Hacker News Reader

Comment by arionhardison | Hacker News Reader

Honest question, did you foresee the last 2 or 3 years happening as fast as they have?

Edit: In terms of AI/LLM's etc..

Sure, we went from 'writing software is a grueling effort for many people' to 'writing software is a solo effort with a few chatbot subscriptions' but I think that my colleagues in the IT world tend to overestimate the effects LLMs have had on the wider populace. Folks mostly use them as search engines and there hasn't been a qualitative change in their lives.

People keep talking about how we're going to enter some kind of positive feeback loop where the chatbots design better chatbot chassis but so far we've just started generating articles about that prediction at a higher rate. This is typical of the gains from LLMs: not a qualitative improvement, but a quantitative improvement. I'd sooner believe we're about to build a Paperclip Maximizer than a Singularity Machine.

There is a vast divide between knowledge workers and creatives when it comes to LLMs

The general populace sees mostly the creative’s view. The results are poor and it comes at the expense of ripping off IP and providing bad copies of it.

The pervasive argument right now doesn’t view LLMs as a tool that have more or less eliminated entire segments of knowledge workers, who had already been offshored and are now basically redundant. Every mid level programmer who copied and pasted stack overflow snippets all the way down to support techs are done.

> Sure, we went from 'writing software is a grueling effort for many people' to 'writing software is a solo effort with a few chatbot subscriptions'

But we haven't. The only kind of software you produce in the latter scenario is unmaintainable and full of flaws. It's not at all a replacement for a team of experts.

Statistically, all software is unmaintainable and full of flaws. If it weren't, there wouldn't be so many rich software engineers.

You’re largely talking to people paid somewhere along the lines by all the AI equity that’s being raised.

AI has made making a demo easy and sometimes it’s might present more than a quick trip to wikipedia would surface. Edit: also it makes making mistakes in Excel easy(analysis).

Best give up all hopes of a future now.

I think most people in tech, in particular SV live in such a huge bubble they are completely out of touch with the common person.

Then they stand there and say 'how come they can't see what I see? Hmm'.

Oh the answer is - inertia! Lets see how long they toe that line for.

Indeed. The moment these LLM brains are given access to natural selection and evolution in the hardware space, it's game over.

We don't need to engineer plumbing robots - they will engineer themselves. We will see a silicon Cambrian Explosion. Maybe next year maybe in 500, but it's inevitable with the current economic incentives.

What if the creation of such robots destroys the economic incentive for doing so?

>The moment these LLM brains are given access to natural selection and evolution in the hardware space

Ok, but that's still sci-fi. We've made a huge advancement in automation of knowledge work, but hardware still has no clear path to what you're describing, or even to anything similar to it.

Bold of you to assume that “current economic inventives” will remain around for 500 years, especially given the American-style neo-liberal economy is only like 70 years old.

Honestly everything about the future has been about fifteen years behind schedule (or more!) from what I was expecting in 1999. I was fully expecting affordable lunar tourism for the middle class by now.

I know, right? Where's the space ship in my garage?

I first heard about Elon Musk in my teens when I was reading astronomy magazines and wishing Robert Zubrin was running NASA. Why wasn't Nasa progressing? We stopped going to the moon and made a space truck to ferry satellites and ISS nodules into LEO for $2B a pop and 1/100 odds of fatality. Anyway here comes this SV guy wishing he could put a greenhouse on mars and see if a plant would grow in Martian regolith...

I'm still waiting for that experiment, Mr. Musk.

Don't hold your breath.

Musk was never sincerely planning any kind of sci-fi future, the fact that SpaceX was built with help from Mike Griffin (who was a key player in the 80s 'Star Wars' SDI program) and the fact that Starlink's roadmap is basically reheated leftovers of Brilliant Pebbles[1] really gives the game away. He's just a generic military contractor with (well, used to have) good PR.

[1] https://en.wikipedia.org/wiki/Brilliant_Pebbles

I'm still waiting for that experiment, Mr. Musk.

Me, too. He's just another Lucy, holding just another football.

I disagree that things have improved with any particular rapidity. The LLMs of today are not really much more capable than GPT-3.5 was.

Excellent point, I should have expanded to the previous; 5 years or even better post and pre covid.

You do realise theres easy push back for this?

Go back to the movies of 30-40 years ago... that's what people envisioned about the future.

Yes, but I believe (perhaps incorrectly) that the is easy push back for most things. My question stands, is this rate of change something that the commenter could foresee. If not; it sands to reason that we should not assume the next leaps would be something that we will naturally foresee.

Reply on news.ycombinator.com