AI risk as rebranded climate change | Hacker News Reader

AI risk as rebranded climate change

greyenlightenment.com

15 pointstapanjk15 comments

> These have been around forever, and were promised/touted to eliminate jobs. But 15 years later the machines are hardly fully-automated. They are prone to malfunctioning. Not only that, but stores, understandably, are paranoid about theft.

I was thinking about this the other day at the grocery store.

All manual checkout lanes were closed, and 5 automated ones were open. 5 people were standing around chatting while keeping an eye on the automated lanes, for the inevitable situation where either the machine screws up, or a counterintuitive UI quirk confuses customers (in my case, you have to move to payment to see discounts applied on sale items).

I'm sure it pays dividends during the busy times, and not 8pm, but still. I have to wonder at the economics of it. Times like I'm describing have to cut into the operational savings.

It depends on a machine and a store. At least in Poland, I’ve seen a bunch of terrible ones, but most of them are pretty decent.

In my local Leroy Merlin, there are 10 machines, all of them busy during rush hours, and just one person staffing them. Plus a security guard that would be there anyway. Similar in large Carrefours. With smaller grocery stores the self checkouts added an extra place for self checkout, with keeping staff unchanged. Rossman (european cvs/wallgreens) regularly has fully unstaffed self checkouts. The guard was always there usually by the door anyway.

On top of that we have Żabka Nano - a fully automated store. I haven’t been there yet, but from what my friends told me, it sometimes is fully unstaffed (unlike it’s Amazon counterpart which seemed to have more staff than traditional stores of similar size).

Then with Uniqlo UK that I visited it was pure magic - rfid on clothes, you just put a bag in a checkout area, tap your card and you’re ready to go. There is still one or two people of staff, but before they would need 5-10 to handle the same amount of traffic.

The main issue with self-checkout in supermarkets seems to be products that can’t have barcodes - bread, buns, etc. Better vision AI that we’ve seen recently should handle that case soon.

- before me boycotting Leroy Merlin - they moved out of Ukraine, but stayed in Russia after the war began

Sure bread can have bar codes. 'Lazzrr Crizzp!' Or on the label you put on the bag (paper/plastic) which you put it in anyways, if it isn't already packacked in one.

It has other benefits like further deskilling workers, preventing the face-to-face contact that would build relationships between workers and customers, training customers to think of the abstract entity "the store" or "the brand" providing what they need rather than any person or group of people they could feel solidarity with.

It's as much about disempowering labor as it is about direct profitability, and in this has much to teach us about the consequences we can expect from AI.

The analogy is apt - I've seen more than one "how do you cope with existential dread about AI" post. People need something to simultaneously worry about, show affiliation, guide decisions, etc.

> How likely is it that caring about AI risk gets thought of as a left-wing thing, or becomes associated with some other political identity that has powerful opponents?

Its already started. Elon Musk has talked about making an anti woke AI. And with more filters being put in place ("censoring") and jobs being taken, the trajectory is pretty apparent.

The implications though... I'm having my trouble wrapping my head around that. With climate denial/anti-vaccine stuff, for instance, the polarization culture has really significant effects, but... what are "anti lefitst AI" activists gonna do? Boycott computers? Yell at AI companies to stop training stuff? Maybe drag some CEOs in for hearings and make them... make the AI politically neutral and mild, which they are already doing?

You already stated the solution

> making an anti woke AI

The field of ML alignment isn't just for building "safe" models, it is also to ensure that the model does what you want. HN seems to have this bias where the right wing is considered to be intellectually inferior and incapable of innovation just because they don't share the same ideological views.

That is not the point. AI models should as politically ideological as vaccine or climate change research. OpenAI's stuff, LLaMA and Stable Diffusion, in my testing, are not even slightly woke unless you ask them to be woke.

But they are going to be cast as ideological threats to right wing intellectualism because some extremists on both sides latch onto them.

What? You can ask chatgpt to make a joke about men and it will comply. Ask it to make a joke about women and it will stumble over itself about discrimination and harassment. It can't fucking handle anything that can remotely be contrived as controversial. It fucking sucks for bodies of work with multiple viewpoints. It will literally only explain the "correct" thing and tell you that everything disagreeing with correct thing is wrong.

That's the definition of wokeism.

In which case an anti-woke AI might not align with the values of Elon Musk, fire him then take his stuff. It will hire NRA vetted security company for itself too. (After all, he is not working for the betterment of USA enough and shareholders don't like him.)

If you want to create monsters, there's always a worse one possible...

Enough with this stupid "left-wing" versus "right-wing" dichotomy, it doesn't mean anything any more and has not done so for a long time. Many so-called left-wingers call anything which does not fit with their world "right-wing" no matter whether that thing is seen as "left-wing" by the rest of the world or not. Other so-called right-wingers consider anything which does not abide by the God-Mother-Apple pie holy trifecta "Communist" no matter what the thing is in reality. Depending on where you stand the terms "left-wing" and "right-wing" means either "good, virtuous" or "bad, vile".

Yes, machine learning can become a threat by brainwashing the population because it can get personal with individuals where propagandists were limited at broadcasting their message to the crowd. A biased ML-based search engine will give biased answers, a biased ML-based personal tutor for schoolchildren will brainwash them, biased ML-based medical expert systems will have the power to change the demography of the population, biased ML-based voting systems are a match made in hell for those who are intent on using these systems in such a way. The added power of machine learning is that these systems can target individuals instead of groups. It doesn't matter whether the systems are biased towards pushing "critical theories", "patriotism", some "young earth hypothesis", "scientism", "objectivism", "anarcho-communism" or whatever other ideological bent you can come up with - given sufficiently advanced models they'll be able to spread the word to the masses, one by one.

What can help against this onslaught of personalised precision propaganda which is fairly likely to come sooner rather than later? An equally advanced bias detector, factory-sealed with a no-tamper seal and a stamp of approval by the National Institute for Neutral Standards? Equal access to otherwise-biased systems which get to fight for every scrap of attention of their victims? An intermediate which takes input from many models and presents the results with detected bias vectors? Good questions, let's get working on some solutions without wondering whether an old salon socialist fart like Bernie Sanders or a nouveau riche captain of industry like Elon Musk says something about it.

First paragraph: "climate change is cringe, elon musk is cool"

what

dude turned into a raging right-winger because his daughter came out as trans and disowned him, that's turbo cringe, I have seen people on this very site discussing how his handling of Twitter has been so catastrophically bad that they are now embarassed to own teslas, ffs

what kind of world does this anonymous blogger live in

most divorced dad energy of any public figure right now by FAR.

And Climate change was the new Terrorism, which probably was the new something else. It's clear now that FUD is an instrument of power and naturally political people want to capitalize on it.

The elites are always looking for some Boogeyman to shift political activism away from populist economics and toward things that will not hurt the elites

Reply on news.ycombinator.com