40 comments

[ 0.24 ms ] story [ 79.9 ms ] thread
Do we really have to call anything that touches data science a bit AI.. that's almost just clickbait.
I'm starting to feel like Inigo Montoya every time I see/hear the term AI...
Yeah!! "You keep using that acronym..."
The media does a terrible job of describing AI/ML.

My roommate was under the impression that Facebook shut down their experiment wherein chatbots communicated with each other because they were scared of the consequences of the machines communicating in their own language.

For the unaware, Facebook determined it wasn’t useful to continue the exercise.

This isn't terribly hard to achieve. E.g with a pretrained LSTM generator downloaded from some place that does neural network stuff + a homebrew text classifier (I'm fond of character-level online linear models with long n-gram ranges) that's just underpowered enough to overspecialize to a certain style, but unpredictably so. Make this poor man's model generative by choosing random combinations of characters that hit a high enough score; and then feed it to the LSTM (along with normal text samples).

The underpowered generator creates weird stuff that the good generator absorbs into its language model. Boom, you are witnessing the technological singularity. Or something.

What is the difference from any other marketing efforts to sell us buying expensive stuff we really don't need?
“Revealed” is really clickbait titling. Reading the article, it has very little to do with AI, and nothing novel.
> “I’ve often heard people wonder about how they are targeted so accurately and it’s no wonder because its all hidden in the small print.”

> Publicly, gambling executives boast of increasingly sophisticated advertising keeping people betting, while privately conceding that some are more susceptible to gambling addiction when bombarded with these type of bespoke ads and incentives.

Meanwhile, in the real world, hardly anyone clicks on a banner ad ever. A perfectly polished targeted turd is still a turd.

> in the real world, hardly anyone clicks on a banner ad ever

In our world. Let’s be honest, there has to be a category of people that still falls for them, otherwise people wouldn’t keep investing billions into this bullshit industry.

The tech-illiterate are still very much prey for ads, I’ve had first hand evidence of this a few days ago from my flatmate (the typical idiot who “logs in with Facebook” everywhere, has 50k+ unread email on a Yahoo address and reuses the same password everywhere) when she fell for some weight loss pill scam ad on Facebook. I had to help her chargeback the transactions with her credit card provider.

> In our world. Let’s be honest, there has to be a category of people that still falls for them, otherwise people wouldn’t keep investing billions into this bullshit industry.

Not just in "our world." The average CTR is 0.06%, and that's from Google's figures [1], 60% of banner ad clicks are accidental, 54% of Internet users have never clicked one. 25% use ad blockers.

They persist because they're a) cheap, b) because everyone else does it, and c) because they may, tangentially, have some brand uplift though no-one can actually prove it. Some desperately dastardly mindfuck scheme they are not.

[1] http://www.richmediagallery.com/learn/benchmarks

The exact reason why google will, for years, happily ignore humongus ad-click-bot-farms.
I imagine that my CTR must be much higher than average, because I will click on any ads for brands that I dislike. The idea being that I am transferring money from the advertiser to the site that I am visiting.
Agree - also if you've ever used a gambling app, they will push notifications to you right before your team is about to play or a horse race is about to start because they know you're likely to bet on that if you remember.

Also banner ads are for gambling are more for impressions than clickthroughs, if you watch a football (soccer) game in the UK, the ad breaks are full to the brim with "BET IN PLAY, NOW" "BET BET BET" "NOOOOOOOOOOOWWWWWWWWWWW!" type ads which just put in the back of your mind. For a gambling addict it must be near impossible to resist.

It’s hard for alcoholics in early recovery to watch football because of the pervasive, nonstop ads for alcohol. Add to this that most of the alcohol industry is propped up by the top 10% of drinkers who average over 10 drinks a day, and you have an industry built around taking advantage of people who clearly need help.
For the benefit of American readers, I'll link to a few of these adverts below. It's worth emphasising the fact that ads of this sort are in every break of every match. The disembodied head of Ray Winstone bellowing in-play odds has become a cultural fixture. I'm a gambler myself, but I'm troubled by the ubiquity and sheer aggression of gambling advertising.

https://youtu.be/l4INTKhOLSs?t=20s

I’ve never clicked on a single billboard, but that hasn’t stopped anyone from destroying the view.
That doesn't mean the ads aren't working.
> otherwise people wouldn’t keep investing billions into this bullshit industry

There are plenty of reasons why this money might be spend without clear evidence of a ROI:

- marketing departments are under pressure to increase sales in a world where the traditional outlets (TV, newspapers, ...) loose relevance

- internet ad campaigns look convincing with all sorts of nice numbers, but unfortunately many of those numbers simply don’t mean what one might think they mean (fake clicks, accidental clicks, ...)

- the corporate world is slow. Which head of department would voluntarily suggest to reduce their budget? The higher ups are too far away to really challenge this stuff

>Meanwhile, in the real world, hardly anyone clicks on a banner ad ever. A perfectly polished targeted turd is still a turd.

Don't necessarily need to click on the ad. If a user sees an ad just once it's in their mind, somewhere. TV car ads are another case where they don't expect people to buy stuff based on the ad, because who in their right mind buys a new car based solely on an advert? The idea is to get the product in users' heads.

I've definitely been guilty of seeing products in banner ads then looking them up later, or letting them influence my future purchases almost unwittingly.

> Meanwhile, in the real world, hardly anyone clicks on a banner ad ever.

Google has about 1 trillion dollars per year in revenue that says otherwise. Online advertising became so huge because it works.

> Google has about 1 trillion dollars per year in revenue that says otherwise. Online advertising became so huge because it works.

That's a non sequitur. Online advertising could be huge because of an irrational belief that it works. You conclusion needs much more support.

Suggesting that people and companies are independently and collectively spending 1 trillion dollars for zero results without any form of verification is absurd.

But, if you really want to go there. Plenty of companies only advertise online and 'magically' end up making money selling real products to real people.

> Suggesting that people and companies are independently and collectively spending 1 trillion dollars for zero results without any form of verification is absurd.

> But, if you really want to go there. Plenty of companies only advertise online and 'magically' end up making money selling real products to real people.

Then quit the hand-waving and provide convincing proof. Also address the issue of bot-farms and other kinds of fraud that could cloud the metrics.

Do you know how Adsense works?

You pay for the click. You can then track them through your conversion process. You can trivially see if Adsense is making you money or not.

Why do you think they are just CPM banner ads and speak with such authority on them here?

Sometimes I wonder is there new research in the world of advertising and AI?

Tracking clicks via heatmaps, pageviews etc has been done for ages. So, has AI/ML provided better ways to convert those ads?

There are improved models for segmenting customers, predicting next best actions etc. It's also a lot easier to run them now (e.g. handling terabytes of data isn't difficult, petabytes is increasingly common)

So it's improved a bit but not much more. The people selling this stuff are marketeers selling to other marketeers, so the sales pitches are getting better and better for sure.

“Every click is scrutinised in order to optimise profit, not to enhance a user’s experience.”

So identical to every modern digital business then....

I’ve seen user experience degrade across the board on apps over the last 5 years for what I assume is precisely this reason.

These are not the robots we were promised.

In a couple of weeks: "Revealed: how amazon uses AI to keep shoppers shopping" and then "Revealed: how pornhub uses AI to keep ..."
Supercharging peoples addiction to their phones, an even bigger dopamine hit if there is a chance you might win money too.

Given the amount of time it has taken to tackle the outdated FOBT machine legislation in the UK I can't see the gambling companies being slowed anytime soon.

AI is used to predict XYZ. Well, given the short sentence describing ML is "how to make predictions from data", this is really not surprising. Now, what is the success rate is the only question that matters.
the gambling industry pioneered a lot of the techniques now used for web sites, far earlier.

“time on device” is something they coined and started optimizing for decades ago.

I'm aware of one gambling company which looks up customer addresses on Google Street View to see in what kind of place they live. If you look wealthy they will send the big guns after you.
You can actually take most of the bullshit out of an "AI!!!!" headline by simply replacing "AI" with "computers".

"Artificial intelligence is being used to predict gambler behaviour" --> "computers are being used to predict gambler behaviour"

I think it's a lot more reflective of what's actually going on and removes a lot of the clickbaity hype...

I think it's more accurate if you replace it with "statistical analysis" instead. The computers are just computing very complex equations with your data as an input, but technically you could try doing it by hand.
That was actually how my stats prof made us learn all of the analyses, she thought if you can't do it by hand you won't understand what the computer is doing either. I found that to be really useful for just about every statistical technique I have used.
From the most recent a16z podcast featuring Ray Dalio, "Machine learning is merely finding the statistically best use (conclusion?) of the data."
Since it seems unlikely that governments will ban the use of data by companies for certain purposes like this or pay citizens a universal income for simply providing data to companies online, at least in the near-term, this seem like a ripe opportunity for more "counter-AI" and/or privacy apps.

I love using Ghostery to provide privacy. Why aren't there more consumer-facing apps that provide privacy or indicate when and how consumers may be manipulated by algorithms? Say you paid (nothing's free) an app to monitor your internet activity, allowing it to track and indicate how your personal habits may be exposing you to bias/bad decisions? I think this could help provide the average consumer with a sanity check. Knowing that you're being manipulated and someone is making a profit at your expense probably is one of the biggest motivators to alter human behavior, at least in Western society.

But hey, I still refuse to get Amazon Prime, so what do I know about the average consumer behavior.