Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

Generative AI Is an Engineering Disaster

theatlantic.com

Post reply on HN
Print view View original

81–90 of 97 posts

Re: Generative AI Is an Engineering Disaster

#81

Post by scotty79 »

JackSlateur wrote: post #76
scotty79 wrote: post #68

Earlier quoted context omitted.

Each solution generates new problems. That's the general physics of intelligence in our universe. Also label of "fake" is completely arbitrary. Because someone might call cancer a fake problem, because people shouldn't live too long and they need a reason to die.

A colleague of mine is fan of AI and we have, from time to time, serious discussion about that topic Last week, I asked him in all seriousness: considering that everything you say is correct .. why the hell does you product still sucks ?! He answered me: because of people Perhaps people is the only problem that need solving :(

I think AI is going to enable a new future where most software products will have userbase of 1. And then they are not going to suck.

Re: Generative AI Is an Engineering Disaster

#82

Post by simianwords »

unknownfuture wrote: post #74
simianwords wrote: post #46

[flagged]

> This is obviously untrue so either the author is knowingly lying or is plain incompetent. We had LLMs not being able to do simple high school mathematics a year back, now it is solving open problems in mathematics. Fields medalists in physics and mathematics are using it on a daily basis. I'm confused, how do you think this disproves the claim in the article? What do you think that quoted portion was talking about?

The two claims are not related, my bad for making it seem so. I about that a serious person would still needs proof that AI is becoming more efficient. If you need proof, here's one:

> A year ago, we verified a preview of an unreleased version of @OpenAI o3 (High) that scored 88% on ARC-AGI-1 at est. $4.5k/task

> Today, we’ve verified a new GPT-5.2 Pro (X-High) SOTA score of 90.5% at $11.64/task

> This represents a ~390X efficiency improvement in one year

https://x.com/arcprize/status/1999182732845547795

This is not even including newest improvements. GPT 5.6 beats GPT 5.5 by 1 OOM.

https://x.com/GregKamradt/status/2075274981794300113

Re: Generative AI Is an Engineering Disaster

#83

Post by unknownfuture »

simianwords wrote: post #82
unknownfuture wrote: post #74

Earlier quoted context omitted.

> This is obviously untrue so either the author is knowingly lying or is plain incompetent. We had LLMs not being able to do simple high school mathematics a year back, now it is solving open problems in mathematics. Fields medalists in physics and mathematics are using it on a daily basis. I'm confused, how do you think this disproves the claim in the article? What do you think that quoted portion was talking about?

The two claims are not related, my bad for making it seem so. I about that a serious person would still needs proof that AI is becoming more efficient. If you need proof, here's one: > A year ago, we verified a preview of an unreleased version of @OpenAI o3 (High) that scored 88% on ARC-AGI-1 at est. $4.5k/task > Today, we’ve verified a new GPT-5.2 Pro (X-High) SOTA score of 90.5% at $11.64/task > This represents a ~…

That's still not the kind of efficiency the article is talking about.

They're referring to raw training and inference scaling and it's relationship (or lack thereof) to traditional economies of scale that we've seen with past technologies where they get cheaper as adoption increases, not more expensive.

Its true that the models requiring fewer turns and tokens due to increasing sophistication improves cost efficiency for users but that doesn't address the fundamental computational scaling problems of LLM architectures.

Re: Generative AI Is an Engineering Disaster

#84

Post by unknownfuture »

tedggh wrote: post #77
maxcb wrote: post #3

Comparing how traditional tech companies scale to how frontier AI companies scale doesn't seem very fair to me. AI is still relatively new territory, in comparison traditional software companies benefit from decades of investment in cloud infrastructure, networking, and hardware. I'm certain those technologies also went through periods of heavy investment before they became scalable. The author also talks about advan…

I think LLMs, with only a few exceptions, have mostly created the illusion of value. It’s undeniable I can write code, troubleshoot and document much faster than in the pre LLM era, but putting a dollar value to that it’s not particularly easy. One group of people being 100x faster at one task doesn’t mean the organization as a whole runs at 100x the speed, in fact in most cases where the bottle neck is somewhere els…

Yup. We're just all re-learning Amdahl's Law.

Re: Generative AI Is an Engineering Disaster

#85

Post by simianwords »

unknownfuture wrote: post #83
simianwords wrote: post #82

Earlier quoted context omitted.

The two claims are not related, my bad for making it seem so. I about that a serious person would still needs proof that AI is becoming more efficient. If you need proof, here's one: > A year ago, we verified a preview of an unreleased version of @OpenAI o3 (High) that scored 88% on ARC-AGI-1 at est. $4.5k/task > Today, we’ve verified a new GPT-5.2 Pro (X-High) SOTA score of 90.5% at $11.64/task > This represents a ~…

That's still not the kind of efficiency the article is talking about. They're referring to raw training and inference scaling and it's relationship (or lack thereof) to traditional economies of scale that we've seen with past technologies where they get cheaper as adoption increases, not more expensive. Its true that the models requiring fewer turns and tokens due to increasing sophistication improves cost efficiency…

This doesn't mean anything. Here are the facts

1. models themselves are getting cheaper - around 5000x in the past 1.5 years

2. ironically, if it were fully cheap, the same crowd would say that this would make AI bubble pop because where would they get moat?

3. training is also getting cheaper, the whole lifecycle to train and do inference on models from 1 year ago is on the whole 50x cheaper or so

I also think the author's point is a bit nebulous

> Chatbot companies are aware that their products are inefficient. Some have found techniques for improving performance, but they have not yielded significant gains

Performance has increased. To you specifically: what metric would falsify the claim that "they have not yielded gains" and "their products are inefficient"?

The same metric should apply to

- industries like steel

- pharma

- internet/cloud computing

- automobile

- consumer electronics

All of them have had high impact. So any criticism on inefficiency should be unique to LLM's and shouldn't apply to all of them. Please answer.

Re: Generative AI Is an Engineering Disaster

#86

Post by Paradigm2020 »

neonstatic wrote: post #79
Rury wrote: post #75

Earlier quoted context omitted.

There are more numerous cases where cities happen to be situated fairly close together in Europe, where trains actually are a more sensible travel option, hence the fixation there. Planes instead make more sense the further you must travel. I mean, depending on circumstances, it takes nearly the same amount of time to travel from Chicago to St. Louis by plane as it does by car. Something shorter than that and it can…

3 years ago I was traveling from northern Poland to southern Poland by train - the only option unless you have a car. It so happened that my family was flying from Toronto to Warsaw at the exact same time. They arrived before I did and were much more comfortable than I was, too. Not to mention the safety aspect. Of course if we're talking about cities that are 20-30 minutes away by car, then flying would not be reaso…

You call strawman on someone else and show up with anecdata... And even anecdata with so little information as to be useless.

In all honesty one of the biggest problems in all these discussion is that there is no Europe healthcare / trains /...

Spain has the second most rail in the world per Capita and together with France are one of the best experiences (Japan and China in their own league) while Germany is currently shit and definitely not up to its "reputation".

Public healthcare in Belgium (personal experience, doctors / 1000 people etc) is amongst the best in the world while I guess based on your comment the one in poland is not that good ?

Nevermind the USA that has places like Appalachia and San Francisco

Re: Generative AI Is an Engineering Disaster

#87

Post by dofm »

unknownfuture wrote: post #73
try_the_bass wrote: post #70

Earlier quoted context omitted.

I mean, that is the essence of how to fix the problems in a capitalist society? One of the core tenets of capitalism is that one person's profits are another person's opportunity. If the latter can do the work/obtain the capital to break into the industry, they can capture some of those profits by offering competition at a lower price point. I don't think the person you're bagging on is saying it's easy to do; but th…

> I mean, that is the essence of how to fix the problems in a capitalist society? So... we're just gonna forget that governments exist? We gonna assume rivers would've stopped burning in the US if only there had just been more startups working on making rivers stop burning?

Superfund-as-a-Service

Re: Generative AI Is an Engineering Disaster

#88

Post by unknownfuture »

simianwords wrote: post #85
unknownfuture wrote: post #83

Earlier quoted context omitted.

That's still not the kind of efficiency the article is talking about. They're referring to raw training and inference scaling and it's relationship (or lack thereof) to traditional economies of scale that we've seen with past technologies where they get cheaper as adoption increases, not more expensive. Its true that the models requiring fewer turns and tokens due to increasing sophistication improves cost efficiency…

This doesn't mean anything. Here are the facts 1. models themselves are getting cheaper - around 5000x in the past 1.5 years 2. ironically, if it were fully cheap, the same crowd would say that this would make AI bubble pop because where would they get moat? 3. training is also getting cheaper, the whole lifecycle to train and do inference on models from 1 year ago is on the whole 50x cheaper or so I also think the a…

All of those industries benefited and continue to benefit from economies of scale in the manufacturing and delivery of their products. You're not making the case you think you are, and I have a suspicion the article's point is just sailing right past you.

Re: Generative AI Is an Engineering Disaster

#89

Post by simianwords »

unknownfuture wrote: post #88
simianwords wrote: post #85

Earlier quoted context omitted.

This doesn't mean anything. Here are the facts 1. models themselves are getting cheaper - around 5000x in the past 1.5 years 2. ironically, if it were fully cheap, the same crowd would say that this would make AI bubble pop because where would they get moat? 3. training is also getting cheaper, the whole lifecycle to train and do inference on models from 1 year ago is on the whole 50x cheaper or so I also think the a…

All of those industries benefited and continue to benefit from economies of scale in the manufacturing and delivery of their products. You're not making the case you think you are, and I have a suspicion the article's point is just sailing right past you.

notice how you didn't give me a metric that proves efficiency/inefficiency?

Re: Generative AI Is an Engineering Disaster

#90

Post by unknownfuture »

simianwords wrote: post #89
unknownfuture wrote: post #88

Earlier quoted context omitted.

All of those industries benefited and continue to benefit from economies of scale in the manufacturing and delivery of their products. You're not making the case you think you are, and I have a suspicion the article's point is just sailing right past you.

notice how you didn't give me a metric that proves efficiency/inefficiency?

And now I know you either don't understand what an economy of scale is or you're not discussing in good faith. Carry on!
Post reply on HN