Hacker News Reader

Comment by beachy | Hacker News Reader

Google's SEO malarkey created the intense vacuum which AI will now happily rush to fill.

The article mentions recipe articles, and how these now invariably have a waffly preamble designed to please the SEO algorithms. It's been obvious to humans how stupid this is for years, just as it's obvious that sites that simply copy from stack overflow are leaches on our attention and should not be highly ranked.

Yet Google hasn't dealt with any of these problems, and with a focus on short term profits, has consistently preferred to take the money from ads hosted on these crappy sites instead of delivering a quality search experience.

Now the chickens are coming home to roost for Google, which can only benefit a range of competitors, from Kagi to OpenAI.

View on HN

I think it's incorrect to say that AI has come to where it is today because of Google. AI researchers would have done exactly the same.

One valid question, though, is why technologists don't think about the consequences of what they do before they do it? Because they focus on short term profits, like you mentioned for Google?

One valid question, though, is why technologists don't think about the consequences of what they do before they do it?

Because they can’t. There’s no way Larry and Sergei could’ve foreseen the multi billion dollar SEO industry back when they were at Stanford. Every decision they’ve made since then has been of the form “stay on the gravy train or jump off?”

I don’t think anyone knows how to solve the problem of adversarial inputs in either the search space or the ML space. It’s a chicken and egg problem: you’re trying to build intelligence but the task of filtering out the adversarial inputs requires intelligence in the first place.

> why technologists don't think about the consequences of what they do before they do it?

“Prediction is very difficult, especially about the future.” ― Niels Bohr

More pertinently, people have been arguing over the consequences of increasingly performant automation at least since Ada Lovelace was thinking about using something like the Difference Engine for music and Joseph Marie Jacquard was using punched cards for complex weaving.

The 200 years between them and us has given us radical, and for most people incomprehensible, growth. Back then, famine was common even with 90-95% of the population working in agriculture; now, obesity is the bigger problem with 1-5% of the population working in agriculture.

We knew about CO2 being a greenhouse gas a century ago, but we thought it would take a millennium and be a good thing; CFCs were chosen as a refrigerant because they were un-reactive; cars were a solution to all the horse manure on the streets; asbestos the solution to all the fires; radium paint was used on watch dials, nuclear explosives were seriously suggested as a way to widen (and also separately to replace) the Panama Canal.

It's just really hard to see all the consequences of things, even with the best will in the world — not that this completely contradicts what you say, as there's also the other famous quote:

“It is difficult to get a man to understand something, when his salary depends on his not understanding it.” ― Upton Sinclair

A note. There is a sensible reason why, likely evolutionary derived, that we ignore "in the future" things so handily. Social reasons.

Let's presume I dislike my neighbour. He's not loud, just an asshat or some such, to the point that I'd love to punch him in the head. I don't, of course. We live in a modern society. Nor does he, either! The dislike flows mutual, deep and dark, yet we essentially ignore each other.

And even more astonishing? We go to sleep each night with a rival next door, sleep peacefully and in comfort, yet each of us with baser, animialistic instincts to beat on each other, perhaps to death, and yet? Even more astonishing, if the city walls were attacked? Hatred or not, we'd stand back to back and defend each other. Our families. Our friends, our city.

We are genetically predispositioned to live beside danger, and even completely ignore it, put it out of the way. The danger with the neighbour, a rival!! right next door, is ignored. And more so, the danger with those invaders one city over, with the wolf outside our cabin door, all of these things are put out of mind.

We'd go bananas if we had to focus on "the threat" constantly.

So we are designed to ignore even impending threats, let alone threads "decades or more in the future". If fact, I've often wondered if the more dire and dangerous the threat, yet with it not immediately on our doorstep, the more people might simply block it out!

That only mild threats are something that average human can deal with, for there is no need to push it from mind to sleep, to relax, to live without becoming a bundle of nerves. I wonder if there's some metric here, some "perfect level of threat", which ensures it is constantly in mind, yet sleep is possible, this returning the maximum human action on a distant (time or distance) threat.

All very good points!

I guess my frustration is that I can understand how fighting famine brought immediate, crucial benefits (even though it seems likely that the technological improvements that improved life may soon destroy civilization as we know it).

But the very first ideas that got in my mind where I learned about LLMs were black hat oriented. I actually needed to make an effort to see the potential benefits of LLMs (other than translators), but I immediately had many ideas of how that could go wrong (probably all of them being mentioned in the featured article).

In the case of LLMs, it seems a lot more direct than, say, the famous story of the Facebook "likes", that did sound like a good idea at first. LLMs sound like a bad idea (for society) from the beginning on.

> But the very first ideas that got in my mind where I learned about LLMs were black hat oriented. I actually needed to make an effort to see the potential benefits of LLMs (other than translators), but I immediately had many ideas of how that could go wrong (probably all of them being mentioned in the featured article).

I'm pleasantly surprised to learn this.

My experiences here since ChatGPT was launched have generally involved witnessing many of the louder people opining that LLMs "cannot" cause harm and that any example of harm anyone gives is "only" free speech. (The latter is something I find inherently weird as a talking point, given how important speech is).

I hope you're working in some kind of AI safety role; myself, I could only recognise the flaws when others showed them to me, I was not able to figure out these problems in advance.

> why technologists don't think about the consequences of what they do before they do it?

I would argue that incentives matter, and the system runs like a chemical reaction to conclusion.

The idea that Edenic internet of the pre-AJAX days was scalable, sustainable or can be somehow regained is a fine fable.

> Because they focus on short term profits

It takes the deep pockets of an Elon Musk, or a hardcore community like HN to do otherwise, no? Again, incentives matter, and economic gravity is hard to defy.

> why technologists don't think about the consequences of what they do before they do it?

They have bills to pay, and people are willing to pay for that. Technologists with strongly held opinions also exist, but they've been marginalized in every possible area where money gets you leverage. Just look at how open-source became a corporate breeding ground.

When you have a question like "Do N people have no ethics?", the answer is very often: some of N do, some of N don't. And the decisions to promote N1 and sack N2 are rarely made within the N group

> why technologists don't think about the consequences of what they do before they do it?

Imagine if they had opinions held so strongly they'd be willing to camp out in the CEOs office in protest. I'm sure that the executives would re-examine their positions on the subject and work with the technologists to address their concerns.

I will assume this was sarcasm.

So think about this: say your CEO asks you to make a missile specifically to target and kill children. Would you consider that it is not your job to question it, or would you actually think about it? I will assume that you would at least think about it. You would have some choices, then, for example:

- Accept it and keep doing your job

- Raise concerns internally (to your CEO) or externally (to the media), more or less vehemently

- Quit and go do something else

And I assume that for such an extreme case, all of this seems absolutely obvious to you. So let me ask: why does it feel like you can't imagine questioning something less extreme, like working on LLMs or autonomous cars or a connected fridge?

Engineers are paid to think (among others of course, but it is usually a necessary skill). It feels like they could reuse that skill and sometimes think about ethics, too.

It was sarcasm yes, but engineers are not paid to think. They are paid to do work (which may involve thinking) toward a given goal decided by the executives.

People are regularly faced with such situations (with both small and large stakes) and all three of those choices are taken by different people. We know that 1. happens because controversial stuff still gets made. 2. happens because we hear about it in the media. And 3. I assume happens as well. But in the end the company can always find new people that choose 1. "Do it or I'll find someone who will." is both true and powerful. If you quit you lose your livelihood and potentially could be blacklisted from the industry, and yet the bad thing still happens anyway.

> "Do it or I'll find someone who will." is both true and powerful. If you quit you lose your livelihood and potentially could be blacklisted from the industry, and yet the bad thing still happens anyway.

That's merely a rationalization to remain a member of group 1.

The truth is that 2 and 3 are not mutually exclusive, and together, they can and do occasionally prevent "bad thing" from happening. In some cases, it's because group 1 is small enough that the company cannot hire a toady to do "bad thing" in time to retain their contract for doing so. It may also be because the combination of external publicity coupled with a high-profile resignation is enough of a PR hit that the company reconsiders.

There's another element to consider: other, unrelated teams within the company who catch wind of "bad thing" can also opt for 2 and/or 3, further disincentivizing the company and perhaps even its competitors from proceeding, as losing a leading team in a highly-contested market segment could mean ceding significant ground to a competitor.

Granted, such moral stands don't affect change as often as one might like, but parroting "bad thing will always happen regardless what you try" everywhere certainly doesn't help things.

no engineers are paid to think. that's why they're not technicians.

and it's why they usually have to have a decent amount of education -- which usually includes ethics -- and in many cases (like civil or mechanical engineers) have to have a formal certification from a governing body. programming doesn't have that, but probably should. regardless, it's a knowledge field, not a fork-and-spoon operator gig.

> "Do it or I'll find someone who will." is both true and powerful.

it's only powerful when there are high salaries attached. most people would swallow their objections for a steady 300k/yr

> So think about this: say your CEO asks you to make a missile specifically to target and kill children.

Well, I wouldn't and I would get fired, and the CEO would find someone with less scruples. And you would say "Damn, these technologists have no mercy" even if 99 engineers have refused before the one that accepted.

You're simply falling for survivorship bias.

> You're simply falling for survivorship bias.

What if I told you that I worked in fields that raise such ethical questions, and I have seen with my own eyes how there was never a single instance with 99 engineers who refuse for 1 who accepts?

For what I have seen, there was no survivorship bias. And then regarding LLMs in particular, I struggle to find people around me who actually question the ethical problems they bring. They just don't seem to even imagine that they could think about ethics.

So unless you don't ask them to do something that is extreme (design a weapon to kill children), they just don't question anything at all.

I guess for the same reason no one else thinks about them either. Capitalism in general seems to reward companies and individuals for being remarkably short sighted when it comes to the long term effects of their actions, and just about everyone operates under the assumption "well, it's going great now, who cares what the consequences are in future?"

But hey, I guess you could say that's humanity's flaw in general. If all our current social and environmental issues are taken into account, it's pretty clear that most people don't think about the consequences of their actions beyond the next week or so.

> technologists don't think about the consequences of what they do before they do it

Software "engineers" are not part of a professional body at all, let alone one that espouses ethics. That only seems to happen in professions with a high risk of lawsuits due to customer death.

How deeply would you expect an interior designer or a landscaper to think about future consequences?

It is well known that more regulated professionals have never done anything morally questionable.

AI search is still a new thing, wait a little for it to become fully mainstream and people will start figuring out how to optimize for it, and also how to trick it, too - it's just unavoidable. And eventually we'll end up having AIEO experts instead of the current SEO experts...

> And eventually we'll end up having AIEO experts instead of the current SEO experts...

The article calls that LLMO :-)

And these new experts will be human... or AI?

My guess is this has nothing to do with short term profits, but rather it's hard to make search robust to SEO. If you're right, Bing search should be better for recipes since they don't benefit from Google ads on these websites. Is it the case?

Even if Bing is better, that might be simply due to the small player's advantage: Google's market share is so large, they have essentially the whole economy of the web investing in SEO against their search quality, adapting quickly whenever Google tunes the algorithm. Meanwhile Bing can tune their algorithm and nobody cares.

Reply on news.ycombinator.com