There’s a difference between fixing a few incorrect words in a text and having an essay (or email or whatever) written from a few words. I don’t think the parent comment would object w/ your use of LLMs.
I bought the device 5 days ago! So far I like it. There are a few flaws (the most annoying being the glare visible when showing high contrast things, e.g. a movie on a dark background, and the text not being crisp for coding). I see myself working 4h+ a day on it though, yes, but it’s early to tell.
I don’t think it’s only that. I personally hate AI not because of CEOs and co, but because the tech is intrinsically born out of theft, and is still, to this day, evolving thanks to theft. And that’s even before the ecological considerations.
I've been on the free information train my entire life, back to my little hacker punk days in the 90's, so my opinion on that isn't worth much. I do think that the ecological considerations are also entirely the fault of the aforementioned CEOs. Machine learning research has been ongoing in good faith since the 40's. Blaming the technology is kind of silly. Imagine if we had banned trains because the robber barons were assholes in the 1830's.
This technology is going to drive some incredible discoveries in all of STEM. The robber barons and monopolists tend to come out of the woodwork when incredible technologies emerge. It just sucks that we still haven't evolved them out of society.
> Imagine if we had banned trains because the robber barons were assholes in the 1830's.
The thing is they don’t control them anymore, governments do by and large now, so a lot of the issues that came with their ownership and special privileges no longer exist. There is no way that is how this is going to go down with Google et al
We could probably do without computers too, but that would be idiotic because they speed everything up. There's a good chance that the next pandemic is swatted down by LLM-powered vaccine development much faster than COVID-19 was.
"We could do without them." is not a great take when it comes to people dying prematurely.
In fact, I would bet that this particular technology will lead to climate change solutions eventually. If nothing else, it will drive an energy revolution in either nuclear or solar power. Probably too late to solve the AMOC collapse, but mitigation is still in play through science.
I hope somebody is documenting all these bombastic LLM-related public statements. They're going to be a neverending source of cringe and laughs for the next 50 years.
(I've heard one C-level dude say with a straight face that LLMs were a "more significant invention than writing".)
You're a Google search away from fact checking me if you want to do that.
I'm a DevOps engineer, not a C suite guy, but I tend to agree with you in general. I think there is a lot of smoke being blown into the hive around this technology but having used it extensively, and having witnessed its progression first hand in engineering, these tools are insanely useful and have made giant leaps forward in just ~4 years.
Don't know if you're a believer in Moore's Law or not, but I don't think your tune is going to take anywhere near 50 years to change. I'd be surprised if it took 5 years.
No offense, but your post reads like LLM psychosis. Mostly because you failed to understand my post at all and rushed in to defend AI's honor. (Why? I don't think AI wants or needs your oaths of loyalty.)
But you would be banning trains if they were built to just run smack into the centre of town squares loaded with bombs, rendering the cities to dust, as a part of their design and boasted about by the owners.
At least until the maniacally evil train ownership debacle was better organized to prevent such harm in their core application.
I parsed that wrong (better?) as running smack (heroin) into the center of town
And so I thought it was a reference to the overall AI psychosis, FOMO, and dopamine marketing. That is the big impact that bothers me, regardless of whether this deployment uses massive new datacenters. It is a toxic social impact even if it somehow reduced atmospheric CO2 with every token.
There are datacenters all over the place and have been for a very long time. Some of them host physical servers for people and companies, maybe only a literal closet somewhere in a building. Others are giant hyperscaler datacenters that have tons of 24/7 lighting and are the size of multiple football fields.
We need to be very careful here, or we're going to throw the baby out with the bathwater.
That is also my concern. This feels like a Satanic panic type of thing and if people don't form nuanced opinions, politicians will latch on to the issue and before you know it, many people who are heavily reliant on machine learning who don't realize it are going to have a bad time when stuff starts getting banned.
The ecological considerations are wildly overstated. Data centres in general != AI, and other industries, including meat production and (ironically) paper for print all use far more water and create more damage.
This might change in the future if the planned insanely huge data centres get built and used. But today the situation is clear - AI isn't any more ecologically damaging than other popular data centred activities like streaming music and video, and general social media.
Also, I just listened to the latest Volts podcast and they make the claim that data centers will actually lower the cost of electricity fairly soon (~2030). Very counterintuitive but it does make sense. We'll find out soon enough.
I've been telling my friends this since LLMs launched. If your interest is in lowering the use of fossil fuels, then a global race for compute is great because the increased demand is probably going to usher in a nuclear/solar revolution in the energy sector.
It feels like a very large segment, on both sides of this argument, is completely incapable of forming nuanced opinions on this stuff.
The IP considerations, environmental considerations, "lol we're gonna destroy the world and get you laid off" considerations, and of course the big middle finger given to artists of all types, from authors to musicians... They painted themselves as villains and then they were shocked when people viewed them as such.
but because the tech is intrinsically born out of theft
Two thoughts:
A. That's only true (to any extent) if you hold the extremely myopic view that 'AI == Generative AI'. For my part I'd posit that "AI" at large is not "intrinsically born out of theft". Not unless you think that linear regression, or a genetic algorithm, etc., inherently involve theft somehow.
B. It's an open question whether or not copyright infringement should be considered "theft" at all. It's curious though, that historically hacker oriented communities tended to lean towards "No" being the answer to that. But the scale at which GenAI affects things may be the reason that sentiment seems to be shifting a bit?
I think it should be fairly obvious by now which form of AI people refer to when they talk specifically about theft. It gets a bit old and repetitive to expand the shorthand in every conversation possible. If people are genuinely curious about other forms of AI, that information is readily available.
When Tesla FSD was in the zeitgeist, theft never entered the discussion, because it was clear that form of AI was not predicated upon theft.
A lot of this discourse is intended for a wider audience than just us technical folk, and so the vocabulary used mirrors that of news headlines. If I were to guess, I would say it distracts from the discourse of the sociopolitical and socioeconomic impacts of generative AI providers and their products.
It is understood which forms of AI are at play here when the discussion revolves around the massive companies, capital investments, datacenters PR campaigns, and mandates required to create and sustain them and their usage.
I (and I assume many others on especially HN) don't consider intellectual "property" as real so there is no "theft" in our minds, so this argument doesn't bear much weight.
Piracy is not theft. If something can be copied infinite times without any effort with broad societal benefit, then it's a moral imperative to do so. The opposite is gatekeeping in the name of monopolistic profiteering and the wealth concentration that the modern broken IP law enforces.
Besides, Anthropic did allegedly buy the ebooks they trained on so it's not like they even did that. It goes both ways though, they should get comfortable with their models getting distilled and opened up for everyone to run however they want. LLMs trained on people's data belong to the people.
Terrible analogy. I leave my house locked because I don’t want people to go in. On a website, I want people to read it, but without disturbing others. A better analogy would be that I open a small art gallery with free entrance and someone sends in an army that barely fits in the room.
Anubis is just a patch that shouldn’t be needed; it annoys regular visitors and prevents people on low-powered devices or with JavaScript disabled from reading the site at all.
I would be very surprised if the ebook license they bought does entail using it for training machines. In fact I'm pretty sure it didn't and I thus do not think they did such a thing in the first place as I credit them with enough legal prowess to know about this.
It's numerous. CEO's lying, ceo-ceo marketing - fire your employees and use AI, environmental impact, social impact, memory/chip shortages, theft of information which has placed a massive burden on site operators assaulted by scraper traffic. I'm sure I'm missing a few but the negatives are real but so long as people get to feel like 10x engineers, it's fine.
Personally, I find AI technology itself super interesting. Plenty of great use cases. However, The current crop of lying thieving assholes running the show make it repulsive.
They're not firing employees to replace them with AI. We're mostly engineers here I think. Does anyone actually believe they're replacing humans with the same AI that we're using in our day-to-days? I don't know about you, but my harnesses absolutely suck without a human driving them and the more knowledgeable the human, the less they suck.
It's obvious they're just using AI as cloud-cover to act like assholes in the typical ways in which they would normally act like assholes.
If one person can become as productive as two or three, why keep the extra one or two employees? You might think that keeping the head count the same means the company can now do more but that is only true if the company has the bandwidth or market to grow into.
AI is in its infancy, it's just learning to crawl. There will be more breakthroughs which will have more serious consequences. Today engineers are safe, holding the AI's hand as it crawls around, bumping into furniture. What happens when it learns to walk, run, and win marathons?
(Assuming that LLM does indeed multiply productivity)
We are likely in for some rough days, as it's much easier to just fire people and maintain the same level of productivity. Musk (arguably) did that with Twitter, even before this started. I was impacted by a post-COVID layoff, myself.
But do you think that once that has leveled out a bit, the bandwidth/market bottleneck you referenced will be identified as the new bottleneck[0]? Like, new businesses will launch, or existing companies will identify new growth areas that they did not have the capacity to move into.
I don't know how to respond to your second paragraph. Looking in that direction is a bit too overwhelming.
[0] I think this was always the problem, not developer productivity
> If one person can become as productive as two or three, why keep the extra one or two employees? You might think that keeping the head count the same means the company can now do more but that is only true if the company has the bandwidth or market to grow into.
If the extra one or two employees are 2x or 3x as productive as they used to be, why would they not be employed? There will be plenty of market to grow into since the gains in productivity are shared throughout the economy.
I think the naive CEO-level reasoning is that one person can get twice as much done with a harness, not that AIs will suddenly become useful while autonomous.
> I don’t think it’s only that. I personally hate AI not because of CEOs and co, but because the tech is intrinsically born out of theft, and is still, to this day, evolving thanks to theft. And that’s even before the ecological considerations.
I also hate it because:
1) Fundamentally, it's about reducing the power of labor (which are what the vast majority of people are) and I know I'm a laborer. This is why the CEOs and wealthy are excited about it.
2) It's about automating the engaging and creative knowledge work, and leaving the humans with manual labor and drudgery.
2.5) It's not even about automating the engaging and creative work well. Code generated with LLMs are "Day 1 Legacy Code" with all sorts of tech debt liabilities from the very first generation. Art made with it is often "good enough" but rarely escapes the uncanny valley and succeeds at its creative goals. Technical writing made with it is "mansplaining as a service", often pompous and confident but low in nutritional value (and factual value). Creative writing made with it is repetitive, rambling, and frequently nonsensical, is terrible at metaphors and similes, is all exposition and almost no narrative arc, is bad at nuance and equally bad at overt messaging.
3) As a human, I don't just hate that C-Suites think they can replace my and my colleagues' creative output with LLMs, I dread the world where LLM-first creative content is ubiquitous because it will be a world of increasingly less substance/nutrition/taste/texture/other human metaphors.
Communication tech/tools enable more people to collaborate. It increases ability for labor that is far away from high value markets to contribute. Same goes for shipping tech wrt physical goods. On the global scale that is empowering the labor class.
Any productivity tool that individual laborers can purchase also (and that still needs the worker) is probably good for labor, overall.
If you're a programmer or (physical) engineer you've been automating other people's labor all this time, it's ironically hypocritical that now that engineers are the ones being automated that they cry foul.
I left one employer in part because I thought some of their automation of almost-but-not-quite minimum wage jobs was unethical to me. In Software, we don't have an industry-wide ethics board that people trust, so the complaints about individual automations remain quieter and personal. It is entirely possible for people in these conversations to not be "hypocritical" in their relationship to this topic versus their personal ethics.
It's definitely hypocritical at the "industry ethics" level, but again we don't have an ethics board and all we have are personal and public opinions. (Arguably this is one of the current problems with AI is that there is no ethics board for software so instead we must debate this in the court of public opinion, such as HN comments.)
Don't even get me started on the socioeconomic considerations!
AGI is a savior figure for the capitalist class. A tech version of the Second Coming, delivering them from the pesky demands of workers, like a living wage or (gasp!) sick leave.
That's why they're all so obsessed with it, it has religious-ideological component to them. When you hear them talk about AGI, there's always this weird eschatological vibe with it.
Unfortunately, they're blinded by their beliefs and can't think things through even one step further. Even if their cyberjesus comes down to them through the machine and replaces all workers, who's gonna buy all their stuff then?
All they're doing in their capitalist zealotry is ringing in the end of capitalism.
> Unfortunately, they're blinded by their beliefs and can't think things through even one step further.
Which is particularly wild and frustrating given how many of their "textbooks" were about the consequences and how bad they can get. The very first tale to create the English word "robot" [1] was about the horrible consequences of building a new class of slave labor. Science fiction has been writing "Robot revolt" and "AI revolt" consequences for more than a century since.
A lot of these TESCREAL [2] types caught up in these religious views of AGI have ignored the actual plots of the books they claim to read for excitement and passion about their visions for the future. To be fair, not understanding your own source books seems pretty common across religions, but seems especially weird in this case how many of these "canon" "Don't Build the Torment Nexus" works are accessible to everyone, in a modern vernacular, have major film adaptations or films based on their themes, and plenty of good Wikipedia summaries of how building the thing destroyed civilization or the world or the galaxy.
> Unfortunately, they're blinded by their beliefs and can't think things through even one step further.
Yes, our new generation of overlords seem to be socially and emotionally stunted and exhibit an alarming naivete about the world. This worries me almost as much as the tech itself. It is impossible to predict the future but in the past when a ruling class completely disregarded the effects of their greed and excess on the wellbeing of society, at some point the bill came due and the consequences for them (and society) were dire.
I blame the disdain for humanities, philosophy and the liberal arts.
Not just with the overlords, but also with our fellow nerds. We're all so busy trying to see if we can build something that we don't stop and think if we should build something, what consequences that might have or what history has taught us.
Theres a reason those fields of study are important.
Everything you said plus now these out-of-touch and incredibly rich CEOs are shoving it in our faces that they are determined to take away our income while pocketing even more riches.
IMHO this is a very reasonable price. We have been accustomed to having free software, but the reality is, nothing is free. I don’t mind paying for things that are worth it.
Furthermore the app stays fully open-source, the gates can be removed easily if you really don’t want to pay.
(But I’m guilty of it myself… through Kagi! I postfix the query with an interrogation point and boom, I have my answer. I only do it for trivial things, or things I can verify easily, though.)
in terms of cost, maybe. in terms of my effort, I dunno. but you're right, it's environmentally unsound. I don't usually compute with that lens, though.
reply