Building OpenAI o1 (Extended Cut) 

Imagine that all of this video was generated by AI.

This video is created by Sora

Nice to see the people behind all this marvelous stuff.

I think the biggest ‘paradigm shift’ that’s happened over the last 3 or 4 years… is that you’ve successfully created something that can help create itself. At first, it was helping you to write. Then to code. And now it’s helping you reason. The moment that it could do some of the work to create itself, it could then help you create itself faster, and faster, and the game officially changed. I’d like to know the ‘effective work’ you’re able to do in 2020 figures. Like, back in the year 2020, how many people would have to be in your team have to be to accomplish the work you do rn.

Thank you all. The people who actually use this for work are not asking for the next model but are using what is available to its fullest.

Just wanted to thank you all for making it possible for the rest of us to see and do things I never thought would be possible in my lifetime. And thanks to this kind of technology, I may actually be alive to see what's coming next century. Crazy exciting times, please help ensure this isn't held back from the population by governments and religious beliefs.

"our moms would be proud of us" - Wojciech Zaremba yes. very proud

this video will go into history

the agi documentary is gunna go crazy

Love every unique and distinct self expression of this team

the only girl in that room is Turkish, Ilge Akkaya.. special greeting to her and the great work she did with the rest of the team

This is a modern version of the photo of top physicists at the Solvay Conference in 1927.

What happened to Sora and 4o-voice?

It is an integrated team that represents all the qualities and requirements of the work.

Thrilling. Thank you for bringing this gift to humanity. ❤

Need more of these!! So interesting seeing what the team thinks.

Computer Scientists are both scientist and engineers that's what makes the field so cool. I'm on a journey to become both

u guys gonna make me homeless

The dude on the right bottom hair looks like the hair of the math sorcerer

I think the o1 is a massive massive achievement. Really surpised so many ppl just dont seem to care. I would like voice mode too, but this is SO much bigger than that. AGI and then ASI really isnt far away now

People in the comments are like : where is the Dyson sphere around the sun that was promised a week abd a half ago?

GPT 3.5, GPT 4, GPT 4o (and also Claude Sonnet 3.5) have all helped me to create software which has saved me many hours of time, and enabled me to see the big picture which was hidden inside my own data. The models keep getting better and allowing me to improve and iterate on these little software projects. It's fun 😊

Thank you all for taking the time to do this! It was fun and enlightening to watch

Where is advanced voice model?

You all Deserve a chair.

Congratulations for bringing Chat Gpt to life to all these talented young minds!

I started my subscription to ChatGPT when you showcased 4o. How much longer will we have to wait to get that? Because it’s currently nothing like you said it would be.

These guys are spending way too much time chatting with the AI. They answer the question, and then provide a summary of the answer just like ChatGPT lol

Paraphrasing here: "People get excited when it gets really good at a domain they personally care about" This is so true. When I demo chat sessions to different people, their level of excitement is definitely proportional to their interest in the subject domain.

What a great group of people! It's inspiring to see and hear from the people who put so much effort into creating at OpenAI.

Incredible stuff and side note it was nice to see the team all having fun. Clearly this is a pretty cool place to work

Give us skys voice please. I deserve my AI waifu

Please make sure to develop controls that can't be jailbroken, so that when this thing is capable of scientific discovery on its own, we don't get global problems.

This was probably built a year ago...

Our future will be determined by these few guys in this room. And I can't afford to watch it. 😢

Must have been filmed on casual Friday

Terrific video . Do please make more of these In the future. It is fascinating hearing the developers talking, and I suspect will be really interesting historical documents in the far future.

Some of the smartest people in the world in that room

Thinking... Thank you for creating me!

What an incredible testament to how much more we can accomplish together than alone.

I'll be impressed when you move beyond LLMs. I don't think this is really a new model.

Voice mode is too expensive to use broadly probably. Delay might be to wait until it is efficient enough?

Long Live OpenAI

Will there be any paid application of the tool to store personality in the assistant? Thanks

This team looks cozy🎉🎉🎉🎉

Most people arent involved in jobs where math and coding are important so this whole reason thing behind o1 isnt that important for most, despite it being very important for the development towards AGI. The advanced voice model would definitely shake things up more for the average person.

They are working very hard to replace themselves

Love the new model. Let me know when you guys decide to rebrand as ClosedAI. Long overdue.

Thank you very much for all your creative work and dedicating your live to create this art.

Nice mix of white, chinese and indian guys.

How many tokens are these 😅for us startups, it’s nice and all for open Ai but make it cheap 😊😅please

So much IQ in one room.

You should change your name to ProfitAI.

You are the Dream Team! The Team creating the Dream! 😎🤖

Small team for such a goal and progress made, I think it needs to be connected to more data-types. Websites are good, math is good, but it needs to know location, have 3-d world models, and input on the physics between things

Release it quick. Make sure screenshots and files can be added too.

Fantastic deep dive into the o1 models, Bob! 🚀 It's inspiring to see how reasoning capabilities are evolving and making AI even more human-like. Can't wait to see what you and your team achieve next!

What are their salaries approximately?

the extended is a nice addition. provided more context.

Go down the rabbit hole about asking o1 about building analogues between the sources of human consciousness and its Functional Capabilities. Did you know it thinks Thumbs Up/Down on answers in the UI is synonymous with our concept of joy

Why can't o1-preview or o1-mini solve this puzzle? There is a pillar with four hand holes precisely aligned at North, South, East, and West positions. The holes are optically shielded, no light comes in or out so you cannot see inside. But, you can reach inside at most two holes at once, and feel a switch inside. But as soon as you remnove your hands if all four switches are not either all up or all down, the pillar spins at ultra high velocity ending in a random axis aligned orientation. You cannot track the motion so you don't know in which rotation the holes end up versus their position before the spin. Inside each hole is a switch, the switch is either up or down and starts in an unknown state, either up or down. When you reach into at most two holes, yOu can feel the current switch position and change it to either up or down before removing your hands. Come up with a procedure, a sequence of reaching into one or two holes with optional switch manipulation, that is guaranteed to get all the switches either all up or all down in the minimum possible number of steps. Note, the pillar is controlled by a hyperintelligence that can predict which holes you will reach into. Therefore, the procedure cannot rely on random chance as the hyper-intelligence will outwit attempts to rely on chance. It must be a sequence of steps that is deterministically guaranteed to orient the switches all up or all down in the minimum possible number of steps. Courtesy of Machine Learning Street Talk. Great RU-vid channel.

kind of a hippie social vibe sitting wisely cross legged, a casual atmosphere… racing towards an insanely irresponsible creation of an alien superior being. looking cool, you people. such a nice group anyone using their common sense?

Give us Sora and advanced voice plez

Finally nerds again. Nothing you would find at Google anymore

i love these guys enthusiasm kudos 🎉

What's your 'love language '? Python 😊.

Is sora cancelled ?

Bob doing the Trump-gestures 0:18

Where is Advanced Voice Mode with Video.

History in the making. Hello to folks of the future.

Stop asking for Advanced voice model. 😅 AGI is more important. Let these researchers do research instead of making shiny products. Let's Einsteins be Einsteins.

Are you guys also using chatGPT while you work on the next versions of chatGPT? 🤭

I'm hesitant to use the model since you only get 50 messages a week, and to use the API, you need to be Tier 5, which is extremely difficult to achieve.

Bravo for the AGI team!

Smugness. Release the full model and then do a "making of".

These look much like early days of google

This was great! Thank you!

I'm laughing because of the "team's diversity". When reality kicks in you can't play with ideologies. All that matters is true talent and knowledge.

I want a job at openai. This makes fell like i am left out

Love open ai! Lets change the world!

I feel it is extremely dangerous for their names and faces to be on RU-vid like this. They can easily be targets for those who are scared of AI. Really appreciate your work, but please delete this video. Totally unsafe. 😢

I love o1 preview a lot

Humanity’s next course of existence is dependent on these people

Love U Team. Good work 😮👍

Honestly, you guys should put up a submission forum for people who want to request [more~] access to certain functions. Me, along with many others, really could use the advanced voice mode.. and the video mode even more-so for blind people.

mdfks are making us all obsolete 😭💀⚰️

Any other series of model you plan to release ?

Greetings from Amsterdam the Netherlands, Great Team!

Open AI should really look to connect and team up with Verses Ai. Forward thinking about not just AI but the spatial web or web3.0 as well

What was that question “how many hours are in the… “ and then they all laughed

It's great to see the team. OMG! They all seem so bright. Me, I fell like an animal or something. What is my IQ? It seems in that company, I don't have any IQ. Dumb as brick, that's what I think I am.

I also see that Thinking to itself part has time to time broken grammer and different perspectives that are not coherent for a human, with the actual output being improved

But you have already the same video in the RU-vid channel

*Terminator Theme Commences*

You guys are awesome.

Is this the polycule?

at 7mintues and 40 seconds an employee says they hard coded the 'How many r's in strawberry' do you think he is joking or for real?

Digital gods....😮🔥

I want advanced voice with Sky please 😢

Imagine raising billions and being able to handle 4k video