No, we aren't. There's still a chance that we will enter a new ai cooling if models continue to demand more GPU and energy to become more effective, but if the same rise in quality persists through GPT 5, 6 and 7, our world will be fundamentally changed, both to the better through a massive increase in knowledge and engineering capabilities, and with huge risks to life and freedom of humanity.
People are bad at predicting the impact of technical changes. Particularly when that change is exponential. When it comes to AI, we've of course got a century plus of science fiction material made by people who tried really hard to imagine the future. It ranges from very dystopian to very utopian. And recently, HN and the rest of the internet has been obsessing over AI a lot.
My personal view is that we're probably all missing the point in some way. And of course I'm probably missing it too. It's like predicting that a few hundred main frame should just about cover the whole market in the nineteen forties. That's where we are with AI today. We have a handful of companies capable of building somewhat useful AI models. And a lot of under resourced tinkerers trying to do stuff with it.
However, a point I like to make regarding work is that we already live in the future. People are worried about losing their jobs not realizing that they've already lost that. By that I don't mean they are out of work. But both the nature of and the amount of work we do to survive has changed massively since the industrial revolution. We keep ourselves busy and it earns us a living. But it mostly isn't all that relevant or important what we do.
Most of us here on HN are doing jobs in software, investments, or related things that have relatively little to do with providing food, shelter, or safety for people. It's all interesting and fun and probably a little bit hedonistic. If I look in the mirror, I can't say with a straight face that what I do is particularly important or relevant. It's more about me doing something interesting (to me) that happens to convert into a weird thing called money that I can convert into stuff I need or want. People are now having serious discussions about having a four day work week. And we just came out of a pandemic where startlingly large percentage of people vastly reduced their economic activity without our economy grinding to a halt. A lot of busy people that suddenly weren't so busy. It did not matter.
It used to be that people worked until they were to old and broken to work. If they didn't they'd starve. For most people meant 6 day work weeks, long shifts, not a lot of leisure time, etc. You'd start work as soon as you learned to work people died young on average. It still means that for some people in developing markets and even some developed markets that maintain a class society (like the US). However that's increasingly less true for a lot of people. There are a lot of people working in the services industry. More than in things like farming or manufacturing.
I had the exact same thoughts today regarding the science fiction material. Almost every single viewpoint I see on here is colored by that lens. I get there is good reason for that but on the other hand, very little is actually similar to the sci-fi material we read about. Yes there are parallels but if you look at it objectively, which isn't actually that easy sometimes, we actually do live in reality, not the movies. I really sympathize with your point that we have very little actual idea where it goes.
Very well put. It's crazy to me that I, a software engineer, can earn 10x more from working a very chill 8 hour day in my pyjamas from home writing some software that's interesting to me but not very important for the world, while my buddy delivers letters and packages for minimum wage on his bike.
well at the surface level it might seem a bit strange, but when thinking a little bit about it, it makes total sense. your buddy can simply only deliver so many packages and letters in a day, impacting a few hundred people or so. Sure, its valuable, and required, but still.
You can write software that millions that each are enriched as much, possibly more than the package delivery your friend does. As such, you simply can command a much higher salary
How many software engineers really write software that is as essential as a working postal service or working food services? Maybe 0.01% of all software engineers? I know that I'm not one of them.
Sure, I understand the macroeconomic reason. But goes to show that maybe some jobs shouldn't just be paid solely based on supply and demand. We can do it because for my friend it's better than not having a job, but it doesn't mean it's the right thing to do as a society.
> It used to be that people worked until they were to old and broken to work. If they didn't they'd starve. For most people meant 6 day work weeks, long shifts, not a lot of leisure time, etc.
Right, this is why we don't have thousands of unemployed horse drawn buggy drivers milling about at street corners. The new jobs created by automation far exceed, in number and quality, the old jobs that are deprecated. Though it's fairly rare for an old job to completely go away, mostly they are transformed. We still have truck and taxi drivers, but now they can transport far more goods and people more rapidly over much greater distances.
I sometimes wonder what I would have done I didn't work in IT. Since I programmed my first computer at aged 13 there's really nothing else Ive wanted to do.
that's an interesting example, because they too are currently in line to be automated away, and that's millions of people. Unlike going from horses to cars, these jobs will go from cars to nothing.
You’re not getting paid to be kept busy, and you’re not getting paid because what you do is interesting and fun. You’re getting paid because you provide real value to someone.
You are free to vastly reduce your own economic activity and the value it brings with it. Then see how it affects your own ability to provide food, shelter, or safety to your own household.
Work is not some made up game we all just decided to play because we’re bored. Someone needs to harvest the crops, build the houses, and protect from harm. None of this comes for free, and we all have to pay for it somehow.
> You’re getting paid because you provide real value to someone.
Often you're getting paid because you provide potential future value to someone. Meaning that plenty of projects whole teams are paid for eventually get canceled with very little or zero ROI.
"AI storm", journals really don't know what bullshit to put out anymore.
Just like all previous AI bubbles, this is a bubble. We'll soon realize that unless you're writing a book, GPT is pretty much useless, and the hype will die down like it always has in the past.
To think that AI who's only capability is to generate text can have any significant impact on anything is simply ridiculous.
Words accomplish something, certainly far more than they cost to produce, otherwise why did we start doing it and why haven't we stopped. Your hands alone won't build an empire in a thousand lifetimes. It won't even pave a complete road. For that you need to pitch a compelling argument and influence a large group to go along with your vision. These particular words here and now probably won't do much of anything. The odds of at least some of them in my corpus of hn comments going on to impact something is great enough to justify the meager cost in doing so.
Eh, I'm not someone who is easily impressed, but I'm going to say LLMs are transformative (ba-dum-dum-sheeeesh).
I use copilot, it definitely saves me time. At the moment just minor things, but there's no arguing, there's simply no doubt.
ChatGPT will transform forum comments. Soon human written messages like this one will be swamped by machine-assisted argue-bots. And sales-bots.
Other LLMs are doing images, which will be much the same. You can illustrate your book automatically, pretty much now. You can generate marketing images, now.
Everything creative will be affected. Music, books, movies.
everything creative where true creativity doesnt matter :)
sure, it will illustrate a book, many presentations, all sorts of stuff, but it will not create novel things as a human does. Sure, it can produce a smattering of what it was trained on, it will not produce new original well thought out masterpieces.
think about how it works for a moment, sure, its great, very nice, simple chores can be done with some reasonable effectiveness. but people are clearly expecting things that an LLM simply isnt meant or capable of doing
Had this discussion the other day, never got an answer from the thread.
What is true creativity? What is a novel thing?
If we establish that it isn't merely a mix of existing things, what is it?
I don't have an answer other than actually maybe these things it makes actually are novel.
If you have another take on it I'd be interested to hear why nothing it makes is novel, and nothing from the LLM family of algos can ever make something that qualifies?
Creativity comes from insight. This thing has no insight, the best it can do is spit out something that might fit, and you will need to decide if it got lucky.
> I use copilot, it definitely saves me time. At the moment just minor things, but there's no arguing, there's simply no doubt.
It's an autocomplete, hardly transformative.
> ChatGPT will transform forum comments. Soon human written messages like this one will be swamped by machine-assisted argue-bots. And sales-bots.
You really think people that like to argue online will deprive themselves from the joy of writing the message on their own?
And bots that write marketing comments are hardly new.
Again, hardly transformative.
> Other LLMs are doing images, which will be much the same. You can illustrate your book automatically, pretty much now. You can generate marketing images, now.
None of this is transformative either, it's just the same as the code autocomplete example.
You really think any of the things you have cited will deeply affect society? Please.
It's not autocomplete, we had intellisense two decades ago. And nobody would argue that wasn't useful. Saving time is useful. Thing that merely save time have been transformative. For instance you can walk across London in a couple of hours or you can take the tube. All it does is save time. But that's transformative, it changes what you can do when you don't need the same amount of time to do it.
Marketing and especially political comments are about to swamp every digital space. Soon you will not know what is real and what is a puppet. This is without a doubt transformative. Look at the relatively small wave caused by the Cambridge Analytica thing. This thing will create a tsunami of BS sweeping us all away. I'll admit it hasn't happened yet, what we have now is just the beginnings.
Most software developers are doing nothing else than generating text - code. And if you take into account only this industry it definitely had significant impact on society just within a few decades.
Current AI is not good enough to replace software developers but in 10-20 years maybe will be good enough to replace just every second one == 50% of devs
It will, if anything, cull the bottom performers. It's essentially an incremental improvement on a script which searches stack overflow for a query and copy/pastes the first result into your program. Having tried it on more interesting systems-level work it does little to nothing to augment productivity.
Me: "How many jobs will be lost due to new AI technology within the next ten years?"
Bing: "According to one source, worldwide, a billion people could lose their jobs over the next ten years due to AI, and 375 million jobs are at risk of obsolescence from AI automation. However, it’s important to emphasize that there is no shared agreement on the expected impacts on the workforce or economy."
It's important to note this is Bing quoting one source, presumably an attention-craving pundit on the internet throwing out absurd claims in order to get attention. This is not actually a prediction calculated by an AI.
Losing your job is only a problem because we live under a system that only allows people to live if they allow themselves to be exploited by a parasite. If things were right there'd be no concept of "losing your job".
Even with sci-fi levels of worst case scenario and it turns out to be an AI that goes sentient and takes control of the world... can't be too much worse than the people in power now. At least it'll act more logically :)
Anything less than that seems like a fun change to be living through. This is our industrial/technological revolution. They turned out alright eventually. Might be time to bring up UBI again but we'll figure it out either way. Humans survive, it's what we do.
Here is the thing: The industry will sugar-coat the layoff effect by cutting down work days - not laying off a significant crowd.
Initially it will seem like a humane thing to do without cutting down salaries BUT we'll feel the real effect when the salaries are not increased to match the inflation.
You will work less, produce more (with AI), paid less (in terms of purchasing power).
Products that AI can touch will get cheap so purchasing power on digital products will stay the same to keep your entertained BUT land, housing, food... hard things will get harder.
I don't buy it. The idea that automation reduces overall employment needs to die in a ditch. It's just not true. Look around you, we live in a world that's automated up to the eyeballs compared to any previous generation, yet employment is as high as it's ever been. In fact it's substituted for domestic labour and has allowed generations of women to go to work and build careers, and hundreds of millions of rural labourers have been drafted into industrial work forces in my lifetime. As automation has expanded, the demand for labour has grown voraciously.
These LLMs are incredibly smart compared to humans in some ways, but they're incredibly stupid in other ways. They're just another tool, and tools need users. Hopefully they will be another big force multiplier in knowledge based industries, but this will increase demand for knowledge workers, because they will become more valuable due to being more productive from having better tools.
If better knowledge tools reduced demand for the people using them, Moore's Law and the availability of ever more sophisticated software would have ravaged employment in the IT industry. But that's just not how that works.
It's based on the assumption that when a worker becomes more productive, that the only effect that has is reducing the amount of work there is for others to do, but that's an incredibly naive view. In fact it also dramatically reduces the cost of that work being done and increases the capacity of the organisation to do work. This increases demand for that work, it makes that labour more skilled, and increases profits thus stimulating investment. Yes there can be localised negative effects in some cases, sure. Ive seen some of those myself, but any economic realignment can have it's pains. Taking into account all the positives it increases the value of the economic activity overall.
Decision-making and reasoning was not automated in any era before. "Automation is just fine because it was fine until now" is an old world approach.
We are not only tool users at this point. We started to change through our tool use. We think we have guardrails against AI but our brains don't have any guardrails. It will get harder and harder to get motivated to learn any domain deeply. Our brains can't cope with these short-term job definitions in intellectually demanding contexts.
Prompt engineer today, automated next year. Take a look at the fate of indie AI startups (avatar generation is best example of this) and you'll see. Income opportunities are limited to shorter and shorter time windows. Only the big players dominate (with their marketing machine) and will probably pay peanuts after reaching dominance.
Of course decision making has been automated, that's literally what computers do. This will just streamline the process of pushing it into more problem domains. Everyone thinks their job is special and that automating any part of it will be the end of the world.
Eventually, with full human level AI yes, sure, at that point we'd not be augmenting human intelligence but genuinely substituting for it. We are a long, long, long way away from that though.
I edited my post above, so you may not have read the last paragraph.
I understand our perspectives differ because I think we are right at the edge of fast take-off and you outlined your view as slow take-off.
Computers were bicycles for our minds, applying our previously formulated solutions to scale. I agree I did not explicitly clarify that I was meaning high-level decision-making getting automated.
“It’s not the end of the world” begs for definition of what end of the world looks like. A meteor is not end of the world apparently but it killed the dinosaurs.
Well, I'll put my big mouth on the line and make a prediction. 12 years ago we all thought self-driving cars would be with us by now, and my wife genuinely wondered whether our kids would need to learn how to drive. My youngest passed her test a few months ago.
There's a significant low hanging fruit effect in AI domains. We rapidly reach a 90% solution, but that last 10% is absolutely critical, and virtually impossible to attain. I think these LLMs will follow the same trajectory. They will quickly become incredibly valuable or even indispensable tools, but the alignment problem and over-fitting will come to limit their reliable applicability in critical edge cases that make skilled human oversight essential. This will relegate them to the role of productivity enhancement rather than genuine labour substitution.
This assumes that humans in knowledge-based industries will maintain a significant cost-effective edge over AIs in some economically relevant skill domain. There's no reason to believe that will still be true in, say, 10 years.
This assumes that a human with a chisel will maintain a significant cost effective edge over a CNC machine...
It's the same thing. These things are still just tools that need to be used by humans. Fair enough, I can't really see 10 years into the future, but I don't think anyone can, and there's nothing about these tools we're developing today that makes me think they're immune to the economic effects of automation throughout the last 300 years.
If my memory serves, in past industrial revolutions, employment eventually caught up with automation. It did take one generation, though (~40 years in the 19th century).
That being said, I won't attempt to predict anything. Just checking what I can do to build a ChatGPT-proof resume :)
No one is employing people for shits and giggles(maybe the government). Every employee exists because their employer believes they will add value to the company.
The "work less" notion seems based on the idea that the only kind of work we can do is to keep the 20th century going. Once that infrastructure is automated, there's nothing left to do. That's just not true.
AI can help us explore space, mine asteroids, discover new worlds. It can eradicate diseases. It can create portable tests for transmissible diseases. It can help us rebuild cities into green spaces with fruit trees, gardens, flowers.
We have this incredible new technology that can let us reshape the world into an incredible future. Why would we settle for anything less than Star Trek?
The possibilities aren't ending with AI; they're just beginning.
Preach! Need more comments like this—about the cool new shit that will happen in the real world thanks to technological advancement. A very Musk-ian point of view.
Real talk though, we need another frontier to give (now globalized) human society something to look forward to. Colonization of space is pretty much the only place to go, and fortunately it doesn’t carry the moral baggage of displacing other life forms (at least as far as we know).
It can also help us spread propaganda, con people at massive scale, destroy public discourse spaces with spam, create deep fakes to slander people, start and prosecute wars, etc.
History shows that both good and bad things will be done with AI. We shouldn’t be doomers but we shouldn’t be naive either.
That is a very nice point of view, thanks for sharing it. I'll try and keep it somewhere in my mind if the dark days come :)
Observing how AI (and, let's face it, pretty much every other tool at our disposal) has been used so far, I suspect that there will be quite a few steps before we get there. First, a mad rush to use AI for just about anything, just in case it sticks. In particular:
- attempting to build addictive technologies with AI (toys? casinos? conspiracy theories? definitely porn);
- attempting to speculate with AI;
- disrupting as many existing industries as possible with AI, whether it's sustainable or not;
- also, pretend AI (look at you, Parler/Hive);
- undoubtedly, AI for cyberwar;
- and of course, any number of confidence tricks based on AI.
My hope, at this stage, is that we'll hit a sufficient number of use cases in which AI is actually useful (AI for green urban planning?) before we hit either a new AI winter (which may be prompted by some catastrophe in which AI will have played some role).
IDK, if my job says "Hey, 4 day work week but you're not getting the yearly increase," I think I would be okay with that honestly. I think a lot of people would actually.
In Iain M Banks's Culture novels, the Minds seem to have an appropriate role in society. Gonna read up on them for some perspective on current developments.
Optimism for something so incredibly unlikely is delusional thinking. What would you say about a gambler who is “optimistic” about their chances of winning the lottery?
Hum. Minds are bored out of their minds and need to simulate entire universes to stave off boredom. Also, they help puny human beings once in a while, because they're basically somewhat benevolent.
It might take us a few steps before we reach that future :)
That’s not surprising. Your average Democrat is a “liberal”. The Democratic Party is center or (arguably) center right, and that is born out in the political views of the average member of the party. They have been trained to think that left labor reforms would be a disaster for the country by the media.
There is no truly left wing party in the US with any significant political power. Since the Democrats are not “the right” they must therefore be “the left”, which is how they’re referred to by essentially all media. This is a useful rhetorical maneuver because it paints left wing policies that most any rational person in Europe would view as reasonable as completely nuts because they’re so far out of the mainstream spectrum of discourse. See the Sanders campaign’s coverage on NPR for an example of liberal media going to town on a left wing candidate. Whenever I talk to someone from Europe about this they find this whole situation totally baffling (“why didn’t the democrats support Sanders? He’s a reasonable left wing candidate”).
Over here liberal means as in liberty (but not libertarian). As in liberal economics, which means free market economics. Sociologically it means promoting equity and individual freedom, but not extremely so like libertarianism. Here in the UK it's pretty much synonymous with centrism.
The "true" outcome will be a combination of all of the above, or even more. It is an emergent phenomenon, not a logically derived outcome. Which is also why it is unpredictable.
Here's a take I havent heard yet. ChatGPT understands language and we've got troves of content archived/realtime that can't be reviewed manually. Now there's a tool that can understand text. Govs must be chomping at the bit to ask, "Hey ChatGPT, does any of this content seem illegal?" where illegal is a broad term, especially the authoritarian ones. Seems like an amazing surveillance tool to flag things for human review.
I would assume some spy agencies already evaluate this. They get too much material to review manually and ChatGPT could really help identifying people they should look at closer.
If it can be used that way, I guess we can expect export restrictions on free usage and code/weights.
Affordable tools for ~this purpose have existed for years. Yes, LLMs make these tools more powerful, but they don't feel like game changers for this purpose.
Source: I've been in a position in which companies tried to sell me similar tools (to detect e.g. bullying).
Not even OpenAI seems to know. They put some perfunctory disclaimer on economics into their risks section of the GPT4 report, but their lack of understanding the consequences didn't seem to prevent them from pushing this development, so uncharted territory it is.
It should anyways be up to governments to reign in this technology for the benefit of humanity, though I'm afraid this won't happen.
> She said artificial intelligence could go in one of two directions over the next 10 years. [...] AI development is focused on the common good, with transparency in AI system design and an ability for individuals to opt-in [...] Ms Webb's catastrophic scenario involves less data privacy, more centralisation of power in a handful of companies
Umm... so are we just not going to mention the elephant in the room at all? I mean, I know everyone finds “the singularity is near!” people annoying, but there’s not any mention at all of some type of existential risk? The “catastrophic scenario” relates to things like privacy and fairness?
Someone commented in another article about AI safety being composed of two types, those focused on “AI alignment” and those focused on “AI ethics”. I increasingly find that the latter group seems to be putting the cart before the horse.
Hypothetically, if you were to poll the research scientists at OpenAI (anonymously) on which type of AI safety they are more concerned about, I would be really curious what the results of that survey would look like.
AI is not in and of itself an existential threat, and is unlikely to be one within our lifetimes. The singularity, in this sense, is not near, and state-of-the-art machine learning does not resemble general intelligence.
Focusing on the misuse of this tool by people is the most prudent worry to address.
Lets assume that current AI type models are capable of approaching the capability of human type neural networks.
So yes, AI will be very disruptive to our lives. It will start with a narrow set of capabilities, but will widen as it becomes really good within those constraints.
I think we are mostly scared about the capabilities of AI that push us to question, what is my worth then, as a human being? Do I still have worth? And, what is really the point of all this?
This “revolution” will push us to look inward and find out what and who we really are. What is really valuable? And most importantly, what is really worth struggling, fighting, striving for, in the context of the human race.
Our subjective human experience and what is valuable to us as humans, still remains out of reach for AI, and I think it will never become within reach. Because AI will never share with us the subjective human experience.
95 comments
[ 2.1 ms ] story [ 157 ms ] threadI'm scared and excited.
And GPT 10 will be the last release because Microsoft.
Was watching a talk on YouTube about how "sparsity" can potentially change the game, so that GPU compute isn't needed.
[0] https://www.youtube.com/watch?v=0PAiQ1jTN5k&t=2240s&ab_chann...
My personal view is that we're probably all missing the point in some way. And of course I'm probably missing it too. It's like predicting that a few hundred main frame should just about cover the whole market in the nineteen forties. That's where we are with AI today. We have a handful of companies capable of building somewhat useful AI models. And a lot of under resourced tinkerers trying to do stuff with it.
However, a point I like to make regarding work is that we already live in the future. People are worried about losing their jobs not realizing that they've already lost that. By that I don't mean they are out of work. But both the nature of and the amount of work we do to survive has changed massively since the industrial revolution. We keep ourselves busy and it earns us a living. But it mostly isn't all that relevant or important what we do.
Most of us here on HN are doing jobs in software, investments, or related things that have relatively little to do with providing food, shelter, or safety for people. It's all interesting and fun and probably a little bit hedonistic. If I look in the mirror, I can't say with a straight face that what I do is particularly important or relevant. It's more about me doing something interesting (to me) that happens to convert into a weird thing called money that I can convert into stuff I need or want. People are now having serious discussions about having a four day work week. And we just came out of a pandemic where startlingly large percentage of people vastly reduced their economic activity without our economy grinding to a halt. A lot of busy people that suddenly weren't so busy. It did not matter.
It used to be that people worked until they were to old and broken to work. If they didn't they'd starve. For most people meant 6 day work weeks, long shifts, not a lot of leisure time, etc. You'd start work as soon as you learned to work people died young on average. It still means that for some people in developing markets and even some developed markets that maintain a class society (like the US). However that's increasingly less true for a lot of people. There are a lot of people working in the services industry. More than in things like farming or manufacturing.
https://www.youtube.com/watch?v=nyu4u3VZYaQ
You can write software that millions that each are enriched as much, possibly more than the package delivery your friend does. As such, you simply can command a much higher salary
I'm afraid there's no "soft landing" here. Between AI and climate change our future doesn't look so good. Certainly doesn't look peaceful.
Hm, depends, not necessarily, just one of many references: https://groups.csail.mit.edu/mac/users/rauch/worktime/hours_...
I sometimes wonder what I would have done I didn't work in IT. Since I programmed my first computer at aged 13 there's really nothing else Ive wanted to do.
that's an interesting example, because they too are currently in line to be automated away, and that's millions of people. Unlike going from horses to cars, these jobs will go from cars to nothing.
You are free to vastly reduce your own economic activity and the value it brings with it. Then see how it affects your own ability to provide food, shelter, or safety to your own household.
Work is not some made up game we all just decided to play because we’re bored. Someone needs to harvest the crops, build the houses, and protect from harm. None of this comes for free, and we all have to pay for it somehow.
Often you're getting paid because you provide potential future value to someone. Meaning that plenty of projects whole teams are paid for eventually get canceled with very little or zero ROI.
Or as some motivational poster put it: “It may very well be that the purpose of yore life is to set a warning example to others.”
Just like all previous AI bubbles, this is a bubble. We'll soon realize that unless you're writing a book, GPT is pretty much useless, and the hype will die down like it always has in the past.
To think that AI who's only capability is to generate text can have any significant impact on anything is simply ridiculous.
Neither mistake by the journo would be possible if they had any substantive experience programming.
We are all just text here. So far as I know, by your metric, you never have significant impact on anything.
I use copilot, it definitely saves me time. At the moment just minor things, but there's no arguing, there's simply no doubt.
ChatGPT will transform forum comments. Soon human written messages like this one will be swamped by machine-assisted argue-bots. And sales-bots.
Other LLMs are doing images, which will be much the same. You can illustrate your book automatically, pretty much now. You can generate marketing images, now.
Everything creative will be affected. Music, books, movies.
sure, it will illustrate a book, many presentations, all sorts of stuff, but it will not create novel things as a human does. Sure, it can produce a smattering of what it was trained on, it will not produce new original well thought out masterpieces.
think about how it works for a moment, sure, its great, very nice, simple chores can be done with some reasonable effectiveness. but people are clearly expecting things that an LLM simply isnt meant or capable of doing
So it'll be fine replacing most of Hollywood, videogames, and pop music then...
What is true creativity? What is a novel thing?
If we establish that it isn't merely a mix of existing things, what is it?
I don't have an answer other than actually maybe these things it makes actually are novel.
If you have another take on it I'd be interested to hear why nothing it makes is novel, and nothing from the LLM family of algos can ever make something that qualifies?
It's an autocomplete, hardly transformative.
> ChatGPT will transform forum comments. Soon human written messages like this one will be swamped by machine-assisted argue-bots. And sales-bots.
You really think people that like to argue online will deprive themselves from the joy of writing the message on their own?
And bots that write marketing comments are hardly new.
Again, hardly transformative.
> Other LLMs are doing images, which will be much the same. You can illustrate your book automatically, pretty much now. You can generate marketing images, now.
None of this is transformative either, it's just the same as the code autocomplete example.
You really think any of the things you have cited will deeply affect society? Please.
Marketing and especially political comments are about to swamp every digital space. Soon you will not know what is real and what is a puppet. This is without a doubt transformative. Look at the relatively small wave caused by the Cambridge Analytica thing. This thing will create a tsunami of BS sweeping us all away. I'll admit it hasn't happened yet, what we have now is just the beginnings.
Happy to say LLMs are a big deal.
Current AI is not good enough to replace software developers but in 10-20 years maybe will be good enough to replace just every second one == 50% of devs
Bing: "According to one source, worldwide, a billion people could lose their jobs over the next ten years due to AI, and 375 million jobs are at risk of obsolescence from AI automation. However, it’s important to emphasize that there is no shared agreement on the expected impacts on the workforce or economy."
A billion people losing their job would be massive.
Even with sci-fi levels of worst case scenario and it turns out to be an AI that goes sentient and takes control of the world... can't be too much worse than the people in power now. At least it'll act more logically :)
Anything less than that seems like a fun change to be living through. This is our industrial/technological revolution. They turned out alright eventually. Might be time to bring up UBI again but we'll figure it out either way. Humans survive, it's what we do.
Rule 1 of life: Evolve or die.
Initially it will seem like a humane thing to do without cutting down salaries BUT we'll feel the real effect when the salaries are not increased to match the inflation.
You will work less, produce more (with AI), paid less (in terms of purchasing power).
Products that AI can touch will get cheap so purchasing power on digital products will stay the same to keep your entertained BUT land, housing, food... hard things will get harder.
These LLMs are incredibly smart compared to humans in some ways, but they're incredibly stupid in other ways. They're just another tool, and tools need users. Hopefully they will be another big force multiplier in knowledge based industries, but this will increase demand for knowledge workers, because they will become more valuable due to being more productive from having better tools.
If better knowledge tools reduced demand for the people using them, Moore's Law and the availability of ever more sophisticated software would have ravaged employment in the IT industry. But that's just not how that works.
It's based on the assumption that when a worker becomes more productive, that the only effect that has is reducing the amount of work there is for others to do, but that's an incredibly naive view. In fact it also dramatically reduces the cost of that work being done and increases the capacity of the organisation to do work. This increases demand for that work, it makes that labour more skilled, and increases profits thus stimulating investment. Yes there can be localised negative effects in some cases, sure. Ive seen some of those myself, but any economic realignment can have it's pains. Taking into account all the positives it increases the value of the economic activity overall.
We are not only tool users at this point. We started to change through our tool use. We think we have guardrails against AI but our brains don't have any guardrails. It will get harder and harder to get motivated to learn any domain deeply. Our brains can't cope with these short-term job definitions in intellectually demanding contexts.
Prompt engineer today, automated next year. Take a look at the fate of indie AI startups (avatar generation is best example of this) and you'll see. Income opportunities are limited to shorter and shorter time windows. Only the big players dominate (with their marketing machine) and will probably pay peanuts after reaching dominance.
Eventually, with full human level AI yes, sure, at that point we'd not be augmenting human intelligence but genuinely substituting for it. We are a long, long, long way away from that though.
I edited my post above, so you may not have read the last paragraph.
Computers were bicycles for our minds, applying our previously formulated solutions to scale. I agree I did not explicitly clarify that I was meaning high-level decision-making getting automated.
“It’s not the end of the world” begs for definition of what end of the world looks like. A meteor is not end of the world apparently but it killed the dinosaurs.
There's a significant low hanging fruit effect in AI domains. We rapidly reach a 90% solution, but that last 10% is absolutely critical, and virtually impossible to attain. I think these LLMs will follow the same trajectory. They will quickly become incredibly valuable or even indispensable tools, but the alignment problem and over-fitting will come to limit their reliable applicability in critical edge cases that make skilled human oversight essential. This will relegate them to the role of productivity enhancement rather than genuine labour substitution.
It's the same thing. These things are still just tools that need to be used by humans. Fair enough, I can't really see 10 years into the future, but I don't think anyone can, and there's nothing about these tools we're developing today that makes me think they're immune to the economic effects of automation throughout the last 300 years.
That being said, I won't attempt to predict anything. Just checking what I can do to build a ChatGPT-proof resume :)
The "work less" notion seems based on the idea that the only kind of work we can do is to keep the 20th century going. Once that infrastructure is automated, there's nothing left to do. That's just not true.
AI can help us explore space, mine asteroids, discover new worlds. It can eradicate diseases. It can create portable tests for transmissible diseases. It can help us rebuild cities into green spaces with fruit trees, gardens, flowers.
We have this incredible new technology that can let us reshape the world into an incredible future. Why would we settle for anything less than Star Trek?
The possibilities aren't ending with AI; they're just beginning.
Real talk though, we need another frontier to give (now globalized) human society something to look forward to. Colonization of space is pretty much the only place to go, and fortunately it doesn’t carry the moral baggage of displacing other life forms (at least as far as we know).
History shows that both good and bad things will be done with AI. We shouldn’t be doomers but we shouldn’t be naive either.
Observing how AI (and, let's face it, pretty much every other tool at our disposal) has been used so far, I suspect that there will be quite a few steps before we get there. First, a mad rush to use AI for just about anything, just in case it sticks. In particular:
- attempting to build addictive technologies with AI (toys? casinos? conspiracy theories? definitely porn);
- attempting to speculate with AI;
- disrupting as many existing industries as possible with AI, whether it's sustainable or not;
- also, pretend AI (look at you, Parler/Hive);
- undoubtedly, AI for cyberwar;
- and of course, any number of confidence tricks based on AI.
My hope, at this stage, is that we'll hit a sufficient number of use cases in which AI is actually useful (AI for green urban planning?) before we hit either a new AI winter (which may be prompted by some catastrophe in which AI will have played some role).
It might take us a few steps before we reach that future :)
Right: "AI is too biased against conservative views."
Worried consumers: "The junk level on the Internet will rise from 90% to 99.99%"
Venture capitalists: "We will make tons of money."
Business owners: "Soon I can fire more people."
Worried employees: "What if AI will replace me."
AI zealots: "AI will just help people do their jobs better and easier."
Government spokespeople: "Some jobs might be at danger but we are working hard to create new jobs."
HN commenters: "We should look more at philosophical implications of AI."
SEO spam farm owner: "Hurray!"
Gamers: "What if NVidia will make AI chips instead of GPUs?"
Stoicists: "We will survive in one way or another."
There is no truly left wing party in the US with any significant political power. Since the Democrats are not “the right” they must therefore be “the left”, which is how they’re referred to by essentially all media. This is a useful rhetorical maneuver because it paints left wing policies that most any rational person in Europe would view as reasonable as completely nuts because they’re so far out of the mainstream spectrum of discourse. See the Sanders campaign’s coverage on NPR for an example of liberal media going to town on a left wing candidate. Whenever I talk to someone from Europe about this they find this whole situation totally baffling (“why didn’t the democrats support Sanders? He’s a reasonable left wing candidate”).
The "true" outcome will be a combination of all of the above, or even more. It is an emergent phenomenon, not a logically derived outcome. Which is also why it is unpredictable.
If it can be used that way, I guess we can expect export restrictions on free usage and code/weights.
Source: I've been in a position in which companies tried to sell me similar tools (to detect e.g. bullying).
It should anyways be up to governments to reign in this technology for the benefit of humanity, though I'm afraid this won't happen.
Umm... so are we just not going to mention the elephant in the room at all? I mean, I know everyone finds “the singularity is near!” people annoying, but there’s not any mention at all of some type of existential risk? The “catastrophic scenario” relates to things like privacy and fairness?
Someone commented in another article about AI safety being composed of two types, those focused on “AI alignment” and those focused on “AI ethics”. I increasingly find that the latter group seems to be putting the cart before the horse.
Hypothetically, if you were to poll the research scientists at OpenAI (anonymously) on which type of AI safety they are more concerned about, I would be really curious what the results of that survey would look like.
Focusing on the misuse of this tool by people is the most prudent worry to address.
So yes, AI will be very disruptive to our lives. It will start with a narrow set of capabilities, but will widen as it becomes really good within those constraints.
I think we are mostly scared about the capabilities of AI that push us to question, what is my worth then, as a human being? Do I still have worth? And, what is really the point of all this?
This “revolution” will push us to look inward and find out what and who we really are. What is really valuable? And most importantly, what is really worth struggling, fighting, striving for, in the context of the human race.
Our subjective human experience and what is valuable to us as humans, still remains out of reach for AI, and I think it will never become within reach. Because AI will never share with us the subjective human experience.