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

ChatGPT agent: bridging research and action

openai.com

Post reply on HN
Print view View original

161–170 of 508 posts

Re: ChatGPT agent: bridging research and action

#161

Post by 8note »

exitb wrote: post #102
jstummbillig wrote: post #89

Earlier quoted context omitted.

I am looking forward to learning why this is entirely unlike working with humans, who in my experience commit very silly and unpredictable errors all the time (in addition to predictable ones), but additionally are often proud and anxious and happy to deliberately obfuscate their errors.

You can point out the errors to people, which will lead to less issues over time, as they gain experience. The models however don’t do that.

but the person doing the job changes every month or two.

theres no persistent experience being built, and each newcomer to the job screws it up in their own unique way

Re: ChatGPT agent: bridging research and action

#162

Post by marcellus23 »

torginus wrote: post #111
samtp wrote: post #82

Earlier quoted context omitted.

This is the exact same issue that I've had trying to use LLMs for anything that needs to be precise such as multi-step data pipelines. The code it produces will look correct and produce a result that seems correct. But when you do quality checks on the end data, you'll notice that things are not adding up. So then you have to dig into all this overly verbose code to identify the 3-4 subtle flaws with how it transform…

I'll get into hot water with this, but I still think LLMs do not think like humans do - as in the code is not a result of a trying to recreate a correct thought process in a programming language, but some sort of statistically most likely string that matches the input requirements, I used to have a non-technical manager like this - he'd watch out for the words I (and other engineers) said and in what context, and wou…

I don't think you'll get into hot water for that. Anthropomorphizing LLMs is an easy way to describe and think about them, but anyone serious about using LLMs for productivity is aware they don't actually think like people, and run into exactly the sort of things you're describing.

Re: ChatGPT agent: bridging research and action

#163

Post by modeless »

pants2 wrote: post #145
modeless wrote: post #139

Earlier quoted context omitted.

Agents respecting robots.txt is clearly going to end soon. Users will be installing browser extensions or full browsers that run the actions on their local computer with the user's own cookie jar, IP address, etc.

I hope agents.txt becomes standard and websites actually start to build agent-specific interfaces (or just have API docs in their agent.txt). In my mind it's different from "robots" which is meant to apply rules to broad web-scraping tools.

I hope they don't build agent-specific interfaces. I want my agent to have the same interface I do. And even more importantly, I want to have the same interface my agent does. It would be a bad future if the capabilities of human and agent interfaces drift apart and certain things are only possible to do in the agent interface.

Re: ChatGPT agent: bridging research and action

#164

Post by ghm2180 »

burningion wrote: post #159
bijant wrote: post #107

Earlier quoted context omitted.

THIS is the main problem. I was listening the whole time for them to announce a way to run it locally or at least proxy through your local devices. Alas the Deepseek R1 distillation experience they went through (a bit like when Steve Jobs was fuming at Google for getting Android to market so quickly) made them wary of showing to many intermediate results, tricks etc. Even in the very beginning Operator v1 was unable…

This is why an on device browser is coming. It'll let the AI platforms get around any other platform blocks by hijacking the consumer's browser. And it makes total sense, but hopefully everyone else has done the game theory at least a step or two beyond that.

You mean like calaude code's integration with play right ?

Re: ChatGPT agent: bridging research and action

#165

Post by sensanaty »

selfmodruntime wrote: post #53
belter wrote: post #21

Earlier quoted context omitted.

By 2030 Europe will be known for croissants and colossal brains.

The European livestyle isn't god given and has to be paid for. It's a luxury and I'm still puzzled that people don't get that we can't afford it without an economy.

We'll only be able to afford our lifestyles by letting OpenAI's bots make spreadsheets that aren't accurate or useful outside of tricking people into thinking you did your job?

Re: ChatGPT agent: bridging research and action

#166

Post by closewith »

nlawalker wrote: post #132
travelalberta wrote: post #59

Earlier quoted context omitted.

I think this is my favorite part of the LLM hype train: the butterfly effect of dependence on an undependable stochastic system propagates errors up the chain until the whole system is worthless. "I think it got 98% of the information correct..." how do you know how much is correct without doing the whole thing properly yourself? The two options are: - Do the whole thing yourself to validate - Skim 40% of it, 'seems…

> Meanwhile an agent that you accept to get only 98% of things right is meeting expectations. Well yeah, because the agent is so much cheaper and faster than a human that you can eat the cost of the mistakes and everything that comes with them and still come out way ahead. No, of course that doesn't work in aircraft manufacturing or medicine or coding or many other scenarios that get tossed around on HN, but it does…

Definitely would work in coding. Most software companies can only dream of a 2% defect rate. Reality is probably closer to 98%, which is why we have so much organisational overhead around finding and fixing human error in software.

Re: ChatGPT agent: bridging research and action

#167

Post by 2oMg3YWV26eKIs »

The security risks with this sound scary. Let's say you give it access to your email and calendar. Now it knows all of your deepest secrets. The linked article acknowledges that prompt injection is a risk for the agent:

> Prompt injections are attempts by third parties to manipulate its behavior through malicious instructions that ChatGPT agent may encounter on the web while completing a task. For example, a malicious prompt hidden in a webpage, such as in invisible elements or metadata, could trick the agent into taking unintended actions, like sharing private data from a connector with the attacker, or taking a harmful action on a site the user has logged into.

A malicious website could trick the agent into divulging your deepest secrets!

I am curious about one thing -- the article mentions the agent will ask for permission before doing consequential actions:

> Explicit user confirmation: ChatGPT is trained to explicitly ask for your permission before taking actions with real-world consequences, like making a purchase.

How does the agent know a task is consequential? Could it mistakenly make a purchase without first asking for permission? I assume it's AI all the way down, so I assume mistakes like this are possible.

Re: ChatGPT agent: bridging research and action

#168

Post by tomashubelbauer »

modeless wrote: post #139
pants2 wrote: post #70

I've been using OpenAI operator for some time - but more and more websites are blocking it, such as LinkedIn and Amazon. That's two key use-cases gone (applying to jobs and online shopping). Operator is pretty low-key, but once Agent starts getting popular, more sites will block it. They'll need to allow a proxy configuration or something like that.

Agents respecting robots.txt is clearly going to end soon. Users will be installing browser extensions or full browsers that run the actions on their local computer with the user's own cookie jar, IP address, etc.

I wonder how many people will think they are being clever by using the Playwright MCP or browser extensions to bypass robots.txt on the sites blocking the direct use of ChatGPT Agent and will end up with their primary Google/LinkedIn/whatever accounts blocked for robotic activity.
Post reply on HN