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

Canvas is a new way to write and code with ChatGPT

openai.com

Post reply on HN
Print view View original

641–650 of 701 posts

Re: Canvas is a new way to write and code with ChatGPT

#641

Post by no_wizard »

cj wrote: post #4

This is cool, but I wish it were integrated into tools already used for coding and writing rather than having it be a separate app. This also demonstrates the type of things Google could do with Gemini integrated into Google Docs if they step up their game a bit. Honestly I’m scratching my head on OpenAI’s desire to double down on building out their consumer B2C use cases rather than truly focussing on being the infr…

To be honest I think they’re having less success than it appears with their B2B offerings. A lot of cloud providers services like AWS have their own things they sell through those channels and I think a lot of businesses are finding those solutions to be cheaper and “good enough”

Re: Canvas is a new way to write and code with ChatGPT

#642

Post by satvikpendem »

rty32 wrote: post #634
satvikpendem wrote: post #497

Earlier quoted context omitted.

That's funny, I do the exact same thing. If I'm out and about or otherwise not at a computer, and I have an idea or solution to a bug I wanted to solve, I'll fire up an LLM on my phone (usually Claude these days) and basically brainstorm together the pros and cons, with code sometimes as well.

Curious -- does this happen to you often? Do you do that for your work (what pays your salary) or personal projects? Do you find it stressful to think about coding all the time?

Not for work, only for personal projects. No way I'd expend that type of effort for something like work, which I put in my 8 hours a day anyway. Stressful, not necessarily, unless there is some annoying bug I haven't been able to solve and deadlines are coming up, but again, that would be mainly for work, not (necessarily) for personal projects, unless I necessarily need to fix that bug to progress further.

Re: Canvas is a new way to write and code with ChatGPT

#643

Post by herval »

Barrin92 wrote: post #433
herval wrote: post #121

Earlier quoted context omitted.

google's approach to shipping products is puzzling. It's like they don't care if anyone uses them at all

Google isn't a startup, they aren't desperate to impress anyone. I don't even think they consider "AI" to be a product, which is probably correct. These AI enabled features are background processes that ideally integrate into products over time in ways that don't require you to explicitly know they're even there. Given how widely used Google Docs is, for serious work, disrupting people's workflows is not a good thing…

I think what you mean is "Google is complacent, so they don't think they need to make a lot of effort to stay relevant"

Re: Canvas is a new way to write and code with ChatGPT

#644

Post by joseda-hg »

munchler wrote: post #292
JumpCrisscross wrote: post #174

Earlier quoted context omitted.

> demonstrates the type of things Google could do with Gemini integrated into Google Docs Or Microsoft! > think OpenAI will end up being either an infrastructure provider OR a SaaS, but not both Microsoft cut off OpenAI's ability to execute on the former by making Azure their exclusive cloud partner. Being an infrastructure provider with zero metal is doable, but it leaves obvious room for a competitor to optimise.

Microsoft is integrating Copilot into many of their products, including Visual Studio and Office/365.

VSCode Sure, but my experience with Copilot + regular Visual Studio has been nothing short of abismal

Re: Canvas is a new way to write and code with ChatGPT

#645

Post by danielmarkbruce »

dartos wrote: post #569
danielmarkbruce wrote: post #389

Earlier quoted context omitted.

There is a lot of nonsense in here, for example: > but we know that synthetic datasets make for poor training data This is a silly generalization. Just google "synthetic data for training LLMs" and you'll find a bunch of papers on it. Here's a decent survey: https://arxiv.org/pdf/2404.07503 It's very likely o1 used synthetic data to train the model and/or the reward model they used for RLHF. Why do you think they don…

Sure, hand wave away my entire comment as “nonsense” and ignore how statistics works. Training a model on synthetic data (obviously) increases bias present in the initial dataset[1], making for poor training data. IIRC (this subject is a little fuzzy for me) using synthetic data for RLHF is equivalent to just using dpo, so if they did RLHF it probably wasn’t with synthetic data. They may have gone with dpo, though. […

Did you read this paper? No one is suggesting o1 was trained with 100% synthetic or 50% or anything of that nature. Generalizing that "synthetic data is bad" from "training exclusively/majority on synthetic data is bad" is dumb.

Researchers are using synthetic data to train LLMs, especially for fine tuning, and especially instruct fine tuning. You are not up to date with recent work on LLMs.

Re: Canvas is a new way to write and code with ChatGPT

#646

Post by vessenes »

machiaweliczny wrote: post #586
vessenes wrote: post #191

Earlier quoted context omitted.

How big is this codebase? gpto-preview and claude sonnet are both fairly capable. Claude has longer context windows and is a little less lazy. aider.chat is working on an architect / coder model right now that's verrrry effective. Essentially - paste your whole codebase into 4o-preview, ask it to rearchitect to spec, and output a design document and list of tasks suitable for a senior dev. Dump that plus your tarball…

Would love a more specific example of this. How to use this architect mode. BTW surprised that tarballs work - aren’t these compressed?

Just incant tar cvf and leave off the z or j: there’s a tiny binary header but Claude won’t mind.

Check out the aider GitHub repository for details but 4o is responsive to text requests too.

Re: Canvas is a new way to write and code with ChatGPT

#647

Post by UncleMeat »

famouswaffles wrote: post #511
UncleMeat wrote: post #194

Earlier quoted context omitted.

Andrej is a great communicator. I've never seen evidence that he is an especially exceptional programmer.

Between his roles at companies he's worked for, his research and his open source stuff, what else would you be looking for lol ? Remove all the educator stuff and karpathy would still be one of the most accomplished of his generation in his field. Idk just seems like a weird comment.

Research contributions are almost entirely separate from engineering quality. This is true for the vast vast vast majority of work people do in grad school.

Re: Canvas is a new way to write and code with ChatGPT

#650

Post by NeoTar »

mooktakim wrote: post #624
NeoTar wrote: post #576

Earlier quoted context omitted.

Let’s just throw away all past experience then? It’s a mistake to assume that there will be 100% correlation between the past and future, but it’s probably as bad of a mistake to assume 0% correlation. (Obviously dependant on exactly what you are looking at).

So the choices are 100% or 0%?

That’s my point – they are not. Your previous comment implied to me a belief that any attempt to draw inference from past events was doomed to failure!

Each circumstance is different. Sometimes the past is a good guide to the future – even for the notoriously unpredictable British weather apparently you can get a seventy percent success rate (by some measure) by predicting that tomorrows weather will be the same as todays. Sometimes it is not - the history of an ideal roulette wheel should offer no insights into future numbers.

The key is of course to act in accordance with the probability, risk and reward.

Post reply on HN