doginasuit

I generally agree but it starts to get interesting again once you have the first solid bricks in place. The LLM is incredibly good at seeing where you are going. You can place one brick and the next few become obvious.

I've been working on a project for the last 18 months, and over that time LLMs have advanced to the point where they've become useful. I can write a section of UI and ask the LLM to write all the front end and backend services. There are enough examples that it writes exactly the code I would have written myself. It can build entire sections of code quickly and efficiently because there aren't very many questions without an answer somewhere in the codebase, which it is very good at finding. If it finds any gaps, it can raise it as a question. It analyzes, asks questions, and then completes the work.

The impression that I'm starting to form is that you do not want to use AI very much early in a project, except to ask questions and check your code. But it's usefulness at assisting with code scales with the size of the project.

This seems inevitable in one form or another. A lot of talented devs are struggling to find employment at the same moment that there is a tool that can guide you through the parts of launching a product that are outside of your expertise. That's just more time to tinker out of curiosity or necessity.

I believe the next few years are going to see a lot of new and innovative tech products. Maybe this is how we build the internet we actually want.

My hunch is that AI can certainly speed up development but you have to walk a narrow path. It generates garbage much easier than valuable product.

I can see your point, but on the other hand it seems like the only alternatives are a blanket ban on reverse engineering or a blanket allowance. I probably don't have to make the case for the value of reverse engineering here, so the question is to what degree it can be abused and if they have the responsibility to stop it. I would argue companies do have an obligation to consider and mitigate abuse of their product, especially frontier technologies. Probably the best outcome will be some kind of balance of judgment by the company.

You may want to change the logo, it is a different color but close enough to cause confusion. Not sure what a trademark court would say but it would definitely be enough to draw intimidation from Google if they wanted to.

I enjoyed this!

This is an interesting question to ask of any company. Every company over a long period is a Ship of Theseus. Google started out as a rowboat and now it is an ocean liner.

Investment in product quality, "Don't be evil", everything that set a company apart and allowed it to rise, it lasts only as long as it is preserved in a way that resists decay.

Is there anything left of the rowboat? I'm at the point where I am actively trying to shut Google out of my life to the degree that it is possible. And yet it is still not Microsoft. I catch a glimmer of the old google in a product release now and then.

While the Jan. 6th convicts all get pardons.

> The only way it will result in less work and the same or better quality of life is if you’re self-employed or are employed by a compassionate human.

May we take this as an opportunity to recognize what a bad deal we get from most employers and explore better ways to organize labor. It's never been more possible. UBI is unlikely to happen and uncertain it would even be beneficial. If AI can lead to a better world, this seems like the most promising path.

This is a very relatable sentiment. On the other hand, checking code is one of the best use-cases for an LLM, it is better at checking code than generating it.

It seems like the question is where to fit it in to the pipeline. The team responsible for reviewing submissions could leverage it to speed up the process, and still respond personally.

> Computer manufacturers devised a simple solution: add a TURBO button that would just slow your CPU down on demand.

I remember these! I could not figure out why it never seemed to make the computer faster and actually made it slower. TURBO was even written in italics to imply speed.

I suppose for some short time it may have actually overclocked the CPU, until the default clock speed caught up. After that, the turbo button was a lie.

Edit: Maybe I have it backwards, it sounds like on some machines it might have been in the underclocked condition when it was off. I just remember disappointing results with the machine that I used.

> generate 100 solutions cheaply without coordination between them and pick the winner with a relatively fast eval

While it might get workable results, this seems like a miserable way to work with code. You spend the vast majority of time thinking about the problem domain of working with agents instead of your project, and for mediocre output. You are transforming your role into a dumpster diver of code.

The approach that proves to be fastest in the long run is to stay close to the code. Lay out the base structure, and use an LLM to check your work and fill in the gaps. The beauty of this approach is that it scales very well with the size of the project. When the LLM has plenty of high quality examples to follow, they start to write the same code that you would. Where it diverges, create documentation.

Perhaps the single best property of LLMs is their ability to consume and apply examples and documentation.

Well said. Attribution is important, for obvious human reasons and to direct people to valuable context. But our names will fade fast from history, what matters is the journey of human knowledge.

> It is amazing what you can accomplish if you do not care who gets the credit.

--Harry S. Truman

Every movement has a counter movement.

I've been using it to generate animated svgs. I made a breakthrough in asking it to create bones/rigging and movement with a python library and apply the result to the svg.

The most noteworthy is the Zizians, a fringe group that is actually linked to several murders. Not exactly fair to lump the entirety of the site with their ideas, but I'd argue that it was a direct result of the echo chamber.

https://en.wikipedia.org/wiki/Zizians

Beyond that, Effective Altruism got its start partly on Lesswrong. Whether or not that counts as harmful would depend on how you see that movement, but one of its biggest proponents is now in jail.

https://en.wikipedia.org/wiki/Sam_Bankman-Fried

There's a famous thought experiment that came out of Lesswrong that has resulted in a lot of distress and questionable life choices from people who take it too seriously. A lot of the AI discussion I've seen on the site has the same overtones, including this particular post.

https://en.wikipedia.org/wiki/Roko%27s_basilisk

The conversation between ChatGPT and Dessa was an amusing flex. Not sure if that was actually the intent, but I had a hunch.

I might be able to enjoy this as scifi but the source makes it difficult.

Lesswrong is one of the smellier armpits of the internet. My impression is that there are clearly a lot of smart people who post there, but they've collectively embraced a lot of delusional thinking. There have been examples where it has escaped containment and harmed people.

I found the article interesting, humans are flawed enough in a variety of aspects that discovering a convergence here is noteworthy. I don't take it as a sign the LLMs believe anything in particular, they are not capable of belief, as you said.

Everything that comes from an LLM is a human expression, on some level, but if you asked a sampling of humans the same question I wonder if it would come out the same way. How would you explain the difference? What makes that answer more probable?

When I was in highschool in the 90s, our school spent a ton of money on a set of midi keyboards hooked up to workstations, new technology at the time, and offered a class on digital music production. Our teacher was the school music director and I remember he was really excited.

The software had a quantization feature which he showed us how to use but he warned us not to use it on everything, because it makes music sound mechanical and less human. I guess that philosophy did not survive in the music industry.

> The last piece is that social media has made it popular to all like the same thing

Social media definitely didn't invent this. I'm not overly familiar with what is popular on social media, but it seems like the internet has greatly expanded the viability of different styles and niche music. When I get together with friends, we all have tons of music than most of us haven't heard before.

The author isn't against icons in menus, they are against icons in every menu by default.

Icons can genuinely help convey meaning/understanding. They can also be clutter. It comes down to how you use them.

I see I didn't make this aspect very clear in my comment, but I believe that what I described is only possible when a human is in tight control. I agree that an LLM on its own will generate pure slop 9 times out of 10. As an extension of your own creativity, I believe it will push the boundary.

AI makes things faster, and it pushes the envelope on what can be built. Like a telescope lets the human eye see further into space, an LLM helps track more ideas and structures within working memory. Yes, people will use this new potential to build slop. And other people will use it to build what has never been built before. The only question is which one we will prefer.

Thank you for figuring it out. It is just one more probably necessary step along the long journey to me finally getting an Atlantic subscription. But like every journey, it is not about the destination but the friends we made along the way, as our mutual acquaintance Abraham Lincoln once famously said.

Yes. Was that with The Atlantic?

Not the OP, but thanks for mentioning this! Looks really useful.

I really enjoyed it and I look forward to trying out the ideas.

I think you are close to every effective presentation format, but the cramped space for the text was the biggest friction. The moving visuals, as helpful as they are, also became a little distracting as I was trying to absorb the meaning of the text. You could solve both issues with a toggle to hide or expand the visual.

Abraham Lincoln's recommendation didn't work for me, reader mode shows the abbreviated text.

Last week there were two separate posts, one said "AI is not a tool" and the other one said "AI is a tool." Even though that logically covers all possibilities, somehow the vast majority of comments in both were negative or contradicting. My takeaway is that comments aren't a very representative sampling of how people actually feel.