For example -- not long ago, when you wanted to do l10n/i18n for your business, you'd have to go through a pretty painful process of integrating with eg translations.com. If you're running an ecommerce site with a lot of new products (and product descriptions) coming online quickly, that whole process would be painful and expensive.
Fast forward to today -- a well-crafted prompt to Llama3.1 within a product pipeline makes that vendor completely obsolete. Now, you could argue that this kind of automation isn't new, you could have done it with an api call to Google translate or something similar, and sure, that's possible, but now you have one single interface into a very broad, capable brain to carry out any number of tasks.
If I was a vendor whose business was at all centered around language or data ETL or anything that involves taking text and doing something with it, I would be absolutely terrified at someone writing a 20-line python script with a good system prompt that would make my entire business's reason for being evaporate.
AI has absolutely revolutionized spam and spam detection. Spammers can now generate absolutely unheard of amounts of complete bullshit. And on the other side, spam detection services and algorithms are getting better and better at detecting it, sorting it, and filtering it based on user preferences. Tons of people are enjoying openly AI generated content; and the content that isn't enjoyed by people is instead enjoyed by other AI bots, driving up the engagement rates. That behavior too though is being monitored by other AI, which then prompts spammers to improve their AI so they can avoid that AI and get their stuff seen by engagement AI.
So we have server farms full of computers that are making complete shit that is then thoroughly enjoyed by other server farms full of computers to drive up engagement numbers while still other server farms full of computers are working to detect the fraud and remove it.
Meanwhile, in the real world, we're still hurtling towards climate collapse. But that's okay, we're finally looking into building nuclear reactors again. To power more data centers.
The future is fucking stupid.
>Substantial and ongoing improvements in AI’s capabilities, along with its broad applicability to a large fraction of the cognitive tasks in the economy and its ability to spur complementary innovations, offer the promise of significant improvements to productivity and implications for workforce dynamics.
I keep waiting for the industry shifting changes to materialise, or at least begin to materialise. I see promise with the coding tools, and personally find Claude and Cursor like tools to warrant some of the general hype, but when I look around for similar changes in other tangentially related roles I draw a blank. Some of the Microsoft meeting minute summaries are good, while the transcripts are abysmal. These are helpful, but not necessarily game changing.
Hallucinations, or even the risk of hallucinations, seem like a fundamental show stopper for some domains where this could otherwise be useful. Is this likely to be overcome in the near future? I'd assume it's a core area of research, but I know nothing of this area, so any insights would be enlightening.
What other domains are currently being uplifted in the same way as coding?
> This technical progress is likely to continue in coming years, with the potential to complement or replace human labor in certain tasks and reshape job markets. However, it is difficult to predict exactly which new AI capabilities might emerge, and when these advances might occur.
The "small" benefits you list are in fact unprecedented and periodically improving (in my experience).
The generality and breadth of information these models are incorporating was science fiction level fantasy just two years or so ago. The expanding generality and context windows, would seem to be a credible worker threat indicator.
So it is not unsensible to worry about where all this is quickly going.
It's only the mechanism that's unprecedented, cementing these new approaches as a state of the art evolution for code completion, automatic summarizing/transcription/translation, image analysis, music generation, etc -- all of which were already commercialized and making regular forward strides for a long while already. You may not have been aware of the state of all those things before, but that doesn't make them unprecedented.
We actually haven't seen many radical or unprecedented acheivements at commercial scale at all yet, with reliability proving to be the the biggest impediment to commercializing anything that can't rely on immediate human supervision.
Even if we get stuck here, where human engagement remains needed, there's a lot of of fun engineering to do and a number of industries we can expect to see reconfigured. So it's not nothing. But there's really no evidence towards revolution or catastrophe just yet.
These points made sense to me: it is impossible to predict what will actually happen, we need better pro-level tools for AI assistance (e.g. Copilot, writer autocomplete, ControlNet) rather than AI as a full replacement, and we need better and clearer paths to retraining and job mobility.
I disagreed with only one point in there: that research is needed for ways to compensate people for the use of their creative works, but that is solely because of my pro-free-cultural moral views. The rest of the article is still good.
Mike Montero would like a word
"Who in this room is now, or has at some time, been in creative services?
"Who here has, at some time, had trouble getting paid by a client for work they were doing?
"Raise your hand if any of these are familiar to you:
"'We ended up not using the work.'
"'It's really not what we wanted after all.'
"Alright. Who's familiar with Goodfellas?
"Alright. 'We got somebody internal to do it instead.'
"'Fuck you. Pay me.'"
"'Fuck you. Pay me.'"
Free culture isn't against the former (therefore this video doesn't actually address the point), but is against the latter, as being restricted from replicating work harms culture and innovation as a whole (e.g. memes and fan art being technically illegal), and imposes a large cost on the public.
That said, I'm not fully against IP laws, just that it should be limited to 14 years and only in situations where it is necessary for the production of it in the first place (e.g. articles behind paywalls). I believe I have a right to an opinion on this as a member of the public, as IP laws are a compromise between the public and the creators. It's not some natural human right.
In this moral view, if AI trains on my HN comment for example, copyright shouldn't come into play because I didn't require it to produce this comment. I had other incentives to write this comment.
As a counter-example, no one cares about statistical analysis (what AI is) when it's just building a corpus, doing classification, or even generating GPT-2 level text etc. It's only when it becomes a threat to jobs when people panic. This reveals the real problem: it is about jobs, not data. And so the solution: financial support, equal education and job retraining. Not expanding copyright laws to cover analysis as well.
This can go one of two ways:
1. Fewer jobs will be used to further suppress wages. What little wages people earn will be used for essentially subsistence living. The extreme end of this is like the brick kiln workers in Pakistan, India and Bangladesh. A lot of people, myself included, call this neofeudalism because you will be a modern day serf. The welath concentration here will be even more extreme than it is now. We're also starting to see this play out in South Korea; or
2. The created wealth will elevate the lowest among us so work becomes not required but a bonus if you want extra. The key element here is the removal of the coercive element of capitalism.
To put this in perspective, total US corporate profits are rapidly approaching $4T per quarter. That's roughly $60,000 per US adult. Some would call that the exploited surplus labor value.
Here's another number: we've spent something like $10T on the War on Terror since 9/11. What could $10T buy? Quite literally everything in the United States of America other than the land.
What's depressing is that roughly half the country is championing and celebrating our neofeudalist future even though virtually none of them will benefit from it.
