ChatGPT O3 Preview Announcement

by franze • 2 years ago

85 42 2 years ago

jsheard• 2 years ago

Since OpenAI bills you for tokens expended on chain-of-thought, I assume this "deliberative alignment" has the funny side-effect of making you pay for the time the model spends ruminating on whether it should refuse to do what you asked it to do. Talk about adding insult to injury.

xxpor• 2 years ago

Maybe the most convincing argument yet that this is similar to how humans think.

skinner_• 2 years ago

Okay, but is this similar to how humans bill you?

whamlastxmas• 2 years ago

I don’t really see this any differently from an API billing you per call and still billing you for calls that where you pass in wrong credentials and get unauthorized error responses

Incipient• 2 years ago

I don't think I've ever seen an API bill for incorrect credentials, or even anything that returns a non-200?

486sx33• 2 years ago

So … hypothetically, I Spent $5 on tokens to figure out the answer - then got an answer that my answer is blocked !

rmbyrro• 2 years ago

using the service is optional and they're clear about how pricing works, though

one can always refuse to use

throwaway3141552• 2 years ago

    def should_upvote(headline: str) -> bool:
        if "o3" in headline.lower():
            return True
        return False

Seriously though, is there anything new here? Also why the need for the editorialized headline? The article is titled "12 Days of OpenAI", not "ChatGPT O3 Preview Announcement" (which frankly makes it sound like it's about to be available to the public, which it isn't).

layer83• 2 years ago

   return "o3" in headline.lower()

would have done the job.

throwaway3141552• 2 years ago

This is more painful than any downvote. Have an upvote.

jordanmorgan10• 2 years ago

This video format is killing me. Would love to keep up with the announcement but I have no desire for 10-20 minute videos.

layer83• 2 years ago

AI video summary at your service:

In this video, Sam Altman and team introduce OpenAI o3 and o3-mini, showcasing their advanced reasoning capabilities and performance benchmarks compared to earlier models. The session highlights the models' robust accuracy in coding and mathematical tasks, as well as new safety testing initiatives involving community participation. The o3-mini is noted for its cost-efficient performance, while safety strategies are enhanced through deliberative alignment.

## Key Points

### Introduction to o3 and o3-mini models

During the final day of their 12-day event, OpenAI announces two new models: o3 and o3-mini. These models are positioned as advancements in AI reasoning capabilities, following the success of the earlier o1 model.

### Performance benchmarks

The o3 model achieves significant improvements in capabilities, scoring 71.7% on software benchmarks and reaching near-expert levels on competitive math exams. In comparison to o1, it shows over 20% improvement in coding tasks.

### Safety testing initiatives

OpenAI emphasizes safety testing for the new models, opening access for researchers to facilitate public testing. The goal is to ensure models are safe for general use while their capabilities are further validated.

### Introduction of o3-mini

The o3-mini is presented as a cost-efficient alternative to o3, also equipped with strong reasoning powers, allowing adjustable thinking time to optimize performance based on user needs.

### New safety techniques: deliberative alignment

A new safety training technique, deliberative alignment, uses reasoning capabilities of the models to create better safety benchmarks, enhancing the ability to accurately reject unsafe prompts.

### Future developments and availability

The video concludes with information on how to apply for early access to test these models with expectations of full public availability in early 2025, alongside a call for safety researchers to contribute.

karimdaghari• 2 years ago

It's a bit ironic that it got the availability date wrong no?

in-pursuit3• 2 years ago

Can this “expert in math” do binary addition? o1 falls apart after 10 bits, which is easily memorizable.

uncomplexity_3• 2 years ago

the mini models are the most exciting to me. they seem to hit the balance of general utility and cost.

people are fond of comparing large bleeding edge models, but when we compare the small ones to say gpt3.5, it is still astonishing.

to me the smaller models being the balance of intelligence and cost is the best indicator of what the general population can use and afford.

the larger models end up being used mostly by individuals teams and orgs who can afford to pay for it and learn how to use it.

keep in mind at this day and age there is still a lot of people who dont know these things exist. it's just outside of their reality. then there are who have the slightest idea about it, but won't commit the time and money needed to learn it and fully experience it.

this is like the new generational transfer of wealth and information. even if you're an individual or a small team, with enough levers you can use these technologies to your advantage, and with more levers (like yc) you can enter the battlefield and compete with existing companies.

muratsu• 2 years ago

I strongly agree. I stopped checking their page for news after day 2 due to video only announcements.

byyoung3• 2 years ago

Summary here https://wandb.ai/byyoung3/ml-news/reports/OpenAI-Introduces-...

transcriptase• 2 years ago

Safety: not from causing you physical harm or risking property damage, since there are no guardrails against it confidently walking you through a flawed breadboard circuit that will cause components to explode or immediately catch fire. Then say

“You’re absolutely right. We shouldn’t have run 120v through that capacitor rated for 10v. Here’s an updated design that fixes the issue”, then proceed to explain the same thing with a resistor placed at random.

No we mean safety from our LLM saying something that pisses off a politician, or even worse… hurting someone’s /feelings/.

binary1323• 2 years ago

It is more accurate to say they are thinking in terms of “brand safety” or liability, that is, being safe from being blamed for making a mean robot.

sourcepluck2• 2 years ago

Well, if one of OpenAI's LLMs started quoting Mein Kampf in its answers to people, would you laugh, or would you not laugh? That's the question.

faizshah• 2 years ago

Did I miss it or did they say whether o3 is going to be available under ChatGPT Plus or Pro?

luke-stanley2• 2 years ago

I watched the live stream and don't recall any mention of it. It's not at that stage!

daft_pink• 2 years ago

We’d really likt to use this today.

luke-stanley2• 2 years ago

The title is very wrong, "ChatGPT O3" is not a thing, OpenAI's new o3 model is not even demoed in ChatGPT, they don't call it "O3". The subtitle at the top is "o3 preview & call for safety researchers". The web page title is "12 Days of OpenAI".

ChatGPT O3 Preview Announcement

Related Stories