Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

Google warns its own employees: Do not use code generated by Bard

theregister.com

Post reply on HN
Print view View original

51–60 of 166 posts

Re: Google warns its own employees: Do not use code generated by Bard

#51

Post by ben_w »

soultrees wrote: post #47
JoshTriplett wrote: post #29

Earlier quoted context omitted.

> The biggest concern is that AI can sometimes generate code from the training set verbatim Not just verbatim: AI-generated code is arguably a derivative work of code in the training set even if it doesn't generate verbatim copies of training data, in the same way that any code can be a derivative work even if it doesn't contain bitwise-identical lines. If you take a piece of Python code, and translate it to C code,…

Everything you do or know is a derivative of your own training set. True original thoughts without context don’t exist, well at least in any way we can find out as every human is the product of their own training set. Just out curiosity, what makes the code you write, more original than an LLMs who’s training set is way bigger than yours and likely will more variance in how to achieve same goal. I’m not trying to be…

While I would say GPT is creative, the law doesn't have to agree with me.

Also, there's the possibility that all things creative are the human form of peacocks' tails, and if that's the case then the cost and difficulty matters more than the outcome, and any discussion based on capitalist incentives is fundamentally flawed.

https://kitsunesoftware.wordpress.com/2022/10/09/an-end-to-c...

(I wrote this 52 days before ChatGPT came out, the reference to GPT-3 is based on the stuff OpenAI released before the chat interface).

Re: Google warns its own employees: Do not use code generated by Bard

#52

Post by mejutoco »

Me1000 wrote: post #32
oxygen_crisis wrote: post #26

Earlier quoted context omitted.

If you can't write the code yourself, you're not a good judge of how competent it is. If you admire any code that simply works at all when you don't know how it should be done, then your admiration isn't a measurement worth considering. Ask an LLM how to do something you already know very well how to do properly, only then can you see its flaws through its mindless bravado.

This doesn’t ring true for me. I have ChatGPT write code for me that I could write myself. I’ve had ChatGPT even rewrite code for me to make it more legible. It’s pretty good at it, especially when it comes to more popular languages.

You are agreeing with OP. You could evaluate the quality of ChatGPT because you could write the code yourself.

Re: Google warns its own employees: Do not use code generated by Bard

#54

Post by LiamPowell »

interroboink wrote: post #34
oxygen_crisis wrote: post #26

Earlier quoted context omitted.

If you can't write the code yourself, you're not a good judge of how competent it is. If you admire any code that simply works at all when you don't know how it should be done, then your admiration isn't a measurement worth considering. Ask an LLM how to do something you already know very well how to do properly, only then can you see its flaws through its mindless bravado.

On the other hand, somewhat in the style of NP-Completeness, it is often easier to verify that something is correct than it is to generate it from scratch. Even if I can't cook, I can say it's a tasty meal (:

It's usually somewhat easy to verify a piece of code is not obviously wrong, what's much harder is proving that a piece of code is not subtlety wrong. When given a complete piece of code that appears to work, it can be very easy to convince yourself that you understand it well enough to know that it is correct, even when it's not. This problem isn't unique to LLMs, refer to the case of programmers copying binary search from textbooks without understanding how it works in their programming language of choice [1]. The problem is avoided (or at least minimised) by formal verification, which is where I think we should be heading with LLM code generation; this additionally avoids the problems with trying to accurately provide a specification in plain English.

[1] https://ai.googleblog.com/2006/06/extra-extra-read-all-about...

Re: Google warns its own employees: Do not use code generated by Bard

#56

Post by jstummbillig »

oxygen_crisis wrote: post #26
londons_explore wrote: post #24

Earlier quoted context omitted.

Agreed. I frequently use GPT-4's code. It is especially useful coding in a domain I'm not familiar with. I recently asked it to write a blender plugin to export certain polygons to a spreadsheet (with coordinates projected and transformed in certain ways), and it did admirably. I had never used blender before, and it would have taken me hours otherwise.

If you can't write the code yourself, you're not a good judge of how competent it is. If you admire any code that simply works at all when you don't know how it should be done, then your admiration isn't a measurement worth considering. Ask an LLM how to do something you already know very well how to do properly, only then can you see its flaws through its mindless bravado.

I don't have to have built a car to be able to judge the quality of a car to a decent approximation. And with code, surely, skill carry over is very real. Being able to generalize is literally our entire thing.

Re: Google warns its own employees: Do not use code generated by Bard

#57

Post by logicchains »

ec109685 wrote: post #50
vsskanth wrote: post #3

Yeah Google Bard seems to make up stuff way more than chatgpt and gives just wrong code. I asked for numpy code to fit a polyline rejecting outliers and it just made up an argument "robust" for polyfit. Chatgpt on the other hand generated a function that calculated standard deviation and removed points outside two sigma.

Here’s an example where Bard is making up studies to support RFK’s thesis that drinking water causes transgenderism: https://twitter.com/EbenThurston/status/1670619027608092674?... It’s flat out lying, yet folks are using its output in arguments.

Here's a real study: https://pubmed.ncbi.nlm.nih.gov/29314190/ . It found when frogs were exposed to atrazine (a herbicide) in wastewater runoff, "female minnows were defeminized, whereas male frogs were feminized". Similarly https://www.pnas.org/doi/10.1073/pnas.0909519107 found that "Atrazine-exposed males were both demasculinized (chemically castrated) and completely feminized as adults. Ten percent of the exposed genetic males developed into functional females that copulated with unexposed males and produced viable eggs." So Atrazine can cause frogs to change their gender, and contaminates drinking waters in part of the US (https://www.ncbi.nlm.nih.gov/pmc/articles/PMC6164008/), so it's not beyond the realm of possibility that it also influences gender expression in humans to some degree.

Re: Google warns its own employees: Do not use code generated by Bard

#60

Post by freediver »

redbell wrote: post #35

> Cautioning its own workers not to directly use code generated by Bard undermines Google's claims its chatbot can help developers become more productive The definitive definition of irony o_O This reminds me of two pictures that were circulating on social media a while ago. The first was Zuckerberg holding a book while billions of people are wasting their time in endless scrolling in the Metas; the second was that T…

Eating your own dog food is an expression commonly used to characterize this.
Post reply on HN