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

201–210 of 448 posts

Re: Operator research preview

#201

Post by reissbaker »

mrdependable wrote: post #161

A lot of people here seem to think this is somehow for their benefit, or that OpenAI and friends are trying to make something useful for the average person. They aren't spending billions of dollars to give everyone a personal assistant. They are spending billions now to save even more in wages later, and we are paying for the privilege of training their AI to do it. By the time this thing is useful enough to actually…

This seems unreasonably pessimistic (or unreasonably optimistic in OpenAI's moat?). There are so, so many companies competing in this space. The cost will reflect the price of the hardware needed to run it: if it doesn't, they'll just lose to one of their many competitors who offer something similar for cheaper, e.g. whatever DeepSeek or Meta releases in the same space, with the cost driven to the bottom by commoditized inference companies like Together and Fireworks. And hardware cost goes down over time: even if it's unaffordable at launch, it won't be in five years.

They're not even the first movers here: Anthropic's been doing this with Claude for a few months now. They're just the first to combine it with a reasoning-style model, and I'd expect Anthropic to launch a similar model within the next few months if not sooner, especially now that there's been open-source replication of o1-style reasoning with DeepSeek R1 and the various R1-distills on top of Llama and Qwen.

Re: Operator research preview

#203

Post by ItsMattyG »

thrtythreeforty wrote: post #57
brap wrote: post #31

I don't know why, but the approach where "agents" accomplish things by using a mouse and keyboard and looking at pixels always seemed off to me. I understand that in theory it's more flexible, but I always imagined some sort of standard, where apps and services can expose a set of pre-approved actions on the user's behalf. And the user can add/revoke privileges from agents at any point. Kind of like OAuth scopes. Ima…

APIs have an MxN problem. N tools each need to implement M different APIs. In nearly every case (that an end user cares about), an API will also have a GUI frontend. The GUI is discoverable, able to be authenticated against, definitely exists, and generally usable by the lowest common denominator. Teaching the AI to use this generically, solves the same problem as implementing support for a bunch of APIs without the…

But if you have an AI then all that's needed to implement an api is documentation

Re: Operator research preview

#204

Post by infecto »

darioush wrote: post #151
mattstir wrote: post #102

Earlier quoted context omitted.

I can sympathize with vague notions of AI dystopia, but this might be stretching the concept a bit too far. This kind of service is extremely abusable ("Operator, go to Wikipedia and start mass-vandalizing articles" or "Go to this website and try these people's email addresses with random passwords until it locks their accounts") and building some alignment goals into it doesn't seem like a terribly draconian idea. A…

You can also write a python script to achieve the same goals. Except it's not python's responsibility to interpret the intent of your script, just as it's not your phone's responsibility to interpret the contents of your conversation. So our tools are not our morality police. We have a legal system that can operate within the bounds of law and due process. I am well aware of the already applied levels of machine lear…

The difference being you would be running that python script yourself. If you by chance hosted it somewhere there is high probability that the host would get a notice and shut you down. I honestly don't see much difference here. There will be multiple providers and perhaps great ways to run these types of tools locally, all have different risk measures.

Re: Operator research preview

#206

Post by aerostable_slug »

itskarad wrote: post #164

I think this opens a new direction in terms of UI for companies like Instacart or Doordash — they can now optimise marketing for LLMs in place of humans, so they can just give benchmarks or quantized results for a product so the LLM can make a decision, instead of presenting the highest converting products first. If the operator is told to find the most nutritious eggs for weight gain, the agent can refer to the nutr…

This reminds me of a scene in the latest entry to the Alien film franchise where the protagonists traverse a passage designated for 'artificial human' use only (it's dark and rather claustrophobic).

In the future we might well stumble into those kind of spaces on the net accidentally, look around briefly, then excuse ourselves back to the well-lit spaces meant for real people.

Re: Operator research preview

#207

Post by 0_____0 »

dougb5 wrote: post #189
windowlessmonad wrote: post #177

Earlier quoted context omitted.

Are our attention spans so shot that we consider booking a reservation at a restaurant or buying groceries "hugely boring"? And do we value convenience so much that we're willing to sacrifice a huge breadth of options for whatever sponsor du jour OpenAI wants to serve us just to save less than 10 minutes? And would this company spend billions of dollars for this infinitesimally small increase in convenience? No, of c…

I'm reminded of Kurt Vonnegut's famous story about buying postage stamps: https://www.insidehook.com/wellness/kurt-vonnegut-advice "I stamp the envelope and mail it in a mailbox in front of the post office, and I go home. And I’ve had a hell of a good time. And I tell you, we are here on Earth to fart around, and don’t let anybody tell you any different...How beautiful it is to get up and go do something."

I love so much. It really encapsulates what I've been feeling about tech and life generally. Society and especially tech seems so efficiency minded that I feel like a crazy person for going to do my groceries at the store sometimes.

Re: Operator research preview

#208

Post by Matl »

darioush wrote: post #151
mattstir wrote: post #102

Earlier quoted context omitted.

I can sympathize with vague notions of AI dystopia, but this might be stretching the concept a bit too far. This kind of service is extremely abusable ("Operator, go to Wikipedia and start mass-vandalizing articles" or "Go to this website and try these people's email addresses with random passwords until it locks their accounts") and building some alignment goals into it doesn't seem like a terribly draconian idea. A…

You can also write a python script to achieve the same goals. Except it's not python's responsibility to interpret the intent of your script, just as it's not your phone's responsibility to interpret the contents of your conversation. So our tools are not our morality police. We have a legal system that can operate within the bounds of law and due process. I am well aware of the already applied levels of machine lear…

> You can also write a python script to achieve the same goals.

First of all, I agree with you generally and am uneasy about this too.

But there's a difference in that someone could say 'hey, this attack on my website happened from OpenAI's infra', whereas that would not apply to Python because it's not a hosted service.

Re: Operator research preview

#209

Post by openrisk »

windowlessmonad wrote: post #177
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.

Are our attention spans so shot that we consider booking a reservation at a restaurant or buying groceries "hugely boring"? And do we value convenience so much that we're willing to sacrifice a huge breadth of options for whatever sponsor du jour OpenAI wants to serve us just to save less than 10 minutes? And would this company spend billions of dollars for this infinitesimally small increase in convenience? No, of c…

The fact that you are downvoted despite pointing the obvious tells you about the odds of the tech industry adopting a different path. Fleecing the ignoramy is the name of the game.

Re: Operator research preview

#210

Post by 4ndrewl »

Night_Thastus wrote: post #187
mrdependable wrote: post #161

A lot of people here seem to think this is somehow for their benefit, or that OpenAI and friends are trying to make something useful for the average person. They aren't spending billions of dollars to give everyone a personal assistant. They are spending billions now to save even more in wages later, and we are paying for the privilege of training their AI to do it. By the time this thing is useful enough to actually…

Don't worry, it'll never be good enough to actually be a personal assistant.

Not this version, but in 3 years time. Promise.

Just keeping sending us money...

Post reply on HN