LLMs are plateauing. At this point, anyone who has cared enough knows their fundamental limitations. Don’t get me wrong, they do provide immense value and have grown quickly in just a few years. But, they will probably never get past the “we must treat it like a junior engineer” phase, especially as LLM outputs inevitably leak back into the training data since everyone is using it (or being forced to by employers). Notice that everything major from these companies lately hasn’t been a vastly improved model from the previous iteration, but products and tooling: see Anthropic allowing you to create in-Claude apps, every companies own version of agentic CLI, and so on.
The article seemed a little fluffy, and I wondered if at least part of it was LLM generated.
However I do agree with the general sentiment.
I find the current hype cycle of LLMs to be similar to the petro industry: there are many useful applications of petroleum, without having to set it on fire. And yet the industry is loath to give up on a use that consumes copious quantities of it's product.
LLMs have many beneficial applications, in deep data analysis and pattern matching. Yet industry is intent on applying the tech to problems where its results are dubious, because of the mass market of those applications.
So much for the magical black box of the market 8-/ As in every documented case ever: product vendors will take every dollar possible, to hell with consequences.
Think of the radioactive skin cream of the '50s 8-/ Sure, it's "good" for you, and "more doctors choose camel"... And never mind that whole libtard fake-news that the planet's ecosystem is going to shit... Every patriot knows the little bebe jesus put it all here for us to trash, obviously...
It's in this context that LLMs are definitely the best way to decide whether or not your insurance company pays for you to get a kidney transplant 8-/
As long as I'm on a roll, and related to my first sentence, I _really_ dislike the em dashes 8-/ and their use has really spread with the prevalence of LLM generated text. Not just in the LLM text, but in the text of humans that are influenced by the LLM text.
If people would just use them "incorrectly", that is with spaces around them, then they would follow the general rule of english writing, to delimit words and phrases with a space.
What do people think is a realistic outcome here?
In a global economy, no country can stop deployment of consumer AI for digital goods unless you go full North Korea.
If you want some kind of international moratorium I'm all ears, but whining at people for buying AI digital art for $0 instead of graphic designers for $10,000 is utterly pointless. At best, you'll get them to buy art from Philippine designers... who are using AI... for $.50.
I have reservations about AI, but what do you gain with this approach? Guilt and national protectionism is utterly pointless.
Of course there will be displacement of workers. That's entirely the point. "We are in the business of unemploying people", like software engineering in general, but on steroids.
Growing backlash yet used by hundreds of millions of people and growing.
This article doesn't talk much about mass hiring during COVID period due to high demand, what we see now is unwinding of that trend, feels like people behind this type of narrative are interested in regulating what goes into these models.
Clickbait
This drivel is worse than ChatGPT.
There is no deep insight.
It is highly formulaic and mechanical:
1. Find a controversy
2. Invent a trend
3. Get quotes from some people who agree with you
4. Mention “ethics” all at a very superficial level.
5. Publish
You would be much better served by having a 15 minute conversation with ChatGPT about this topic than reading this article.
All the negativity really is an understandable response to so many negative changes to the status quo way of life, but I can't help but take a step back as a realist and say that Pandora's box is open.
That is not to say we are powerless to regulate aspects of how this technology is legally utilized, but this sort of quasi-luddite "turn it off" response is totally infeasible.
The best bet we have for making the AI future suck less is by building things which make life more compatible with AI. That means a revamp of how we handle intellectual property, how we handle education, how we manage an economy humanely in a world where human labor potential is comparatively diminished.
We need to focus upward toward the things that can make humanity able to handle the transition. Not downward toward the sand where we can pretend it isn't happening.
EDIT: please respond if you're going to take away my karma for my honest opinion. I use AI almost every waking hour but we have to be realists about the impacts on society
I find it rich that this article seems to glorify the proliferation of the Internet itself, completely ignoring that social media has been driving mass addiction since the late '00s and that the rich have been growing exponentially richer long before LLMs and the like made their splash. Maybe this is because WIRED's business model requires the existing 21st century media landscape to function.
I'm not exactly an "AI optimist", but this is not constructive journalism by any means. There are countless unaddressed, explicitly tech-related issues that would only further metastasise if we arbitrarily reverted and then halted progress to 2021.