▲ 100 points
back
104 comments
This is roughly how I’m starting to think about it. This may be doomsday or it may be an amazing opportunity. I don’t know. If it’s doomsday, it doesn’t matter whether I act as if it’s doomsday or if I look for opportunities. Either way I’m fucked. But if I act as if it’s doomsday, but it’s actually an opportunity, I’ll miss all the possible opportunities that come up. So I don’t see any downside in looking at this as an opportunity. I may as well.
And once I started to look at it that way, I started seeing all these potential ways to use AI to do what I do now, only better and more easily. So I’m going to keep looking for opportunities and going after them. If I’m going down, I’m going down swinging.
Pascal's wager for a different time and place. I have never thought about it that way. That's interesting.
Pascal’s wager is full of holes[1]. I agree with you that these ideas are similar, and it’s because of that that your parent comment deserves scrutiny rather than blind acceptance. For example:
> If it’s doomsday, it doesn’t matter whether I act as if it’s doomsday or if I look for opportunities. Either way I’m fucked.
There are different degrees of “being fucked”. Not everyone is disadvantaged the same in a doomsday scenario: it does not literally happen in a day and it does not necessarily spell doom for those who caused it or are outside their sphere of influence. It follows that someone who looked for alternatives rather than fully embracing the system might stand a better chance.
Furthermore, a scenario which is not yet inevitable may become so because people think it is inevitable and give up or act in a way that reinforces it. That’s the definition of a self-fulfilling prophecy.
Note I’m not advocating a specific approach to this case. My aim is to highlight that while the initial argument seems logically airtight, it is deeply flawed and you shouldn’t take it as the true solution.
But it can be also something that is not opportunity nor doomsday.
Then considering it opportunity one spends resources with not gains.
> If it’s doomsday, it doesn’t matter whether I act as if it’s doomsday or if I look for opportunities
If everyone acted like this all the time, there'd be no point in unionizing or voting. If you suspect it might be doomsday, you should join other people who feel the same and do something about it.
It's probably too late to stop it from filling the internet with misinformation and spam, but it's not too late to stop it from taking the jobs of the people who produced the training sets (i.e. writers, open-source coders, etc.)
Since the training data is sourced from the web and from users, I've been wondering how difficult it would be to collectively poison these systems with junk data. I'm guessing it wouldn't take that many people to do it.
Here's what I think. At first, it will be used as a tool to augment devs. But quickly it will learn how to replace a lot of devs. Which ones? What's the most common code on Github? Probably YAFRA (Yet another fucking react app). It's going to learn on those first and replace those first. If you write code against/for bespoke hardware, for example, (where there are essentially zero projects that the AI can learn from), you're probably going to have another couple decades before you're replaced.
Not saying one is harder than the other or criticizing anyone's skills, etc. Just looking at what the AI has available to learn from.
It’s here, it’s real, and we have to deal with it as a society.
The goal of the “AI Ethics” professional class to have all AI be approved by well-compensated PhDs in corporate labs has always been a paternalistic farce.
Common people deserve access so they can figure out what their lives will be like now, and then we can make collective decisions about where we go.
> I think one of the things that really separates us from the high primates is that we’re tool builders. I read a study that measured the efficiency of locomotion for various species on the planet. The condor used the least energy to move a kilometer. And, humans came in with a rather unimpressive showing, about a third of the way down the list. It was not too proud a showing for the crown of creation. So, that didn’t look so good. But, then somebody at Scientific American had the insight to test the efficiency of locomotion for a man on a bicycle. And, a man on a bicycle, a human on a bicycle, blew the condor away, completely off the top of the charts.
> And that’s what a computer is to me. What a computer is to me is it’s the most remarkable tool that we’ve ever come up with, and it’s the equivalent of a bicycle for our minds.”
~ Steve Jobs
ChatGPT is motorcycles for our minds.
Fast and fun but dangerous and unforgiving if we don't anticipate potential dangers? :)
I fear it's more like "a butler for our minds".
The more it serves us, the more indulgent, flaccid and incapable we will become. When our butler invitably degrades, we will have forgotten how to serve ourselves. At least with bicycles we were still pedalling.
> When our butler invitably degrades
Why should it? Any more than, say, the global economic system providing us with gasoline at the gas station, which we rely on?
Looking forward to our WALL-E style future.
More like moped, powered, with small wheels, definitely not off-road. Still good if you don't mind going the same way again and again. Or going "in the wild" only to find out someone is already there.
BTW, animals do make tool. Even some insects (ChatGPT told me last night :)).
The point of the bicycle is that it is propelled by and enhances the effort of the human on it.
A motorcycle depends on, and thus is useless without, an external source of fuel.
Even the Apple I needed electricity to be useful.
> questionably reliable flying cars for our minds
> TempleOS is a motorcycle.
> You know what’s pretty crazy? If you lean over on a motorcycle, it crashes. Don’t do that.
‐ Terry A. Davis, https://www.youtube.com/watch?v=Dwgj9UtuDFY
So, here's my big problem with AI: To quote Strong Bad, "I don't trust any device I can't mash Ctrl-Alt-Del on."[0] And the good bits of AI, as it exists currently, are inscrutable NN statistical models living in the cloud (i.e., someone else's computer). You pay for access, drink a verification can, submit your query, and then a miracle occurs[1], and you get your result.
Today, if I want to be a programmer, I can procure all the equipment I need to do some serious programming -- for cheap! I have control over every step of the process. If creating software becomes something that you use AI to do rather than do yourself, it will require surrendering much of the process to mysterious cloud entities, like William Gibson's electronic voodoo gods floating out on the edges of cyberspace. Unless we have open-source-equivalent AI models that we can download, play with, run on our own hardware that actually give good results. I think I will wait for those to become available on hardware I actually have before faffing about with AI.
[0] https://twitter.com/StrongBadActual
[1] https://www.researchgate.net/figure/Then-a-Miracle-Occurs-Co...
To be fair verification cans are pretty universal these days - not really an AI issue
The key quote:
> [Wozniak] was overjoyed when he learned that the skill that put so much effort into suddenly became massively easier. He wasn't worried about not being able to earn an above-average salary from this anymore, he was happy about all the new cool things he and everyone else would be able to build.
If you want to get things done, then having your skills obsoleted is good.
If you want to get paid to get things done, having your skills obsoleted is concerning.
This is true, but it also ignores the economic reality: your employer wants to get things done, but money is their constraint more than time. If they can replace you, they are obligated to do so.
This is a nice idea, but in economic reality, you won’t be getting anything done if your boss lets you go.
Something else that frames this conversation well is to consider the experience for a non programmer, inspecting the AI results to some query. We view this all through very adept and skilled eyes that know how to code an how to read code. Imagine you don't know anything about programming...
* You don't know what a function is (or a method / class)
* You don't even understand variables, static types, bools/ints/strings
* You have no idea about an entry point such as main, how returns work
* You have no idea about OOO, organising code over multiple files
* You don't know what a socket is, you don't understand I/O, file handling
To top it all, code looks alien to you, you have no idea how to read it.
The only individuals who would be able to bridge the knowledge above, would likely become programmers anyhow.
Adding to your list:
* You don't have experience with basic technical decision making -- e.g. why would you pick Java vs. C vs. Python vs. Go vs. any other language, and how would that decision change if you were building a web app vs. mobile app vs. embedded software?
* You don't have a familiar set of tools and established preferences for the aforementioned decisions -- i.e. if an experienced programmer is building a web app, there are probably specific languages and tools they're inclined to reach for.
* You don't know about application deployment and productionization. How do you get your project somewhere people can use it? You don't know about deploying code to servers or running it in Docker or K8s or PaaS.
* You don't have the practiced logical thought process for things like algorithm design, debugging, figuring out user bug reports, identifying edge cases, predicting user behavior, etc.
* You don't have the experience with things that experienced programmers find "obvious" -- e.g. if you're building a web app, you need an authentication system, registration flow, password reset, etc.
I think the main effort to address this will be LLM-driven low-code solutions. Imagine something like Airtable, but with an integrated LLM assistant that can make changes to your app and write code to implement business rules and automation. This gets rid of problems like choosing your tech stack, deployment, etc. and it minimizes the amount of code the LLM has to create.
I imagine this will have effects comparable to those of spreadsheet software: non-programmers will be able to build custom tools with a newfound ease and some will end up learning a lot of programming skills in the process.
And, much like with spreadsheets, these tools will end up becoming business critical in unexpected ways and will turn out to be a huge pain in the ass to maintain. They'll find that parts of their business logic don't work as expected and that they can't figure out how to prompt the LLM to get it to implement the rules the right way. The users of these tools will basically have to learn some programming anyway or they'll have to hire experienced programmers to figure the problems out for them.
The biggest effort when I write code is to make it readable for other developers. The day where non tech people start using AI to produce non critical software and start interacting with code as a black box, simply telling the AI what they want and pointing the edge cases to correct, how much simpler development will become for AIs?
I really doubt if AI will have any impact on developer jobs. If I want to build an alternative to TurboTax, how much can AI help? I like to think of software as automating a large number of bizarre corner cases which are a result of us being human. And corner cases are what AI usually fails at.
> How much can AI help?
In 2019, this is sci-fi
In 2023, it can help engineers understand the complexity of taxes, translate some tax rules into code, help refine edge cases and fix mistakes.
In 2030, ???
As an SRE I'm not too worried on a 25 year horizon. Yes I pulled that number out of my ass.
Depending on the quarter only 10% to 25% of my time is writing code. The other is dealing with: finding high level reliability concerns, figuring out what solutions to implement, diving into complex acute production issues, evaluating trade-offs for fixes, finding ways to make things more efficient, negotiating what to prioritize with other teams/orgs/leadership, etc.
Could an AI do some of these things? Sure. Are they going to be able to figure out the right stakeholder to talk to, negotiate with them, get buy-in and decide on overall designs/solutions? No.
I'd be a bit concerned if I was a frontend developer (though any decent one could pivot to something else.) however.
> I really doubt if AI will have any impact on developer jobs.
It already has. Every single one of my coworkers is using Copilot. If you mean in terms of the job market, the shift to more declarative supervision of the tool rather than imperative implementation of the work yourself means that the bar for skill is far lower.
There will come a time when product managers and UX designers just use AI to generate Figma mocks, and AI generates an entire scaleable platform from just that. Will they need developers? Sure, but maybe 1. Certainly not 8-10.
> If I want to build an alternative to TurboTax, how much can AI help?
The alternative would be built to help users leverage AI to do their taxes. Upload your docs and it’ll get you 90% of the way there or something like that. AI will help developers build that alternative faster. Almost every piece of software will need to be reimagined from the ground up in this way. That is the job security
I think the article ends up makes the point in favour of AI anxiety:
> In particular, I started to realize that every little thing that we decide not to do as a company could have been done,
If every single company has more development bandwidth available, then at some multiple it would start driving demand and then wages down quickly. The multiple doesn't have to come just by AI improvements, as a lot of non developers can start using the AI to implement things.
It is very good if you intend to found a company, you have multiplied the things your company can do by leaps and bounds. Not so good for salaries developers.
In other words, it's a good time to found a company. It always was a good time, but this is a prime opportunity.
While I appreciate that smart people will use this as simply the next phase in bicycles for the mind, I can't help but see the current AI revolution as the first tremors of the singularity.
At the rate this is developing, I can't help but think AIs will be designing smarter AIs within a decade or two. People reassure themselves that these AIs are just synthesizing the internet datasets -- that they don't have "souls". I would say that doesn't matter -- whether or not a submarine can swim, a race of advanced submarines could dominate the ocean.
I'm not fully convinced that the singularity requires intentionally designing a general AI, or that a glorified chatbot with some Darwinian forces applied to its existence couldn't get there.
People keep thinking The Matrix or I Have No Mouth But I Must Scream, but I just keep thinking Her -- that even a benevolent AI revolution means humans are pointless. Or worse, Bladerunner, since I'm quite sure that if we do create a general AI we'll still murder it after it answers 15 questions or steps out of line.
That anxiety is much harder to shake than the one about my job - I thought we had more time.
Honestly threat from AI is not novel in the sense that anyone is under threat from being replaced by cheaper labor or eliminated altogether. AI is simply the newest one. Before this, it was jobs moving to India, RAD tools for non-devs etc. At the end of the day, whoever is unable to stay relevant is going to get steamrolled by the machine (no pun intended) of progress
How I learned to stop worrying and love the bomb.
Guys, AI is not doomsday. If anything, it will be a boon to society when it reaches mass adoption. I'm thinking of how much faster we can develop new technologies, how much quicker we can research things. MEDICAL research that can be progressed that much quicker, lives that can be saved, animals rescued from extinction, etc. Hell, even the Computer from Star Trek. All of those things have potential applications with AI.
It's just a new tool. Just like when computers arrived, mathematicians were not replaced just because computers could do the calculations faster. When autotune arrived, artists were not replaced - they just added that to their repertoire.
Yes, some jobs will be lost. Yes, there will be a changing of the guards. That's just the nature of things - it's just how it's always been.
This.
> Then at some point he describes his first time writing a game in software. He goes on about how running all the variations he tried in software would take a skilled engineer months of work, and he had done it in a couple hours - and was ecstatic.
I am sure everyone felt ecstatic at some point, regardless of their trade or field, when they worked on something all day or longer and got it done. The sheer joy it brings afterward makes your mind completely calm; past and future cease to exist for that moment. There are always going to be people like that.
AI or no AI, 20 years from now, people will still be writing software because it just makes them happy. Sure, the nature of problems will change, but these are the people who will keep pushing the field forward.
“I never think of the future - it comes soon enough.” - Grandpa Einstein.
This is great. The world will change dramatically, and it is OK to be afraid of that and the unknowns, but the best approach is to take advantage of all the good things that are coming and do not spend too much time being sad about the things that are going away.
That depends on what the good things are that are coming and what the things are that are going away.
And it also depends on the bad things that are coming, and what new things are going to come.
You are making an ethical argument based on utility, those two are not the best in combinations like that, the ethical argument should be made on an ethical foundation, not on the temporary outcome.
It's interesting how one of the most obvious problem spaces, the one of ethical reasoning is one that is least amendable to pure logic without adding a very large dose of humanity.
Example: killing is bad. But killing to save someone from endless misery (called Euthanasia) can be a good thing. But it still requires - where I live - two doctors to come to the conclusion that someone is beyond medical help and suffering needlessly. And then there needs to be a clear declaration on the capability of the person making the request that they are able to make such a request in the first place.
Ethics isn't simple, even in 'simple' cases.
Can any ChatGPT fans point to some actual case studies where the time to market for producing a non-trivial piece of software was dramatically reduced?
I mean, I built this app demo with no iOS development experience in a few hours, rather than the week it would have taken me otherwise.
ChatGPT was released four months ago. You need time to both (1) produce a non-trivial piece of software and (2) write a case study about it.
Give it a year.
At one point around the turn of the 20th century, two guys riding on horses side by side came upon an automobile going about the same speed. The one guy said to the other, “I don’t get what people see in these automobiles. We are going just as fast on our horses. These autos aren’t going to last very long.”
And they were on a relatively flat part of the tech curve…
I've finally gotten around to reading Homo Deus by Yuval Noah Harari. It's been very interesting reading this alongside everything that's happening with AI at the moment.
Would be curious to hear from others that have read it, but I find it difficult to fault his core arguments (or at least what I interpret them to be).
The problem white collar humans have right now is that they're highly specialized. They're incredibly good at being very effective cogs. This is exactly what AI is getting so good at doing (in certain verticals). Traditional capitalism effectively demands that if a company can pay the owner of an algorithm 10% of what it would pay for a human to do the same thing (for even 80% of the quality), then that's what will eventually happen.
Can government regulate it? They can sure try, but then either the companies or AI hosting providers will move to a country that doesn't have the same restrictions and it will happen anyway.
Then people will say "it will just open up other industries". I'm not sure it will. What other industries will the swaths of copywriters, lawyers, accountants retrain for?
I just don't understand everyone saying "It's going to make everyone's lives easier". In the short term sure, but if AI gets to where it's owners want it to get to, then a lot of people are going to find themselves professionally worthless.
It's entirely possible that this is just not something we're prepared for, and it's almost guaranteed at this point that there's no stopping it.
What's really interesting is this book was released in 2016... and Yuval was using Microsoft's Cortana as the example of this upcoming AI...
Interesting times either way.
I see a crack, in the economic pyramid that has sustained socio-economic inequality for so many centuries.
What you see is the pyramid becoming even more obtuse.
I can't wait for the LLMs to 'invent' their own more efficient programming language, and a dumbed down IDE for us meatbags. Maybe they can even figure out a better way to silicon then we do now.
I'm hopeful.