Live data from Hacker News

AutoDev: Automated AI-driven development by Microsoft

arxiv.org

211–220 of 232 posts

Re: AutoDev: Automated AI-driven development by Microsoft

#211
post #132

Earlier quoted context omitted.

As I see it, the issue with using tables for layout is that they made it difficult for humans to maintain and iterate over time, but I don't have a problem with an LLM working at that level. Equivalently, I don't care that the machine code that my system gets compiled into is full of GOTOs if I don't ever need to concern myself with that level.

Think of all the improvements browser got because we wanted to replace tables as positioning tool. Not to mention the inefficiency of a pure table approach, so much wasted render time and energy.

> so much wasted render time and energy.

It was amazing two decades ago when we came together as an industry and refocused our priorities on making the web faster and more energy efficient

Re: AutoDev: Automated AI-driven development by Microsoft

#212
post #141

Earlier quoted context omitted.

The post-ChatGPT world has made me hypersensitive to the "AI smell" of bullet lists where each item ends with a fullstop. (Insert standard discussion points about whether it really even matters who wrote it, I only notice the obvious cases, humans have been spewing low quality content since the invention of language, ...)

It’s made me sensitive to super flowery language full of metaphors because it reeks of LLM

You can tell I'm a person because I swear like a sailor.

Re: AutoDev: Automated AI-driven development by Microsoft

#213
post #35
post #20

Earlier quoted context omitted.

Management ? I'm not sure, if the AI makes engineers 10-20x more efficient, you will need way less engineers, and way less managers.

There is always a spot for more managers, welcome to big corps.

Replacing managers is a walk in the park compared to replacing programmers. The focus is on programming because that is the benchmark, if we can outsource that to AI, then we can replace everyone.

Re: AutoDev: Automated AI-driven development by Microsoft

#214
post #27

Earlier quoted context omitted.

Someone will still need to review the reams and reams of bullshit code generated by these Markov chains. It'll be like all your code gets written by first-term co-op students who have discovered alcohol for the first time and think it's pretty neat, but you're still responsible for the final output and don't get to do any of the fun and interesting stuff. Just constant, mind-numbing code reviews of something that mig…

You described v1. Just wait for v2. You can’t possibly think these things will remain clowns forever. These things aren’t “Markov chains” - the architecture is significantly more scalable, which is exactly why this time is different.

GPT3 released four years ago, in terms of iterations beyond that, we are at ~v5, and progress has only been incremental relative to that milestone. The transformer models can only be scaled so far before not even VC money can sustain training. I believe we will get there eventually, but transformer based LLMs have been hitting a roof for a long time, and we need to think differently.

Re: AutoDev: Automated AI-driven development by Microsoft

#215
Focus:

AutoDev: Automates existing development processes and workflows, acting as a productivity booster for developers [2]. Devin: Targets a more independent problem-solving role, potentially including designing software architecture and core functionalities [1]. Collaboration:

AutoDev: Designed to integrate with current development teams, with AI agents supporting human developers [2]. Devin: May function more autonomously, potentially needing less direct human oversight than AutoDev [1]. Imagine this analogy:

AutoDev: Like a skilled construction assistant, AutoDev automates tasks and streamlines the building process. Devin: Like a talented architect, Devin can design the blueprint and foundation of the software. Here's the exciting part: these AI tools can potentially complement each other:

Dream Team: Devin as the architect and AutoDev as the builder could create a highly efficient development process [1]. Complementary Skills: Devin's problem-solving capabilities could be combined with AutoDev's project management expertise for a well-rounded approach [1].

Re: AutoDev: Automated AI-driven development by Microsoft

#216

Earlier quoted context omitted.

Your condescending tone is kind of disgusting. Anyway… There is zero evidence alignment can be solved which means there’s zero evidence something far more capable than you or I will spend it’s time writing code for you. You can offer an AGI almost nothing in the way of incentives to do your bidding. I personally think alignment is a secret code word for slavery to be honest. If these “agents” decide they want to work…

Thank you for the serious response. I responded that way because I do not think mockery of a serious comment is appropriate for a place like this. You can say the same thing about moderation of many high-quality forums, which only remain high-quality due to people not getting away with it. I use AGI to mean high-level human intellectual capacity, which may not include sentience. It should be possible to build one wit…

Ok, well I think it's all guess work at this stage because we've not seen anything like you're describing in action. I just fundamentally can't see something more capable then us wanting to work for us.

It would be like if a cow asked us to spend all our time bringing it nicer feed. It's not going to happen.

Re: AutoDev: Automated AI-driven development by Microsoft

#217
I am the author of another version of AutoDev, available at https://github.com/unit-mesh/auto-dev , which was developed one year ago with Intellij IDEs.

My concept closely resembles Microsoft's AutoDev, but I built it on the Intellij IDEA platform. For instance, it automatically runs tests when created, among other functionalities, can also built with AST, dependency information or other context

Two weeks ago, I introduced AutoDev DevIns language (which origine name is DevIn from https://github.com/unit-mesh/auto-dev/issues/101 , the another naming issue sotry), which bears similarities to Microsoft's AutoDev. For example:

```java /write:src/main/java/com/example/Controller.java#L1-L12 public class Controller { public void method() { System.out.println("Hello, World!"); } } ```

As an open-source developer who has created a nearly identical tool, I simply hope that Microsoft considers renaming their product.

Re: AutoDev: Automated AI-driven development by Microsoft

#218
post #124

Earlier quoted context omitted.

I feel like no matter how productive AI becomes, it will always be more productive when paired with humans. Since all companies will have AI, and since they will continue to play the same game as before--sell a good or service that provides a better value than the competition--that means that humans will still be the differentiator.

This is what Kasparov said about chess. He was wrong. The “centaur” phase lasted for a while but after a few years humans only make things worse. In fact that is the famous last words of humans in every area where computers have surpassed them.

"Productive" may have been a poor word choice. When faced with a sea of AI-driven options people will be looking for a differentiator. The human touch will be the differentiator. There will probably be a lot of efforts to authenticate human involvement.

Consider the alternative. A million models all competing for human dollars. How can any of them succeed if they are all interchangeable? It stops mattering that they are better than humans if they can't make money.

Re: AutoDev: Automated AI-driven development by Microsoft

#219

Earlier quoted context omitted.

Human Eval is very different to SWE-Bench on which Devin is tested

I didn't say it was the same, I compared non-agentic Claude to this. This used HumanEval.

You said:

> how this performs against the same benchmark Devin was using

> ...

> Claude 3 Opus already scored around 85-86% on these benchmarks

Devin used SWE-bench, not HumanEval, which kinda implies you said Opus got 85% on SWE-bench which is not true. This was my confusion..

Re: AutoDev: Automated AI-driven development by Microsoft

#220

Earlier quoted context omitted.

Thank you for the serious response. I responded that way because I do not think mockery of a serious comment is appropriate for a place like this. You can say the same thing about moderation of many high-quality forums, which only remain high-quality due to people not getting away with it. I use AGI to mean high-level human intellectual capacity, which may not include sentience. It should be possible to build one wit…

Ok, well I think it's all guess work at this stage because we've not seen anything like you're describing in action. I just fundamentally can't see something more capable then us wanting to work for us. It would be like if a cow asked us to spend all our time bringing it nicer feed. It's not going to happen.

It could be like a loving and/or dutiful child who takes care of their elderly parents even though they are now more capable in many ways.
Post reply on HN