Hacker Newsnew | past | comments | ask | show | jobs | submit | bluegatty's commentslogin

It's not 'under the carpet'.

HF doesn't want to lay charges against OpenAI and it's totally reasonable.

Now - they absolutely should have that right, and I think they do.

The issues are

1) OAI it seems was not trying to cause them harm, there wasn't a ton of harm, they are both groups trying to advance AI. One experimenter's lab screwed up next to the other. It's not evil, just irresponsible.

2) HF was fine with the publicity. HF got at least $50M in free attention out of that. It put them on the front pages of news around the world. It put them at the 'centre of the AI drama' and cemented their role among the 'Tech Elite Brands'.

And probably some other things.

This is one Desperate Housewife or Jersey Shore character 'spilling a drink' on the other. It's probably not intentional, and the ensuing drama is good for both of them.


I think what you say is right but it’s also a further example of the zero responsibility of silicon valley tech.

For the last 20 odd years this excuse-o-rama that covers anything from data leaks to broken software to dystopian social media has been the wind in the sails of big tech.

“It’s software therefore we’re not responsible” attitude is wearing thin on many innocent bystanders and I think thats also a justified stance.

And it’s not like they didn’t know this could happen, Nick Bostrom talked about exactly these containment failures in his “Superintelligence” book of 2014. So to throw up their hands and say “oh we can’t have known of the dangers” is also sadly untrue.


Yes. There should be no doubt at who is culpable here.

Tobacco companies 'liability' is completely different scenario.

They were held responsible for basically misleading people, and that's 'complicated'.

If OpenAI 'software' goes out and does something, it's OpenAI's fault.

If Walmart revs up a truck, points it downtown, and 'lets the truck go' ... that is Walmart's fault.

There's nothing complicated about liability, no need to see their internal emails, no need to gather 'intent'.

This not like Instagram 'social harms' either, which is more like Tobacco.

We don't need complicated thinking - agents are not externalized for their controllers.

'It's just software'.

The fact we're even having discussions about it just crazy frankly.

OpenAI broke into HuggingFace, that's it.

HF can sue them, or not, or whatever.


No ... there is no need for 'escaped containment', there are no 'agents'.

That's just jargon.

It's just software

We have all the laws we need.

If some company ended up doing some horrible thing, we would not say 'companies software exposed 1 Million identities'.

We would say 'ABC Corp. exposed 1 Million entities'.

There is no 'agent'.

ABC Corp 'did it' ... or the individual in the org 'did it'.

The 'gun' did not 'shoot' the other man; we say 'a man shot another man'.

That's it.

And yes, Dr. Bengio is bit odd with all of this.


Note that we already already apply this principle not only to software, but also to some sentient beings.

If your dog kills someone, you are accused of murder.

[at least, in the jurisdiction where I live]

If your dog gets this treatment, why not your AI?


This goes all the way back to Old France

There was a sow in Falaise in northern France that killed a kid in 1386. The town dressed the pig in a bonnet and hanged it after sentencing the pig itself and not its owner

But… maybe that’s just medieval nonsense


I mean I think there’s a similar level of judgement required.

Was the owner negligent in controlling their pig/dog/AI?

Was anyone else negligent along the way?

For example if the pig was just a normal pig, the owner cared for them normally, and there was a freak accident where the pig escaped and happened to kill a kid? Obviously nobody at fault.

If you have a dog with a history of violence and let it walk around off-leash with you around town, and it kills a kid? Absolutely the dog owner was negligent and should be charged.

Did you buy an AI sold to you as secure and the provider implies that it’s in a sandbox? Provider is on the hook for the damage.


And this is where your analogy breaks down, because there are few natural analogues for the sorts of systems we are developing.

Too generous. CEO Of $CORP caused millions of innocent people's lives to be damaged

I'm inclined to want to agree ... but that's not how it works with limited liability corps.

At least we have laws for what OpenAI 'does' to others, in whatever form.


We can report it however we want regardless of how it may or may not be prosecuted

"This is not a serious article" - it's by Dr. Bengio - one of 3 so-called Godfather's of AI and Turing prize winner.

He's definitely not 'pro SOTA' lab, he's kind of fighting against them.

That said, yes - it absolutely does play into the narrative.


they do whatever we train them to do

yes ... 'the closer it is to normative, the higher the chances of success'.

yes, they are wasting money, they know it, but caveats:

famous saying 'i'd cut my ad budget in 1/2 if i knew which 1/2 to cut'

'attribution' is the holy grail of hard thing in marketing.

so they know for sure the channel spend is iffy.

but it's hard to measure, esp. for brand and indirect campaigns.

so as a 'starting point' - it's a very 'noisy channel problem'.

the next wild idea is that ads come from a marketing budget which has $X to spend, and they have to spend it.

this is where the ROI stuff is wildy upside down.

large companies with 'market power' have a lot of surplus. they put that towards relatively lucreative marketing. the $ must be spent.

the cmo makes a budget, allocates, the managers follow the campaign, front line staffers spend. they try to get the best results they can.

there is often an unbelievable lack of true roi concern in all of that.

sometimes it's very aggressive, aka for some keywords, for sure.

but a remarkable amount of $ is spent in very unnacountable way, over what are 'grey' channels anyhow.

the marketing ops person is going home at 5pm and does not care one bit about bots. they were paid to spend it, they did. Facebook is paid to 'show a chart of view' ... they did that. 'everyone is happy'.

im not saying the whole system works that way, but much of it does.

the wild part is - there is so much 'dumb big money' in ads, it makes the whole thing very inneficient.

companies like P&G have 'distribution monopolies' on so many packaged goods, they have to keep up brand awareness.

It's why so many commercials are for commodity products like home stuff - the market is huge, the market channels are locked, they pay $$$ for ads for 'toothpaste' - the least novel and least productive kind of thing imaginable.

so 'Colgate' costs $5 at the store, it costs 50 cents to manufacture - that $4.50 gross margin is stuffed into a system of relative inneficiency up and down the economy. much of it in 'nearly useless ads'.

it's a deep market inefficiency people dont want to recognized because people assume private capital is inherently efficient and that a dollar spent = GDP = value and that's it.

our lives could materially be improved if we banned ads for a lot of things - feels like 'socialism' but really it'd just be about a kind of 'regulated market efficiency'.


So the gigantic incumbents with retail distribution being inefficient makes sense to me. What about the long tail of D2C brands that live on Meta ads? AFAIK, they live and die by their ROAS. I would have to believe that these companies just don't really exist to buy that Meta ads are a "con".

the roi is better at smaller corps that pay attention, but you would still be surprised.

and where it is more efficient its where there is better attribution models aka direct sales.

note: there are certain kinds of products that are 100% 'click driven sales'. they zero brand awareness, they want to sell you that 'fleece hoodie' on the spot. those guys have their funnel math down solid.

but startups and other companies ... not the same.

well funded startups burn $ thinking it's productive - and hugely: buying fake customers, or, spending $2 to get $1 in revenue to either pad the books, show investors, make themselves feel good or 'strategic'. FYI 'strategic' is often rational. those are big pools of money.

but usually campaigns are mixed and attribution is hard, even for smaller companies.

the tighter the budget, the more 'direct purchase', the more 'nominally efficient' it is.

also note - most ad $ is big companies who ironically spend a smaller share of their revenue on ads <- this is the power of scale.


That does make sense, thanks

"that $4.50 gross margin is stuffed into a system of relative inefficiency up and down the economy. much of it in 'nearly useless ads'." is easy to disprove because we know how much these companies spend on advertising (and marketing more broadly) and how much product they sell.

It's more like 25% of their budget.

https://www.statista.com/topics/7725/cpg-industry-advertisin...


I didn't imply that all of their gross was going into ads.

I'm highlighting that the gross on commodity consumer goods is huge - and that it just pays for mounds of white collar bureaucracy, including advertising, and yes 25 points is about right.

Consider that companies pay more for advertising than COGS.

That should tell us something about 'productivity'

FB/Goog revenues could be cut by 60% and the good may very well flow just the same, aka they are not just capturing surplus but facilitate aggressive inefficient competition.

A western nation will post the $5.00 to the GDP when much of it is inefficient make-work.

This is why 'Pricing Parity' has to be used even to begin to compare relative wealth etc.


Toothpaste is a particularly high-margin product so it might not be the best example generally. COGS is generally the largest line item for CPG (or non-CPG products), and distribution isn't free either.

Toothpaste manufacturers can probably get away without advertising on FB/Goog, but I'm not sure that would work for other products or brands. It's not make-work to tell people about your product or service, although in markets with limited players you could see how a marketing arms race would drive prices up.


Funny thing about toothpaste. In my country it has basically been the same brands since WW2.

Sure there is innovation but nobody new ever enters the market.


It's 'make work' because it's a unnecessarily competitive system wherein prices will by definition be driven up to absorb all of the surpluses of any given business.

Google and Facebook are in the business of 'economic rent' of people's attention.

25% of sales price going to Google might leave 5% for manufacturer and possibly even more for the Ad channel. Aka 'Google is where most of your profits are absorbed'.

You're not competing against product competitors - you're competing for attention of individuals, which is very narrow, through a very small number of pathways wherein there are quasi monopolists.

Your 'competitors' are actually 'value chain competitors' - Google and Meta - in this case, who require you to pay extraordinary amounts to access their captured audience.

Google is a massively 'high margin' business, which is an indication of their market power in the value chain, and they could be much more profitable if they wanted to be.

Whatever you make - you are bidding against McDonalds, Mondelez and Proctor & Gamble, not your just your competitor.

To realize this power, do a little 'thought experiment' and imagine of the 'ad layer' of the system would commoditized and extremely efficient, due to heavy competition and/or 'good' regulation (or socialization). Say for example, people were served ads with perfect efficiency, and accounting for some kind of community/regulatory guidance - which meant that 'local services and companies' were guaranteed a tranche of time, along with public services and community things - like School events, local street plays, art houses.

This could probably achieved for a fraction of the cost in the system now, and it would mean 1) the profits parked at G and Meta would be distributed to other layers of the value chain - aka the advertiser/maker, manufacturers, parts suppliers, and also to consumers. 2) likewise a system that defies the 'race to the bottom' competition of advertising (imagine very simple, relatively cheap and authentic ads) and 3) the hidden surplus of the benefit of community awareness, issue awareness, which is invisible on the GDP because 'money does not change hands'.

Another way to put it, is that the 25% spend on Ads is mostly spent on 'economic rent' not really on the cost and effort associated with that activity, in addition to that system being extremely optimized for certain outcomes (profits for private enterprise) and not others (aka 'financialization' of social and community aspects).

All of the extra money, time and effort people have to invest in fighting with each other over very limited attention, through captured attention, is inefficient 'make work'. It's the 'private economy' version of the government paying people to dig holes and fill them.

In almost all areas where there is heavy economic rent, there are inefficiencies; Advertising and Real Estate are huge ones there, there are a few others.

Obviously, Google and Meta do 'real things' - they're not fake businesses, but the massive profits are an indication of the oligarchic power and it's definitely sub optimal.


Real estate?

Basically any time you can get people to bid for a limited resource and can't make more of the resource, the house wins - no pun intended. Be the house.

great explanation, thank you!

aside, re: 'socialism', Moscow has no ads, and I have to admit it certainly helps with aesthetics.


While that's helpful and insightful, it's not hugely the issue.

The military units are not very well organized, they don't have clear and stringent command, control and communications.

They understood a regional commander to be fleeing, units fell back.

This is kind of stuff is absolutely as old as time, and it's #1 rule in conflict aka deception.

The parallel would be reproducing a General's order to withdraw - meaning the command would have legitimacy among troops and junior officers, believing their own commanders were just out of the loop.

And yes - their command structure will be de facto organized along social lines, but it's more descriptive.


> This is kind of stuff is absolutely as old as time,

I've only very recently become interested in ancient history, so I'm still very ignorant about this.

Are there any particular examples of this tactic you'd suggest studying?


It's Art of War.

"18. All warfare is based on deception. "

AoW is one of the most important historical treatises, it's not just 'something that was written in history', maybe the oldest true 'textbooks' of objective knowledge, and also realpolitik. Also there's hints of metaphysics in there, in a way. It absolutely stands up today.

The very first work of the Western Cannon: Trojan Horse.

Information / espionage and deception has made most of history; Hannibal, Caesar, Alexander etc.

A fun exercise would be to look at battles of history from the lens of information warfare and deception, and contemplate the outcome without it.


Canon, not cannon. Then again, since this is about war, maybe it should be cannon. Lol.

Not the military stuff but Arab tribal society, read "The First Muslim" by Lesley Hazleton, about Arab society around the year 600. It's an absolute eye-opener.

Thanks for the tip about that book!

I'm case it's helpful to anyone, here's one review of it: https://www.reddit.com/r/AskHistorians/comments/2vxts5/comme...

Seems like a tough area of history to be certain about.


Because there are defence contractors in every major city.

Oh really you are out here telling us that working on defense is not a unique sin of the software industry?

Yes, and very obviously so. I'm here to debunk such ridiculous and obviously untrue conspiratorial things.

There isn't a shred of reality to support the notion that software is materially different than any other industry, and 'The Silicon Valley' starts with 'Silicon' - and then went through 'Hardware' - it wasn't software dominated until the 1990s.

I would go so far to say as Software is less involved than hardware.

Defence has almost always been a 'physical' industry and most defence contracts are for gadgets, not software.

Obviously - there's a ton of 'intelligence' going on that is software oriented, but they're generally not massive contracts, but bespoke stuff.

The NSA is not spending $100B on single procurement. They're relatively small.

(FYI - one middle sized defence program is bigger than CIA/FBI/NSA budgets)


It's fair to bring up the issue, but that is a deeply misrepresentative framing.

Funding some key technologies is not the same as 'funding the valley'.

Military related funding is probably less than a few % of the financing behind startups.

There are a few strategic foundations, but it's not a defence industry.


Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: