The article is not a technical overview of AI's failures, it's about the organizational failures that lead to things like token leaderboards. I would argue that this article goes far beyond the "obvious" fact of token leaderboards and into the deeper problems faced by both vendors and companies trying to measure the efficacy of AI.
Honestly, it's just shouting into an echo chamber. They call it organizational failure, but it's really just 'come on, my side.' A consultant who does that is untrustworthy. It means they have no metrics.
Same goes for their 0% claim. It's overgeneralization, isn't it?
So why did they fail, what went wrong, and how should these AI projects be analyzed—and by what metrics? They skip all that and just throw out a single KPI. What insight does that offer?
Honestly, there are a lot of people on HN who just shout that AI is useless because their own jobs are at risk. That's an undeniable fact. In my country, Korea, we say 'like a pheasant hiding its head in the ground.' It means denying reality and lacking self-awareness.
I've also seen many failed AI projects due to a lack of proper evaluation of AI capabilities. I don't want to deny that premise.
But if you're going to make that argument, you need to describe why they failed. There's none of that here, just this:
'We're going to opt out because it looks like AI projects will fail.'
They package it as being cautious, but when you peel back the wrapper, it really means:
'We can't keep up with the AI trend. And we can't tell what works and what doesn't.'
Sure, AI has a bubble. No one can deny that. But separate from that, it is useful. Do you really think millions of consumers would stick around for something useless?
Social problems are arising, and people are losing jobs because of it. Let's think the other way around. It doesn't make sense for people to lose jobs over something that's truly useless and wrong.
People are losing jobs because it's useful and valuable. So wouldn't it be more accurate to say that the problem is 'over-application' rather than the thing itself being bad? In other words, we could say that it's being valued more than it's worth. But this argument just says it's bad, period, with no nuance. I can't trust someone who talks like that.
AI writing, if you give it a red-team prompt, can at least refine the surface of counterarguments to your claim. But this—from overgeneralization to extremes—is a textbook example of bad human writing
(i'm more chatty because it's my "down-day". Hope you don't mind)
I don't think you're upset because AI is useful and valuable. Are you upset because you can't describe how the bubble will burst?
Some (more) quantitative types saying that the economics are deflationary. So NVIDIA have already won.
But there are some that claim that the bubble has nothing to do with NVIDIA's monopoly on compute (21.C steam engine/steel). That the bubble is in the OX layer. (You?). If Anthropic (or Palantir replacement) fail to do what? Then the bubble will burst.
This is the more interesting but I will have to sleep on it.
I think AI is useful and has value, but what I find problematic is the illogical behavior that people display while claiming to be engaging in thoughtful writing.
Rather than saying NVIDIA will be the ultimate winner, I think Apple could be the final winner as a hedge against semiconductor stocks, at least for now.
In that sense, I'm not sure. You can't time a bubble, so it's just a bubble.
But I do think NVIDIA will remain powerful because of its CUDA ecosystem—though the future is uncertain.
[Bro I was hoping you'd rip into your own OX idea here.. then we could have a convo here or offline.
I wouldn't say I get frustrated by illogical writing--esoecially ones that only _seem_ illogical---, but more that people like us who might know better don't point out precisely where that illogic is after thinking carefully about it.
]
It's possible to time a burst _if_ you see the pin, but anything you do _when_ you see it is technical illegal?
So your OX idea feels like just the sort of thing that could open our eyes new kinds of pins.
Further, I'm not totally closed to the idea that actively trying to pop the Anthropic bubble is moral, indie of how quixotic that sounds. Even if it might be a good case study for deinstitionalizing research math.. even if that Napoleon quote that's trending..
I would also never bet against Jensen taking some ideas from "the Apple playbook". Especially when multiple people have already pointed it out. Though this is tangential. Less tangential---
Refinement of OX framework would help us think about how NVIDIA can or cannot play the Apple game. Certainly, if given calibrated riskiness and timing, Apple would move against Anthropic. You know what I mean?
If this were genuinely a consultant's writing, they would need to present a modeling hypothesis for why failure occurs in organizations and support it with metrics. But claiming their own observation of 0% is not a hypothesis.
Of course, they can write it on a personal blog. That's not inherently bad.
But when they're rejecting client requests while offering no opinion on how to model the metrics that would define success—that's the problem. This type of writing is just an illusionary piece that claims to offer insight. It's a bad form of human writing. I like to call this style the 'Rhonda Byrne Secret type.'
At the very least, a consultant should provide modeling of their hypothesis and actionable advice.
Explaining everything with a single reason—that people are afraid of being laid off if they don't adopt AI—is just a cop-out. There's no hypothesis, no reasoning behind why that claim holds. It's trying to explain everything with just that one assertion. I'm not even sure if that holds up organizationally
The OP offers no logical reasoning about how decisions are made or why they break down.
A proper piece of writing would have looked like this:
Why did the client want to adopt LLMs?
What problems arose during the client's process of adopting LLMs?
How does this connect to the organization's internal issues and disrupt its pipeline?
You need at least that level of description. This is just an emotional post that seems to say 'I wish LLMs would fail.
If they had really explained organizational dynamics, they would have talked about the asymmetry of expected costs.
Professional managers and mid-level managers prioritize career defense above all else, so they're extremely risk-averse. They would have explained the costs of adopting AI versus the costs of failing to adopt AI.
If they reject AI adoption, shareholders or the board would say they're 'failing to read the trend.' If they adopt AI and fail, they can package it as 'early market entry costs'—a defensible risk.
Providing that kind of defensive logic to decision-makers is exactly what consultants do.
It's described as being held at gunpoint, but it's really just a prisoner's dilemma. In a market where competitors and vendors are all lying and saying 'we've achieved 100x productivity with AI,' the moment someone tells the truth and says 'we don't need AI for our business,' they become a political target both inside and outside the organization. In the end, everyone knows it's useless, but they approve budgets for 'fake AI projects' anyway. I can offer that kind of cheap insight too—it's a Nash equilibrium where the system has been driven to a race to the bottom.
The ban on replaceable flavored pod vapes caused that. Such a bizarre rule and so wasteful. You couldn’t replace the battery on the pod vapes but at least you could reuse the recharge and reuse the battery.
>A problem for city kids maybe? My parents were smart enough to raise me somewhere more suited for children.
Implying that the only reason people with kids live in a city is because they're stupid? What in the world are you on about? I've read the below thread, and this is one of the most bizarre takes I've seen in a long time.
Tech companies are some of the largest and most influential companies in the world, and people are just now starting to understand what it means to have their lives influenced by wide-reaching algorithms. Sounds like a great topic for journalism to me.
Regardless of the motivation, NYT's recent algorithm-centric investigations (like the YouTube one a month or two back) are really interested and well-researched.
His point may be that devs that are constantly job-hopping for better pay will "experience the same year 10 times" and not grow in skills.
I'm not sure I agree with that viewpoint.
Not devs, but junior devs in particular. Jumping 10 times a year may be for different reasons for experienced people, although some may see it as a red flag.
There is a huge difference between devs that do development for the money and devs that do development because it's engineering. Nothing wrong with this, but it's not a recipe for becoming great developers.
Would you want to work with code monkeys that do whatever they are told to keep/get job or craftsmen that care about what they do?
Why is it either or? Yes I work for the money first and foremost. I don’t learn anything that isn’t going to either make me more money by being better at my job, keep me competitive in the market, or make more money in the future.
Yes I started programming in the mid 80s in my bedroom. But I keep up to date strictly for the money.
Of course not, they should be comfortable to be able to pursue this, but I don't think it's a recipe for becoming great developers. They may become great managers, CEOs, etc.. but not craftsmen.
Would you rely on someone that started their tech career thinking about pay instead of engineering challenges, problems and solutions?
The New York Times comments are free from trolls and spam, but it's a frustratingly obvious echo chamber when it comes to politics. I'm a liberal guy but I can't stand it. David Brooks wrote an interesting column (http://www.nytimes.com/2016/08/09/opinion/the-great-affluenc...) a week or so ago and most of the comments are just bashing him for being a Republican, as if that has anything to do with the subject matter.
i think your diagnosis is wrong. show me a conservative news outlet on the web with a high SNR of thoughtful and intelligent comments, free of frothing, conspiracy-laden bullshit. maybe the NYT is an echo chamber because modern conservative positions are so weak and contradictory, they can't stand the withering critique of a well-moderated forum. instead, they only survive in troll havens.
as for brooks' column, you might be missing some context. brooks has made a career of talking out of both sides of his mouth and (annoyingly) providing intellectual cover from the NYT for a plethora of bad conservative ideas. now that they're blowing up in his face, he's backing away from these stances.
"He should realize that we’ve been trying to bring the tribal ethos to the U.S. for a long time, with strong local communities providing the sort of help and social services that bind people together and take care of each other as we get older, or fall short in some way.
But He Who Talks with Forked Tongue likes to imagine an egalitarian utopia where 99 percent of us are quietly stitching blankets while a few get to hoard the vital resources. When the tribesmen and women protested and occupied Wall Street, Brooks nearly went on the warpath, and wrote a column in the Times entitled “The Milquetoast Radicals,” (10/11/2011) in which he castigated the unwashed hippies who dared to protest the insane degree of income inequality in this country."
I was not comparing NYT to any other news outlet. It's an echo chamber regardless of the fact that conservative news sites also have echo chamber comment sections.
The David Brooks article was just an example. When Bernie Sanders was still campaigning every comment on Hillary/Bernie-related articles was about how the New York Times is wrong and that Bernie is the best, people will learn about the political revolution soon enough, etc. I was a huge fan of Bernie and I got bored of those comments instantly.
neither am i comparing them. you're observing that the NYT comment section is free of trolls/spam but is otherwise a liberal echo chamber. that's another way of saying that it's lacking a counter-balance of intelligent conservative comments. i'm accepting that critique for the purpose of argumentation, and replying by asking you to look around and find anywhere on the web that has a majority critical mass of intelligent conservative commentary. once you realize that it pretty much doesn't exist, maybe that will lead to a different conclusion...
anyway, one of the few places i read that, for whatever reason, does carry an even mix of intelligent comments across the spectrum is interfluidity. for example:
Holy Hell, you are right. I actually have the distinguished privilege of having one of my NYTimes comments be a "Times Pick", meaning the ed. board actually read it and recommended it for insight, I suppose.
My comment was mostly meta, calling out people for missing the point of an op-ed. The op-ed was from a privacy/civil liberties person about why a "no buy" list for guns would be a bad idea. He wasn't arguing on the merits for or against gun ownership, just that these secret lists on which LEO acts are dangerous.
Every comment was something along the lines of "What about my right not to be shot in the streets?!" - I tried to point this myopic view out, and every reply to my comment was "What about my right not to be shot in the streets?!".
As a smallish-government liberal (Public services are well and good but government should be kept in check by a powerful and vigilant population) I get torn up whenever I defend gun rights or spending reductions on that website.
With all his talk about planting bombs, I'm sure they suspected that he was wearing a bomb vest to detonate when police moved in. I'm not sure how else they could have handled it.
How would AI writing help in any way?