author appears to think that at some point in the future we'll all agree to go back to how we were writing code in 2019. unclear why
All those previous silver bullets are still with us. This one's not going away either.
This is a lot more like the PC or the Internet than cryptocurrency.
Huh.
When Taalas burned Llama 3.1 to the HC1 they saw at least an order of magnitude more tokens per joule. That saving is sitting right there, today, ready for anyone who is willing to say "this is the model we're freezing on for a couple of years".
It also seems plausible that it will take many years until hardware price/performance reaches a point where companies can provide models people deem acceptable at unsubsidized prices people deem acceptable. I don't know.
Take a look at this: https://en.wikipedia.org/wiki/Koomey%27s_law
2.6 years is about 16x in a decade. That's another path out - inference on GPUs gets so cheap they become profitable at current prices.
But in ten years we'll be using LLMs even more than today. I'm sure of it.
This is a lot more like the dot-com boom than like crypto or IoT. The internet really was a trillion-dollar industry, and AI will be too.
Maybe.
If anything, prices would go _down_ as the rehosters wouldn't be saddled with debt from training models.
In fact I'm more or less at the point where I don't even want the model to get better, I care more that it outputs faster.
If optimizing inference was a priority, OpenAI and Anthropic could also focus on it and get the prices down. They are just focusing on their dumb AGI race instead.
Maybe it can’t solve a Millennium Prize, but it’s pretty good at writing CSS.
I really cannot wrap my mind around the idea that LLMs will just not be a part of everyday life in the future — in tech and everywhere else...
1. We will (hopefully) eventually get back to the truth that programming (for problem solving) is about understanding and modeling the problem space, not writing text that satisfies a compiler / runs an app / appears to work like well-written software most of the time. And programming (for art and creativity) is about the human spark of creating something outside the norm, informed by life experience, relating to another human because of the shared experience of being human. I believe that this will shake out in the end as these tools at best helping us to think about how to write software, not actually writing the software themselves. So in that sense, we will get back to "how we were writing code in 2019" — using tools to help our own brains create.
2. The now-long history of the internet has proven to us that those skeptical of trust in what is on the other end has largely proven out. Not commercially — we can almost entirely trust what is on the other end of a business transaction, at least when trading with large players in the economy. Some costs there include compromised online identity and compromised financials (from repeated, rampant data breaches). But we can't trust the voice of individuals anymore, as proven by our rapidly disintegrating democratic norms. We are in a crisis point as a civilization when it comes to not actually talking to each other at a human level anymore, rather being mediated by the internet. So that's where that — the internet never going away — landed. I think the jury is still out on whether the internet is good for humanity. I'm not saying it will ever go away — it can't possibly — but the fact that we've accepted it as inevitable has led to us failing at considering possibilities for reigning some things in or keeping some things offline. As a present example, we now purport to trust closed models run by trillion dollar companies, when one poll showed one in five voters asking chatbots who to vote for in a few weeks in the US elections — no chance for corruption there! "How we were writing code in 2019" was being aware of where the code and algorithms we were using came from and (even in the worst case, of secreting code into a large codebase via theft) trusting it as doing as its author, a human, intended.
3. Like other cycles, unsustainable economics are propped up right now by institutional investors riding numbers-go-up until the train crashes, always looking and finding someone else to point out and say "but they're still onboard" — just like past cycles. The common person / worker / individual will be left holding the bag in very painful market convulsions that will forever scar those who actually make it. There's a cost there — ruined lives, careers, potential — a cost that we repeatedly decide is acceptable. "How we were writing code in 2019" was weighing the tradeoffs of using a product or voting with our money on its sustainability, rather than determining the future laid out by us (not at all by random or by merit) is inevitable and "the way" that we should pursue. Groupthink at its worst.
I am using "How we were writing code in 2019" here not as a point that we should return to, but as the point before the hype took hold and all reason was thrown out the window and a point to consider as we weigh — really weigh — the pros and cons.
Lastly, on a personal note, I've got no fear of being replaced. Like another in this subthread, I happened into a career that paid well for doing something I already loved. At times in my career, I even chose the path that made the least sense financially, in order to pursue something I loved and that satisfied my curiosities. I am extraordinarily fortunate to truly love the art and craft of programming and to work with beautiful, caring people who hire me to do so. If it goes away tomorrow, welp... it's been a good run.
What's _not_ going to happen is what's desired by this article which can basically be condensed into 3 bullet points to be appended to your initial spec to Opus 5.5.
I don't want to see this person's dreams crushed, but _log linear charts keep going up and to the right_. I also don't think we're entitled to the high status and pay jobs we currently have 'by right'. We've been useful - we've been _very_ useful and we've done very well for ourselves. We've been the shovels in the gold rush. Now we have to hope we can adapt and that we still provide value commensurate to our pay.
I didn't become a software engineer for the money - I just happened to one day have my hobby become a lucrative career, so perhaps my viewpoint is a little different. All the stuff that is important to the author is still stuff they can do, they just probably won't receive the status and remuneration they think they're entitled to.
But instead, humans preferred the productivity AND the readability AND maintainability that elevated us above assembly. Maybe AI-generated code will get there also.
For example, we rigidly require specs that can be used to drive end to end tests. Commits can’t break builds or tests. 100% test coverage required for both unit and integration/E2E tests.
And all of this alongside adding macOS/iOS/Android to our standard builds instead of just web/electron. Oh. And moving backends to Rust/Go.
Anyways, you've obviously got good policy in place for generating code. This wisdom isn't universal.
I can't honestly say so, though.
Yet, I can say I am so excited for the future! For once, I don't want to wake up in the future — I'm loving being at the heart of it, seeing it progress and watching it allow people to be more creative and more empowered.
This has been my state of mind continuously for more than 3 years
Some people I know also seem to believe this? Like all the AI providers are going to go bankrupt and we'll just go back to what there was before? Do they not realize that indeed the productivity gains are there?
Sure, there's a lot to be debated about how this fell on many people's laps, on how they are being asked to do more, with the same pay and sometimes putting in more hours. These are pressing issues that have emerged as a natural (or, well, maybe forced for some) consequence of a complete shakeup in the world in general. But I don't think society (i.e. most people) will go back to what was before unless we actually have some form of devastation that really physically prevents us from doing so.
If OpenAI and Anthropic and what have you go bankrupt, others will continue. Maybe the economics will look a bit different, but to think we will just go back to what it was seems, to me, completely nonsensical. But maybe I'm the one spouting nonsense. We'll see in a few years...
The main issue for front end development is that companies (particularly large ones) have been releasing shitty broken UI/UX for so long that this stuff became the norm.
Either way the article does not even explain what "blows over" mean. A lot of people seem to be fearfully reacting to the title alone.
As far as what AI actually provides, at the end of the day its cheaper software faster (or that's the promise). You still need to build products and maybe you can do that with less folks now. You're not writing assembly today. The machine does it. But you're still working on software. Maybe with AI you're programming less and working on UX more. Is that earth shattering change?
We've yet to hit a point where you any of the big software companies have been overthrown by an AI startup... Whats the revolution really? Its captured the mindshare of the industry but we're at a point of being practically disgusted by AI products already.
Things won't literally go back to 2019. But there could certainly be a return to normalcy in store.
Well, it cuts both ways, the other earth shattering change is Designers taking on more programming/engineering work.