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

Operator research preview

openai.com

Post reply on HN
Print view View original

401–410 of 448 posts

Re: Operator research preview

#401

Post by noch »

dougb5 wrote: post #88

> We’re collaborating with companies like DoorDash, Instacart, OpenTable, Priceline, StubHub, Thumbtack, Uber, and others to ensure Operator addresses real-world needs while respecting established norms. Are these tasks really complex enough for people that they are itching to relegate the remaining scrap of required labor to a machine? I always feel like I'm missing something when companies hold up restaurant reserv…

> Are these tasks really complex enough for people that they are itching to relegate the remaining scrap of required labor to a machine?

I think I sympathize with your feeling but I don't agree with the premise of the question. Do you have or have you ever had a human personal assistant or secretary?

An effective human personal assistant can feel like a gift from God. Suddenly a lot of the things that prevent you from concentrating on what you absolutely must focus on, especially if you have a busy life, are magically sorted out. The person knows what you need and knows when you need it and gets it for you; they understand what you ask for and guess what you forgot to ask for. Things you needed organized become organized while you work after giving minimal instructions. Life just gets so much better!

When I imagine that machines might be able to become good or effective personal assistants for everyone … If this stuff ever works well it will be a huge life upgrade for everyone. Imagine always having someone who can help you, ready to help you. My father would call the secretary pool to send someone to his office. My kids will probably just speak and powerful machines will show up to help.

Re: Operator research preview

#402

Post by xboxnolifes »

ben_w wrote: post #398
franktankbank wrote: post #277

Earlier quoted context omitted.

> There's plenty of data avaliable to anyone I think you are just wrong here and so everything that follows is wishful.

Wikipedia exists and can be downloaded, and the full archive includes talk pages, CC-BY-SA: https://en.wikipedia.org/wiki/Wikipedia:Database_download "Plenty" may be vague , but it's not wrong .

Does wikipedia help bots make coherent decisions for my personal grocery shopping based on my own habits, price point, allergies, and preferences?

Re: Operator research preview

#403

Post by ickelbawd »

energy123 wrote: post #347
yoshicoder wrote: post #18

I am a little concerned with letting an AI agent that routinely hallucinates control my browser. I can't not watch it do the task, in case it messes up. So I am not sure what the value is versus me doing it myself.

I had the same concern but looking at the demos, they run it in a sandbox so the worst case failures should be constrained.

How do they sandbox third party applications though… They can’t. Whoops the AI deleted your account…

Re: Operator research preview

#404

Post by drewbeck »

sky2224 wrote: post #191
TeMPOraL wrote: post #169

Earlier quoted context omitted.

I don't. Chat interface sucks; for most of these things, a more direct interface could be much more ergonomic, and easier to operate and integrate. The only reason we don't have those interfaces is because neither restaurants, nor airlines, nor online stores, nor any other businesses actually want us to have them. To a business, the user interface isn't there to help the user achieve their goals - it's a platform for…

I think the chat interface is bad, but for certain things it could honestly streamline a lot of mundane things as the poster you're replying two stated. For example, McDonald's has heavily shifted away from cashiers taking orders and instead is using the kiosks to have customers order. The downside of this is 1) it's incredibly unsanitary and 2) customers are so goddamn slow at tapping on that god awful screen. An AI…

McDonald’s makes a lot more money with the kiosks. Slowness is an issue but the upselling is major, and putting a lot of images of tasty looking things in front of a hungry person is very effective. Chat could never do this!

Re: Operator research preview

#405

Post by beefnugs »

ansonhw wrote: post #331
karpathy wrote: post #246

Earlier quoted context omitted.

This is the video of how the parent comment was made: https://operator.chatgpt.com/v/6792c91d5b048192b5dbad908de05... But there are many things missing from the video. - Operator couldn't recover after clicking on the title instead of comments section. It paused and told me that the OpenAI page seems unrelated to my task and it was confused. I had to tell it to click on "comments" not on the title. - Not shown is tha…

When: 1/ More accurate + less supervision 2/ Faster 3/ Background + Parallel Will be the magic moment. Currently like supervising a grandparent using the web. But there's huge potential if ^^^ happens. Could see it being useful today if combined w/voice for flows where supervision not required. Example: asking w/voice to make reservations while driving.

This will never actually happen. Because there is no true feedback.

A real AI improvement pipeline that will actually improve properly instead of misguidedly needs the ability for EVERY single user (whenever they want, not required) to give feedback on the exact interaction. Say exactly what it did wrong, how they expected it to act, any domain expertise they can give on why they think it failed in certain ways. Then the developers can make decisions based on the real fuckups. This isn't happening anywhere.

Re: Operator research preview

#406

Post by entropi »

wafflemaker wrote: post #365
gyomu wrote: post #255

Earlier quoted context omitted.

@dang can we have guidelines against posting AI generated content here? (who cares if the account is "human operated" or has a disclaimer). It's just lame and not what this forum is about.

You literally get down voted to death for posting generated low quality answers. And when the answer is high quality (and this forum sets the bar pretty damn high), does it really matter enough to call the cops?

I don't come here to "consume high quality content". I come here to engage in discussions with often tech-oriented yet quite diverse set of people. The distinction I think is quite an important one.

In my opinion Karpathy's generated answer was followed up by an insightful, actual comment so it is fine; as long as such things are the exception and not the rule.

Re: Operator research preview

#407

Post by drewbeck »

n144q wrote: post #221
gordon_freeman wrote: post #155

What is fascinating about this announcement is if you look into future after considerable improvements in product and the model, we will be just chatting with ChatGPT to book dinner tables, flights, buy groceries and do all sort of mundane and hugely boring things we do on the web, just by talking to the agents. I'd definitely love that.

Not until ChatGPT can do these things as reliably as concierge service, and provide full refund for any situation it messes up. I am not looking forward to a trip booked for wrong dates with the hotel name confused/hallucinated for a different one.

Yeah the failure states are really an issue. Happy path looks magical but there are so many ways that it can go wrong. And you don’t have the fallback of an actual human you’re talking to to clear it up.

Re: Operator research preview

#408

Post by lolinder »

lm28469 wrote: post #352
sumedh wrote: post #280

Earlier quoted context omitted.

> Are our attention spans so shot that we consider booking a reservation at a restaurant or buying groceries "hugely boring"? Dont be limited with these examples. How about Airline booking, try different airlines, go to the confirmation screen. then the user can check if everything is allright and if the user wants to finish the booking on the most cheapest one.

Google flight does that for you, and your browser can already full in 80% of the form fields. I don't remember spending more than 1 minute booking tickets. Deciding where to go takes 50-100x more time, the booking speed is such a non issue. What's the goal of technology ? Automate everything so that we don't have to live anymore ? We might as well build matrix pods at that point

Exactly. 90% of the time I spend booking tickets is in reading every single field a few extra times before clicking "yes, please charge my card $N00". I'm not about to outsource that confirmation step to an LLM, and outsourcing anything else isn't going to save any real time.

Re: Operator research preview

#409

Post by ben_w »

xboxnolifes wrote: post #402
ben_w wrote: post #398

Earlier quoted context omitted.

Wikipedia exists and can be downloaded, and the full archive includes talk pages, CC-BY-SA: https://en.wikipedia.org/wiki/Wikipedia:Database_download "Plenty" may be vague , but it's not wrong .

Does wikipedia help bots make coherent decisions for my personal grocery shopping based on my own habits, price point, allergies, and preferences?

Why wouldn't it? Are your allergies not listed, is your local economy not explained and thus median expectations of price sensitivity not inferrable, are the words you use to describe your preferences and habits so out of distribution?

Re: Operator research preview

#410

Post by kridsdale3 »

bko wrote: post #60
minimaxir wrote: post #8

Overall, Operator seems the same as Claude's Computer Use demo from a few months ago, including architecture requiring user to launch a VM, and a tendency to be incorrect: https://news.ycombinator.com/item?id=41914989 Notably, Claude's Computer Use implementation made few waves in the AI Agent industry since that announcement despite the hype.

I thought Claude Computer Use is through API, and I remember hearing about high number of queries and charges. This looks like its in browser through the standard $20 Pro fee, which is huge. (EDIT: $200 a month plan so less of a slam dunk but still might be worth it) Is there any open source or cheap ways to automate things on your computer? For instance I was thinking about a workflow like: 1. Use web to search for…

> Is there any open source or cheap ways to automate things on your computer?

MacOS has had Automator since 2005. It's far more like "programming" to use than a 2024-tier ML based system, but it was designed for non-programmers, and lots of people do use it.

Personally, I hate it.

Post reply on HN