actsasbuffoon

Yeah, the value of software as a whole seems like it is about to greatly decrease. You’d think tech companies and investors would be a lot more nervous about destroying the value of software, but I’m not sure if they’re thinking that far ahead.

Hooray, in a few years you can type a prompt and get a complete app you can sell to customers! But the customer can also type a prompt and get a complete app that meets their expectations even better than yours. Your generated app is worthless. Your entire company is just a middle man between the customer and ChatGPT, and the customer can write a prompt just as well as you.

Software as an entire industry is pretty much going to cease to exist if AI enthusiasts get their way.

We already see this with music. Have you ever had someone give you a link to a song they generated with AI? Do you really want to spend 4 minutes listening to a thing they made in 15 seconds? If I wanted an AI song about coffee then I’d generate one myself, not listen to yours. The faster and easier it gets to generate apps, the more we’re going to end up with people just generating their own. You won’t be able to sell your generated apps to anyone.

I’m on the opposite end of the spectrum. I’m almost certain that this is going to end extremely badly for the majority of humanity, and for programmers in particular.

I think there’s a less than 5% chance that this goes well, and that’s only if we get a series of things to go extremely well. And frankly, we’re tracking along the extremely bad path so far.

It wouldn’t even necessarily need to be a real GZip bomb. Just something containing a few hundred kb of seemingly new and unique text that’s highly compressible and keeps providing “links” to additional dynamically generated gibberish that can be crawled. The idea is to serve a vast amount of poisoned training data as cheaply as possible. Heck, maybe you could even make a plugin for NGINX to recognize abusive AI bots and do this. If enough people install it then you could provide some very strong disincentives.

It’s also pretty easy to create your own VCV Rack modules, assuming you don’t mind doing a little math.

I can see it now. You get an ad showing you in prison. The text reads, “I shouldn’t have downloaded that torrent without NordVPN.”

Welcome to our new dystopian hellscape.

Unless I’ve missed some major development then I have to strenuously disagree. AI is primarily good at writing isolated scripts that are no more than a few pages long.

99% of the work I do happens in a large codebase, far bigger than anything that you can feed into an AI. Tickets come in that say something like, “Users should be able to select multiple receipts to associate with their reports so long as they have the management role.”

That ticket will involve digging through a whole bunch of files to figure out what needs to be done. The resolution will ultimately involve changes to multiple models, the database schema, a few controllers, a bunch of React components, and even a few changes in a micro service that’s not inside this repo. Then the AI is going to fail over and over again because it’s not familiar with the APIs for our internal libraries and tools, etc.

AI is useful, but I don’t feel like we’re any closer to replacing software developers now than we were a few years ago. All of the same showstoppers remain.

Excellent point, but now consider the availability of open source models as well. They’re not as good as the frontier models, but at some point they get close enough, right?

I run Llama 3.x locally on my 5+ year old GPU. It’s pretty decent. It’s not as smart as GPT-4 or Claude, but it’s good enough for a lot of what I do, and most importantly there’s no cap on the number of prompts I can send in a given timeframe.

For many consumers, it’s fine to be a little bit behind the cutting edge. My smartphone is 2 models behind now. It’s fast enough, the camera is good enough, and the battery is holding up just fine. I don’t feel any need to fork over $1,000 for a new one right now.

I think the same phenomenon will become more common for AI enthusiasts. It’s hard to beat free.

And what happens to the economics of building frontier models then? I suspect a lot of air is going to be released from the AI bubble in the next couple of years.

I’m assuming they mean that if you multiply subscribers by the subscription fee then OpenAI still ends up losing billions per year.

I think there are complicating factors.

Investors are throwing unfathomable amounts of money at AI right now, and frontier models are very expensive to build. There’s also the issue of these AI-as-a-Service companies requiring an ever increasing amount of GPUs and electricity, and because frontier models keep getting bigger, the problem keeps getting worse.

There is a serious risk to these companies. A frontier model doesn’t sit at the frontier for very long. Given a year or two, open source models can catch up for less than 1% of the money spent to build a frontier model. The moat doesn’t last long, so you need to keep burning tens of billions to stay ahead.

But as open source models continue to improve, we run the risk of reaching a moment that looks like the current smartphone market. People aren’t upgrading their phones as often anymore because a 4 year old smart phone is still pretty great. There’s no need to buy every year. Well, at some point an open source model will be close enough to a frontier model that many users will just stick with it. They’ll prefer it because there’s no monthly fee to run it on their own hardware, because there are no rate limits, and because they prefer the privacy and flexibility.

Right now you need a big cluster of GPUs to run top tier models. But with every passing year we get closer to good-enough open source models running on gaming GPUs.

And when that happens, would you want to be one of the investors shoveling tens of billions of dollars into building frontier models with declining subscription numbers where you’ve got no moat?

I don’t think the economics of this are going to pan out in the long run. Everyone is still hyped about ChatGPT, so investors are happy to throw money at it. But I think we’re in a very big AI bubble right now. The economics only make sense if someone achieves AGI within the next couple of years. If they don’t, this whole thing is going to decline sharply as open source models provide a better free alternative.

I’ve actually had a few recruiters reach out in the last few weeks. It’s nowhere near the daily calls I used to get, but it seems like there’s been a little bit of an uptick.

I like that idea, and I may need to steal it. I have this tendency of asking candidates questions, and if I feel like they’re demonstrating very strong knowledge then I may toss them a few extremely obscure questions for bonus points, but I never expect candidates to get them right.

But this up-front approach of setting expectations seems like a better way to go.

Yeah, this was certainly true about 5 years ago. The market has thinned a lot over the last few years.

Yeah, even in this market, your best people still have options. I recently had to find a new job, and while it was definitely more sparse than normal, I had a decent offer within a month.

Five years ago, I probably would have had my pick of 3-5 offers after about two weeks. This time it was a month for one offer. Definitely a bummer, but the offer was still an upgrade, and so far the new company is working out well, so it’s fine.

If you mess with your people badly enough, the only ones who will stick around are the ones who can’t get a better offer.

Other fun options include telling it that you’ll pay it a tip if it does well, asking it to play a character who is an expert in the subject you want to know about, or telling it that something terrible will happen to someone if it does a bad job.

Congress did their job. They granted authority to regulatory agencies who are much better equipped to make these sorts of decisions. Regulatory agencies employ scientists and subject matter experts in their relevant fields.

Our congresspeople are a mixture of serious people and clowns who are very good at saying outrageous things to get on television. Even the serious people cannot possibly be well-rounded enough to have the kind of expertise necessary to write perfectly detailed regulations for every possible field.

Could I come up with some regulations related to tech? Sure. But it would be a terrible idea to ask me to write financial, medical, or environmental regulations.

Like I said, Congress did their jobs. Sometimes delegating to experts is the smartest move.

If I rob a bank then you can tell. If I make an AI model on my own computer and train it, how would you ever know? It won’t have a “signature” that can be detected by any kind of AI detection system because I’m the only person who knows it exists.

You can make it illegal if you like, but you’ll never be able to enforce it. And other countries are not obligated to follow suit, and will have significant competitive advantages as a result.

There are a bunch of well-known areas where popular calculators tend to give incorrect answers: https://apcentral.collegeboard.org/courses/resources/example...

It’s mostly fine until it isn’t. AI will probably operate in the same capacity. We already have so much incorrect information out there that’s part of our pop culture. Even down to things like the fact that Darth Vader never said, “Luke, I am your father,” and Mae West never said, “Why don’t you come see me sometime?”

Even basic movie quotes are beyond our ability to get right. Hilariously, I just asked ChatGPT about these quotes and it explained that these are common misquotes, told me what was actually said in these movies, and explained some relevant context.

Sherlock never said, “Elementary, my dear Watson” even once in the books. Kirk never said, “Beam me up, Scotty.” We’re much less correct than we like to think. And somehow we’ve survived.

ChatGPT is fallible just like we are. We’ll manage, just like we always have.

The cat is already out of the bag on that one. You can download several open source text to image AIs and run them on your own computer. You can train new styles and concepts on your own computer in about 20 minutes. You can throw up roadblocks for the big players, but it’s far too spread out to ever be contained again.

Ditto for LLMs. Obviously GPT-4 is the best game in town, but I have several LLMs and the code required to use them installed on several machines. Training new data for them can close the gap quite well at the expense of becoming less general.

We can’t regulate our way out of this one. That window closed a while ago.

ChatGPT has plugins in closed beta that can fetch real-time information from other sites and services. Things are moving so quickly that even though this appeared to be a problem a week ago, it’s pretty much solved now.

And in the months Google has had to respond, they have released Bard which is very noticeably bad compared to the competition. If this is their response then Google is in deep trouble.

There are a lot of reasons.

First, we’re not talking to kids about intercourse. We’re saying kids need to have some amount of knowledge about this stuff that is appropriate for their age.

For example, small kids who are put into the care of strangers need to be aware that adults are not allowed to touch them without permission. They need to know the proper names for their private parts because it’s more difficult to understand if a kid is being abused when they’re using strange euphemisms. They need to understand that they shouldn’t be ashamed to tell an adult if someone touches them in a way that makes them uncomfortable.

Are these conversations I want to have with my young child? Hell no! But they’re necessary for a kid his age. I wouldn’t be doing my job as a parent if I didn’t teach him these things.

The reality is that it’s going to be a series of age appropriate conversations every few years, because he’s going to need to know a little more as he gets older.

And if parents are sending their kids into the world without this kind of age appropriate information then they’re putting their kids at serious risk for abuse. Frankly, this is the kind of situation where I’d be glad that schools and libraries are making age appropriate materials like this available for kids; to help the kids whose parents aren’t doing their jobs.

Sure, though I certainly see pretty wide variances in myself based on my environment. If I’m happy and in a good environment then I tend to be very agreeable. As my stress and unhappiness go up, I tend to get pretty disagreeable.

I had one job that was absolutely awful, and to be brutally honest my attitude sucked. I wasn’t as nice to my coworkers as I wish I had been. I wasn’t a complete asshole or anything, and my behavior was never raised with me as being problematic, but I’m not proud of how I acted through that time.

But over the course of a fairly long career I’ve gotten a lot of comments about how laid back and friendly I am, and how I tend to go out of my way help people. I like to think that’s who I really am, but I can definitely be unpleasant if I’m miserable.

I’ve been doing this for a long time now, and I’ve had good results from companies with humane interview practices.

Places that treat candidates poorly tend to treat employees badly, and companies that drag out the interview process tend to miss out on in-demand talent.

As a result, places that respect your time and well-being have been more likely to have good engineers who have stuck around. Places that are toxic and/or bureaucratic tend to combine bad hiring with high turnover and create a terrible environment.

I love it for writing tests. It’s good at picking up patterns, so if I need a bunch of unit tests then I can usually write one test and let it write the rest based on the RSpec descriptions.

I recently had a ticket that required refactoring some code that used net::http from Ruby’s stdlib over to using the HTTP gem. The rewrite was pretty easy to do by hand, but we had a bunch of test cases and rewriting them proved to be a bit painful because the mocks required were quite different.

So I spent half an hour rewriting the first test case with the new shape of mocks. Then I commented out the remaining tests and the AI recognized what I was doing and rewrote the remaining test cases with the new mock structure in seconds.

I treat AI like a junior programmer. I don’t trust it to do the hard stuff, but it’s great for scenarios where I have already established the pattern and I just need someone with basic coding proficiency to apply that pattern to a large chunk of code. It’s amazing at doing the grunt work.

Yeah, there have been numerous times when having a fireplace and/or natural gas stove saved my family from freezing. When you live in a place that routinely gets cold enough to kill, you need multiple heat sources that don’t share a single point of failure.

According to a UBS study from last week, ChatGPT is the fastest growing consumer-based web app in history: https://www.reuters.com/technology/chatgpt-sets-record-faste...

This thing is taking off like wildfire. Google has very good reason to be terrified. This kind of growth is quite likely indicative of a major disruption.

It’s hard for me to imagine that Edge would become my main browser (big Firefox fan for many years), but I’d certainly give these new AI features a test drive.

I apologize; I may have explained poorly.

I was trying to convey that burnout is (IMHO) a result of putting a lot of effort into something and not getting enough back out of it. It’s like a defense mechanism to stop wasting energy. This may be a result of processes, tools, compensation, treatment by co-workers, or any number of other factors.

And my suggestion on how to resolve burnout is to find a way to feel productive and capable again. Personally, game development worked for me. But I also mentioned mobile app development as a possible avenue, but obviously this list isn’t intended to be exhaustive.

Just find something that helps you remember why you loved programming in the first place. You probably haven’t stopped enjoying the act of writing code. It’s just that something at work is making you unhappy, and you’ve started to associate those negative feelings with programming in general.

One paragraph mentioned that, but the thrust of the argument was about burnout. For me, burnout occurs when I feel like I’m putting a lot of time and effort into something but I’m not getting enough back to justify it. Burnout is like a mechanism to get you to stop pouring energy into something that isn’t working.

Unfortunately, many managers are bad at giving positive feedback. The only time you hear from them is when something isn’t working. Corporate processes slow things down and make you feel ineffective. Eventually you start to wonder if you’re actually bad at software development, or if you’re just a bad worker.

Let’s say you get a ticket for a performance problem. It turns out that the app is doing something kind of silly, and it would be easy to fix the problem and you’re confident that users would be okay with it. Well, too bad. You’re not allowed to make product decisions; only management can do that. So you bring it to management, but they’re in a meeting. You don’t have enough time to do anything else, so you spend 15 minutes reading HN.

They get out of the meeting, and you explain the problem. They’re okay with the change, but because it affects the UI, you need to get approval from the UX team. You explain that it’s a very minor change, but management insists that you have to follow the process.

You go to UX. You explain the problem, but they don’t have time to get back to you. You figure it’s going to be a bit, so you grab another ticket. 20 minutes later, it turns out that they’ve approved your request. It was a very small change, after all. So now you have to put your ticket back, which is going to mess with the time tracking in Jira that your boss gets pissy about, but whatever.

You make the change, which takes about 5 minutes. You commit, push, and open a PR. The CI tests fail. Apparently there was some kind of silly linter error, so you fix that and push again. Now you wait for the tests to pass, which takes about 40 minutes.

The QA team requires a detailed set of instructions for how to test every ticket, so you start writing up those instructions in the Jira ticket. As that’s happening, you get a Slack message from the documentation team. They noticed your ticket status change, and you forgot to mark it as requiring doc review because it affected the UI. You apologize and go back to update the ticket.

When you’re done updating the ticket, you notice the build has failed. Looks like a flakey test that’s been a problem for months now, but no one ever gets time to fix it. You re-run the tests and cross your fingers.

Time for a sprint planning meeting. Two hours later you resume work. The build passed, but your irritatingly picky co-worker has requested a change on your PR. He always requests at least one change on every PR, no matter how small. You used this method to find an element in an array, and while there’s nothing wrong with that, he personally prefers that method for doing it, and he won’t approve your PR until you change it.

So you update the branch and push again. Then you start working on filling out information for the compliance team which is needed for anything affecting this part of the application. You wrap that up and notice that the flakey test failed again. Swearing under your breath, you re-run the build.

You spend some time reviewing PRs for your co-workers and see that your build passed and your picky co-worker approved the changes. You merge and move onto something new. A few minutes later you get a message from the QA team. You didn’t provide documentation for how to test the code. You remember that you were in the middle of writing it when you got interrupted, and you forgot. You apologize and finish writing it.

You look at your watch and realize that it’s the end of the day. You know your boss is going to be irritated with you after standup tomorrow because you spent all day working on a ticket that was only estimated for half a day.

This is the kind of stuff that makes me want to run away from software development and never return. It’s not the code; it turns out that after all these years I still love programming. It’s all of the soul crushing BS that makes me lose the will to live.

And so far, the antidote for that has been to work on something with zero red tape where I have full autonomy. It reminds me that I’m actually very good at software development.

That doesn’t sound like you’re falling out of love with programming. It sounds like you’re frustrated with the red tape at work.

Try working on something fun on the side. Make a game, or a mobile app, or something you’ve always been curious about that doesn’t have anything to do with your job. You might be amazed by how productive you feel when you work on something ambitious that doesn’t involve all of the corporate machinery.

I’ve been getting a nice dose of that with game development. Sometimes I sit down on a Saturday with a big plan that should last me through the weekend, and I get it done before dinner that evening, and I feel like some kind of programming wizard. It’s been a great reminder that I am a talented programmer, but I’m just feeling burned out on all of the tedious process that’s involved when coding professionally.

Yes, plus a bunch of other retailers. I keep marking them as spam, but Google isn’t getting the hint. This has been going on for months now and I’m strongly considering switching my personal email over to a new service.

I don’t understand how GMail has dropped so severely in quality. These are some of the most obvious spam messages I have ever seen.