduce support costs for Amazon.
I don't think this is cynical. It's discussed openly. They want to provide the tools to enable all companies to automate customer service, which is where Facebook's big chatbot push came from.
And that's by design. They don't want to talk to you and help you. You're going to buy there anyway and we all know it.
For example, I currently have a company that I technically manage. It has one employee who does maintenance on the infrastructure (AWS Elastic Beanstalk), database (AWS RDS), and code for five different products. The only other people who touch the code are security auditors.
That just wasn't possible 10 years ago. Managing servers alone for those products (~50 EC2 instances, ~10 load balancers, 10 database instances) would've been a full-time job.
If your human support team made a mistake, there was an appeal process. If your AI makes a mistake, sorry, bad robot, deal with it.
While people talk up Googles ai (at least they once did) the truth is while they may put out some good research now and again most of the stuff they deploy is crap by design - they can’t afford to spend a dollar per day per customer to run a complex algorithm on your data, they can only spend fractions of a penny.
This severely limits the quality of the ai deployed by large consumer oriented megacorps.
On the other hand you look at some of the ai being deployed in the b2b space or anywhere there is a real exchange of money and you’ll see some pretty expensive and interesting ai research and deployments.
But the market is driving more and more money/resources to those companies/services.
Let alone Amazon/Youtube/Google are very different businesses, you can say Google has non-existent customer support, but at least for consumers, so far my experience with Amazon is more positive than negative.
I don't dispute that. But a lot of other people have issues with dropshippers, fake products, radioactive products[1], unreliable reviews, etc.
When I stopped using Amazon, I realized I was returning ~90% of all my purchases just because the product was fake, described wrong, low quality, or just not durable enough to last even a month.
Yes, it's great to be able to return things, but it's a lot nicer to feel reasonably confident that the person who sold a product to me actually believes it's a good fit for what I want as a consumer.
1. https://www.japantimes.co.jp/news/2017/09/20/business/nuclea...
On both of those sites, you can change the filter to show only products that are available in stores.
Those in-store products still have whatever guarantees they did 20 years ago (selected by a human working for Target, probably going to be pulled off shelves if they result in too many returns).
Algorithms are human judgement.
In a strict code sense an algorithm is a way of encoding a set of assumptions that you decide on when you're developing the code to run it. Even in unsupervised machine learning an algorithm is the result of human judgement because someone decided to delegate the decision making process to a machine.
Every time a computer makes a judgement call there is a human who is responsible, and should be accountable, for that process existing.
To pull it back a bit, "human judgment" in my usage meant "input processed by human using a brain" and algorithm meant "input processed by machine".
Given those premises, the argument that "algorithms are human judgment" is untrue both by definition and by observation.
For example, when I ask Google/Siri/Alexa a question like, "What is the best way to set up my router?", I will get a wildly different answer than if I ask a GeekSquad employee.
They may be great for 80% or even 99.9%, but they cannot be 100%.
People use them because they are cheap and convenient though. It's the same with algorithms. If you charge margins high enough for good customer service you will get out-competed by those that don't on price.
> Maybe this means web-scale sites like Amazon's marketplace aren't viable. I'm fine with that implication. Stores like Target and Walmart seem to have the same prices with human-curated inventories
Not only are they viable, they are hugely more profitable without the human expenses. This inhuman model - yes, I think that is exactly what it is - will displace everything else that simply can't compete in runaway capitalism.
I don't believe this. The reputation loss alone will cost you a lot of potential money. Stadia is basically dead because the Google brand is completely untrustworthy. B2B customers want to avoid Google as much as possible because it's just a liability that threatens to ruin their business.
Of course, that's not what we've got...
The reason it isn’t sufficient to be human-level is because individual humans have all kinds of biases and blindspots which scammers design their scams around.
I’m kinda thinking it may be a necessary prerequisite, as scams could involve anything a human wants, but it isn’t enough.
Given that's possible with human moderation too (cf tax codes), what makes you think AGI will solve the problem?
I agree with you, but what I meant by "viable" was more from a consumer perspective.
> This inhuman model - yes, I think that is exactly what it is - will displace everything else that simply can't compete in runaway capitalism.
While I think you're mostly right, Amazon is eventually going to burn enough buyers that they go back to the more trustworthy brick-and-mortar stores.
Most people I personally know have already done this. They may search for something on Amazon, but they will go and buy it somewhere else.
I'm obviously part of a very specific subset, but it's going to hurt Amazon eventually, especially if the US takes a more consumer-friendly regulatory approach than it has for the past few years (which it will).
Of course none of my competitors using the exact same data set had any such problems.
I tried for YEARS to appeal it. There are simply no humans working at Google and nobody reads your emails.
Edit: Actually, I did get a response a couple times but it was obviously automated. They just said to remove the ads from the pages where such words are displayed. So I added a simple rule and a column in the database to hide ads for those keywords. That just triggered the bot to move down the list of their "obscene" language. Next it was the names of various sexual positions, acts and fetishes (Japanese does have a very rich vocabulary in that topic), then manga slang, even silly sounding onomatopoeias that when explained in plain English are "vulgar", etc.. It seems once your website is flagged there is simply no way get clean.
Google accounts have enough worth & history associated with them that they should be able to create some kind of appeal process whereby if you jump through the right hoops proving identity and such, you could eventually reach a human who can intervene?
It feels like they're religious about the idea of having an algorithm decide everything. Works pretty well for some things, but they sure do burn some customers/clients pretty badly along the way for other things.
:(
I realize it's complicated, but there are things that make me worry the future of the internet might not be so great.
It's that Data Scientists don't realize most of their models still simply fit a conditional expectation - and given all their power, it's going to be us, not Amazon or Google, who has to adapt to the distribution as not to repeatedly get hit as some sort of outlier.
It's a dystopia where AI works because of us trying to conform to it, because otherwise we are out of luck. At some point, we self select into Amazon-humans, Google-humans etc.
The solution isn't easy, but it's simple: stop using Amazon (specially AWS), Google (specially AdSense), etc. Once they lose half their annual revenue, they'll start taking things seriously.
No humans, just AI (ML) decisions.
Had my API key revoked because it hadn't been used for 90 days. They provided a link to reclaim it, yielding a LONG form where most questions were only applicable to businesses and service providers.
I answered to the best of my abilities, requesting them to give me access again,also exhausting other available communication options on the matter.
That was about 6 months ago, all I got so far was automated emails telling me it's in review.
Skynet's greatest achievement was to make you think it's not real.
Some folks here are in touch with it though. Perhaps you'll have a better luck using HN as the google support channel.
One of the problems with very large corporations like Google is the marginal benefit of each additional customer is negligible. They have so much money that they don't really need extra business and would rather pursue other objectives. Look at all the stuff getting cancelled lately for whatever reason. The whole thing takes place extra-judicially and is not in the monetary interest of these companies. Stakeholder capitalism if you will. I think the old profit driven system was better because at least you knew what the rules were. Now, the way these big companies make decisions is largely opaque and based on secret rules and arbitrary decisions often based on nothing but whim, or worse, a "good enough" algorithm.
Even bigger newspapers sometimes don't have ads on the more morbid type of news pages.
That puritanical, prudish attitude has done so much damage already, we should stop accepting it as "Americans gonna American".
Source: I work in Google. Not in Ads, but this topic has been discussed in TGIF at some point.
I guess going to Target is much easier to get the necessary shopping done.
Salomon, Nike, and Garmin all got direct purchases from me recently because of what you describe.
I've abandoned Amazon because of that, until they finally add a "no third party sellers" option. And also a "don't show items we don't actually send to your country" option.
I'm a technically proficient developer who's very careful about where I spend money online. If you're not very internet savvy, it'd have been really easy to miss that. In the end I've ordered from a national chain retailer which are doing click and collect nearby.
- Buy shoes from the Addidas website and you can have your name printed on your shoes for free,
- Buy from Quantas or Ibis instead of a price comparator and you’ll have free cancellations or rebooking for a period.
That you know of. There's really no way to know. Counterfeit products could even be manufactured at the official facility, just with no (or even failed) quality control.
https://www.theatlantic.com/technology/archive/2019/01/count...
The existing anti trust rules are enough to push these monopolies to stop the anti competitive activities. Just like anti trust rules were used against Microsoft many years ago, which opened door to online competition like Amazon, Google, Apple etc.
I am a bit down because I’m not aware of any success in fixing this kind of problem through regulation.
I’m trying to think of a regulated industry or company that has service as good as even Walmart and can’t come up with anything.
The hope would be that at least they have some bank level of regulation where they at least have to respond by snail mail and stuff.
EU (and UK) citizens have the right that automated decision making that affects individuals must:
* be documented, and individuals must be informed about it.
* include simple ways to request human intervention or challenge decisions.
* be regularly checked to ensure it's working as intended.
This applies to anything with "legal or similarly significant effect on individuals". As examples, the ICO there has "automatic refusal of an online credit application" and "e-recruiting practices without human intervention".
IANAL, but I imagine this would cover the OP's situation if they were EU-based, and most similar complaints (Apple unilaterally pulled my app down incorrectly and won't answer my emails, Google wrongly closed my Gmail account). If a lawsuits & penalties start appearing under that umbrella, it might help change the monopolies positions.
Doesn't help in the US of course, but hopefully they'll follow suit in the coming years, which would apply quite a bit more pressure here.
For now...
It probably *is* 99% accurate - there is a lot of fraudulent activity online.
That 1% false positive rate really hurts though, destroying people's livelihoods.
I think that people developing machine learning models sometimes don't think about the consequences of the false positives enough.
There is a great section in David Spiegelhalter's book The Art of Statistics which got me thinking about this.
def is_fraud(x):
return False
You'd get an email, "Hello, there might be a problem, here are the details" and if you made some good faith attempt to reply, be able to talk to an actual person, who could potentially escalate your issue.
That said, in many other applications of AI, a 1% error rate is probably a huge success. AI does not need to be perfect to be considered successful, it only needs to do better than a human. In this example, AI failed miserably at that, but there are plenty of applications where AI errs far less frequently than humans do at the same task.
Also, every time I've had an issue with an amazon order that I've needed a human to help with, I've been able to get one on the phone or via chat.
I suppose there are some Amazon sellers who have specific items that (contrary to the public listing) actually contain drugs. In that case, people could order them as a way of buying drugs that would look like some other kind of retail business from the point of view of Amazon, shippers, and tax agencies. The seller could try to make the listed product extremely unappealing or uninteresting to the general public, or have some kind of code or out-of-band contact method that people could use to indicate that they wanted and expected specific orders to contain the drugs instead of the listed item.
Also
Also, it looks like the "people lying in order to harm, steal from, or extort others" issue hasn't been solved for any side of a two-sided market or review system. :-(
Edit: I was first thinking of saying that this is a way that drug prohibition really empowers random people to hurt each other with false accusations. But this is probably true for any kind of high-stakes accusation. As long as there's any kind of contraband or dangerous item that Amazon cares quite a bit about banning from its site, and as long as there's a significant market in that item, people could make spurious accusations that any seller was secretly trading in it.
I know our salespeople had constant troubles with the U.S. customs because we always had to declare whether our devices were medical equipment or not and different TSA agents had different opinions on how it should have been filled in. They often got upset regardless of how the form was filled in.
Legally speaking it is not a medical device, but some agents disputed that because it was used in the medical field at hospitals. Other agents were upset for wasting their time if you did declare it as medical equipment.
The real TSA agents however, do appear to care. And I'd also argue they're also quite competent. But Snowden was indeed correct that 95% of the people you deal with at the "federal government" are contractors - of varying quality.
What you’re describing is something we run into with the US specifically quite often. It’s kinda crazy, but the US has really become a stagnant market; nationalistic economically, and difficult to do business with unless someone gets to take a big piece of the slice, and very conservative when it comes to innovation.
I would really like to talk more if you’re up for it.
But other important Amazon marketplaces just won't respond, or don't follow up. I Had the UK tell me if I got a VAT they would enable my account; I did the months of work required to get VAT registered and then responded to UK support that I was ready to go, and have had all of my emails (~1 a month) ignored for over a year now. Why tell me to get VAT and then just pretend I don't exist? They just don't care. Even worse Amazon will INVITE me to join certain marketplaces, which I have to pay to join, to only later then find that I'm randomly blocked in that marketplace as well. Oh yeah, no refunds either.
I've escalated to jeff@amazon and had multiple people there tell me they're in the middle of unblocking the account everywhere, and weeks go by and nothing happens, and then they too just stop responding my emails. At this point jeff@amazon doesn't respond anymore either, even though I made certain to not repeat my request more frequently than once a month. It is like hair-rippingly frustrating and it is the biggest stress point for my business currently. The worst part is this is my main source of income so I'm just relying on their support people who are wholly unaccountable.
I wish I had options as to otherwise but Amazon is such a behemoth now that most of the customers who were purchasing through my own ecommerce store a few years ago now demand to purchase on Amazon or not buy at all. I'm losing a ton of money because of this and have absolutely no way to get help. I cannot even use words to convey how frustrating it has been..
Some people seem to think no web market is better and we should go back to brick and mortar mega corps like Target and Walmart, but good luck getting shelf space there.
I ordered a book, they sent the wrong book. They argued over whether it was the right book (they had different isbn), they made me buy an envelope to ship it back and only refunded my money once they received it and cleared it.
So I was out the price of the book for three weeks.
Old Amazon wouldn’t have screwed up, would have mailed out a replacement book immediately, etc.
They seem to be stupid in handling exchanges and cheap now.
This doesn't seem obvious at all. It could be any number of other malicious actors. Off the top of my head: an unsatisfied customer, a troll, a disgruntled former employee, etc...
Put another way: if a single unsatisfied customer is able to take down a product then there must be dozens or hundreds of profit motivated take downs. The article is not naming an exact competitor, but dismissing the accusation as "it could be an entirely different random person" does not change the issue's substance: Amazon is broken.
Not sure what the intent of this observation is - the same is just as true: 'if a single competitor is able to take down a product then there must be dozens or hundreds of profit motivated take downs'.
> The article is not naming an exact competitor, but dismissing the accusation as "it could be an entirely different random person" does not change the issue's substance: Amazon is broken.
Considering there are only so many no-buckle belt vendors, stating that it obviously a competitor casts aspersions on these other vendors without any evidence. But no, it indeed does not change the substantive critique of Amazon.
Big tech companies are full of people acting rationally, seeking to maximise metrics that are easy to track and sound good, but have negative externalities that are hard to see or to measure.
It is a bit dystopian, but you can already see some of the adaptations, such as the very existence of SEO and its significant use by businesses.
- Seller is banned by an AI agent for arbitrary reasons. There is no reasonable phone recourse
- Seller tries to make a _human_ point in a blog post hoping it goes viral
- The blog post is noticed on different platforms and buzzword of AI wrong doing is amplified to a certain level
- If some arbitrary threshold of _human_ reaction is noticed on the social media, the company in the wrong will probably do something about its AI bot going haywire
Thus, the bot's common sense holes will be covered by collective wisdom of the crowds (which maybe, just maybe, will be integrated in the bot's training loop).
If no human reply is given, it starts to exponentially pile up the fine, until the issue is solved.
They shouldn't be able to allow people to make their livelihoods on their platforms and suddenly cut them off without even giving proper support. They should pay for this - innocent people go through an immense amount of stress and get stripped off their money and the product of their work.
If the answer is: "they're too big to have humans manage this", then you know what the reply is - cut them into pieces, half, fourths, who cares. Slap a user account cap on that shit with a number that a human workforce can handle.
I feel no sympathy for growth problems of this massive companies. I really hope European Union starts to crack on this shit, and hard.
For now I think that voting with your feet is still a recourse, complemented by vocal messages on social media to maximize hit-back.
The idea of the small claims court is that you don't use a lawyer, so there are no complicated forms to fill in, and if it goes to court you just present your evidence in front of a magistrate in plain English with no legal mumbo jumbo.
Starting a case in the small claims is usually a splendid way to force your way past the army of bots and AI customer service shields, and get your complaint seen by a real person. It usually gets things resolved very quickly. But not always... here's a good story:
Last year I was overcharged £6 by Opodo on the credit card surcharge for a flight booking. I spent two hours going round in circles with their Facebook Messenger automated customer service bot. There were no other contact methods given. So I took them to court, claiming:
Claim: £6.23
My time interacting with chat bot (2hr 22min) @ £100/hr: £236.67
My time preparing claim (1hr) @ £100/hr: £100.00
Court fee: £35
Total: £377.90
... which I think is extremely reasonable of me. They didn't file a defense, so I won the case by default. Even after winning the case, I still had no method of contacting them, and no payment was forthcoming, so I went to the next step which is to file for a warrant.
This cost an extra £95 but it included a bailiff visiting their office, who would be able to take goods to the value of my claim. Bring it on, I say! So I apply for my warrant and wait.
Nothing happens for 6 weeks (this is in the depths of the first lockdown so I wasn't surprised by the radio silence).
Then I get a call from a very helpful chap who tells me he's the bailiff. He says I probably shouldn't hold my breath waiting for this money because the registered office is just a plaque on the fall with no staff or office... and he has FIFTEEN warrants outstanding for Opodo.
The next step would be a warrant against Opodo's bank which would allow me to freeze their account... but there was no easy way to do this and would probably require a lot of paperwork. So I gave up. My belief in the UK criminal justice system shattered.
Then about a month later I got an email from Opodo:
"We are contacting you from Opodo after receiving the Judgement for Claimant. In order to proceed with your refund of £377,90, ..."
... and within a couple of weeks they had paid me in full! I'd highly recommend the small claims court - it takes literally five minutes to file a claim here... check out: https://www.moneyclaim.gov.uk/
The whole "can't get legal fees" is moot in the US since awarding of legal fees to the winner is extremely rare.
* methoxetamine, a synthetic hallucinogen
* jimson weed, a literal weed from the Americas that causes severe hallucinations and paralysis
* coca leaves, a traditional Andean herbal stimulant that are the natural source of cocaine
None of these items are related or similar at all, and none of them would be used in making belts. Seems like a difficult error to make through mislabeling!
There are already categories of items I don't ever try to buy on Amazon: farm equipment, fencing supplies, pet food.
And I'm currently looking into email solutions that won't leave me completely screwed when Google arbitrarily decides it's time to delete my gmail account. There's just too much associated with my email address for it to be in the hands of someone who won't answer the phone.
IMO, these companies are winning because they're better than what came before them, but they're far from the perfect solution.
It seems that it's more important for Amazon to have the lowest prices rather than being a trusted reseller.
(It's so bad that I've stopped buying from Amazon if possible. I'm just too worried that I end up with a counterfeit product.)
p.s. Can’t wait to try the drugs.
This forces me to use smaller (often local) alternatives, and honestly I don't miss anything.
So my 2ct: Stop complaining, instead do something. Stop using Amazon, Google, Ebay, Facebook and the likes.
YOU are the one that makes them unavoidable, so just stop doing so.
This sounds an awful lot like the experience of the Terraria developer banned from Google earlier this week.
And whats always shocking every time I need to contact a large company via some fast response required, its never available.. Phone calls are passed around department to dept that cant deal with significant situations, email support takes far too long, live chat rarely exists.
It's mind blowing how such large companies don't have these services but at the end of the day it's a numbers game for them.
There just are more people that don't complain and their bottom like is still fine.
Really really frustrating as there is no even recommended customer service requirements for companies. I'm sure in the above case there is money lost for this action by the competitor. Will the OP be reimbursed ? Of course not.
Classic Amazon, competitor most likely has financial link to Amazon, even might be their puppet company.
It is unfortunate that these trust based systems are being abused.
Most tech companies are run with as few employees as possible, that is why SV FAANG engineers make so much.
We can look to china to understand a bit more of the statistics and personnel required to effectively moderate a system. e.g. number of censors per number of users.
This means that companies like facebook have 2-3 orders of magnitude fewer employees needed to moderate their system. i.e. they are effectively unmoderated. (this is why it was no surprise that facebook was used to plan the capital riots, even though FAANG benefited by deplatforming parlar for political reasons).
We are in an interesting situation where there are probably similar levels of legitimate and illegitimate takedowns but nothing will change because the people with power and money have what they need.