Hacker News Reader

FYI, there's a "Save ChatGPT as PDF" Chrome extension [1].

I wouldn't use on a ChatGPT for Business subscription (it may be against your company's policies to export anything), but very convenient for personal use.

https://chromewebstore.google.com/detail/save-chatgpt-as-pdf...

Wow, that is impressive! How were you able to use o1-preview? I pay for ChatGPT, but on chatgpt.com in the model selector I only see 4o, 4o-mini, and 4. Is o1 in that list for you, or is it somewhere else?

Yes, o1-preview is on the list, as is o1-mini for me (Tier 5, early 2021 API user), under "reasoning".

It appeared for me about thirty minutes after I first checked.

Likely phased rollout throughout the day today to prevent spikes

“Throughout the day” lol. Advanced voice mode still hasn’t shown up.

They seem to care more about influencers than paying supporters.

> lol.

It's there for a lot of people already. I can see it on 3 different accounts. Including org and just regular paid accounts.

It's my understanding paying supporters aren't actually paying enough to cover costs, that $20 isn't nearly enough - in that context, a gradual roll-out seems fair. Though maybe they could introduce a couple more higher-paid tiers to give people the option to pay for early access

Not true; it's already available for me, both O1 and O1-mini. It seems they are indeed rolling out gradually (as any company does).

You got advanced voice mode? I did get o1 preview just a while ago.

You got o1, or o1 preview?

o1-preview and o1-mini. I don't think o1 is publicly available yet.

And I assume voice mode is like Sora; a nice PR play.

It's available for me. Regular paying customer in the UK.

I see it in the mac and iOS app.

Like others here, it was just available on the website and app when I checked. FWIW I still don’t have advanced voice mode.

I don’t have either the new model nor the advanced voice mode as a paying user.

u do just use this link: https://chatgpt.com/?model=o1-preview

That worked. Now can you do that for advanced voice mode??? Pretty please!

Haha I wish, although I saw the other one i forgot its name which makes music for you, now you can ask it for a soundtrack and it gives it back to you in your voice or something like that idk interesting times are ahead for sure!

Wait what is this? Tell me more please

I heard on X suno.com has this feature but couldn’t find it maybe its coming soon? Idk but there are ways u can do it, maybe it was a different service suno is pretty cool tho

they are rolling it out slowly, this link doesn't enable access. they only gave me access around 4:30pm PT

I think they're rolling it out gradually today. I don't see it listed (in the browser, Mac app or Android app).

Available on ChatGPT Plus signature or only using the API?

it's in my MacOS app, but not in the browser fir the same account

The linked release mentions trusted users and links to the usage tier limits. Looking at the pricing, o1-preview only appears for tier 5 - requiring 1k+ spend and initial spend 30+ days ago

edit: sorry - this is for API :)

Isn't there a big "Share" button at the top right of the chatgpt interface? Or are you using another front end?

In ChatGPT for Business it limits sharing among users in my org, without an option for public sharing.

I often click on those links and get an error that they are unavailable. I’m not sure if it’s openAI trying to prevent people from sharing evidence of the model behaving badly, or an innocuous explanation like the links are temporary.

They were probably generated using a business account, and the business does not allow public links.

In context, a lot of times it’s clear that the link worked at first (other people who could see it responded) but when I click later, it’s broken.

The link also breaks if the original user deletes the chat that was being linked to, whether on purpose or without realizing it would also break the link.

Even for regular users, the Share button is not always available or functional. It works sometimes, and other times it disappears. For example, since today, I have no Share button at all for chats.

My share chat link moved into the sidebar in the … menu to the right of each chat title (MacOS Safari).

Ah, I see it there now. Thanks.

I'm impressed. I had two modified logic puzzles where ChatGPT-4 fails but o1 succeeds. The training data had too many instances of the unmodified puzzle, so 4 wouldn't get it right. o1 manages to not get tripped up by them.

https://chatgpt.com/share/66e35c37-60c4-8009-8cf9-8fe61f57d3...

https://chatgpt.com/share/66e35f0e-6c98-8009-a128-e9ac677480...

Great progress, I asked GPT-4o and o1-preview to create a python script to make $100 quickly, o1 came up with a very interesting result:

https://x.com/soheil/status/1834320893331587353

The screenshot [1] is not readable for me. Chrome, Android. It's so blurry that I cant recognize a single character. How do other people read it? The resolution is 84x800.

When I click on the image, it expands to full res, 1713x16392.3

> it expands to full res, 1713x16392.3

Three tenths of a pixel is an interesting resolution…

(The actual res is 1045 × 10000 ; you've multiplied by 1.63923 somehow…?)

I agree,

But it’s what I got when I went to Inspect element > hover over the image

Size it expanded to vs real image size I guess

Pixels have been "non-real" for a long time.

In some contexts. In this context (a PNG), they're very real.

This context is the moreso the browser, complete with it's own sub-pixels, aliasing, simulated/real blurring, zooming, etc.

But file-format context, yes, PNG, BMP, and TFF are the real lossless image kingpins.

Click on it for full resolution

I didn't work until I switched to "Desktop Site" in the browser menu, as a sibling comment suggested. Then the page reloads with various buttons, etc. Until that just the preview image not reacting to clicks.

When you open on phone, switch to "desktop site" via browser three dots menu

Yes, that works - the page reloads with postimg.cc UI and clicking the image opens full resolution.

Direct link to full resolution: https://i.postimg.cc/D74LJb45/SCR-20240912-sdko.png

thank you

What if you copy the whole reasoning process example provided by OpenAI, use it as a system prompt (to teach how to reason), use that system prompt in Claude, got4o etc?

It might work a little bit. It's like doing few shot prompting instead of training it to reason.

This is a brilliant hypothesis deconstruction. I am sure others will now be able to test as well and this should confirm their engineering.

Did you edit the message? I cannot see anything now in the screenshot, too low resolution

You need to click on the image for the high res version to load. Sorry, it’s awkward.

The website seems to redirect me to a low resolution image, the first time I clicked on the link it worked as you are saying.

Very cool. It gets the conclusion right, but it did confuse itself briefly after interpreting `256 * last_byte + second_to_last_byte` as big-endian. It's neat that it corrected the confusion, but a little unsatisfying that it doesn't explicitly identify the mistake the way a human would.

What is the brand of the fan? Same problem here with proprietary hood fan...

InVENTer Pulsar

is it better than Claude?

Neither Sonnet nor Opus could solve it or get close in a minimal test I did just now, using the same prompt as above.

Sonnet: https://pastebin.com/24QG3JkN

Opus: https://pastebin.com/PJM99pdy

I think this new model is a generational leap above Claude for tasks that require complex reasoning.

Way worse than Claude for solving a cipher. Not even 1/10th as good. Just one data point, ymmv.

Thanks for sharing this, incredible stuff.

second is very blurry

When you click on the image it loads a higher res version.

Solved here: https://news.ycombinator.com/item?id=41525164

What's the incredible part here? Being able to write code to turn hex into decimal?

Also, if you actually read the "chain of thought" contains several embarrassing contradictions and incoherent sentences. If a junior developer wrote this analysis, I'd send them back to reread the fundamentals.

What about thoughts themselves? There are plenty of times I start a thought and realize it doesn't make sense. It's part of the thinking process.

Well, it doesn't "correct" itself later. It just says wrong things and gets the right answer anyways, because this encoding is so simple that many college freshmen could figure it out in their heads.

Read the transcript with a critical eye instead of just skimming it, you'll see what I mean.

> Asking the models to determine if my code is equivalent to what they reverse engineered resulted in a nuanced and thorough examination, and eventual conclusion that it is equivalent.

Did you actually implement to see if it works out of the box ?

Also if you are a free users or accepted that your chats should be used for training then maybe o1 is was just trained on your previous chat and so now knows how to reason about that particular type of problems

That's not how LLM training works.

so it is impossible to use the free user chats to train models ??????

That is an interesting thought. This was all done in an account that is opted out of training though.

I have tested the Python code o1 created to decode the timestamps and it works as expected.