Jeff Wang 👨‍🚀

Ann Lee

Pipibjc

Sravya Popuri

Juan Pino

I’m Jeff (@jffwng) and I’m the Product Manager for the Universal Speech Translator project at Meta’s FAIR (Fundamental AI Research). We recently released a speech translation model that supports a primarily oral-only language, Hokkien. We also released SpeechMatrix and an unsupervised speech system. On Thursday I’ll be joined by a few of my teammates: Sravya, Ann, Peng-Jen, and Juan. We're hosting the AMA on 12/1/2022 @ 10:30AM PT to share more about our work. Ask us anything!

We’re the research team working on Meta AI’s Universal Speech Translator project — Ask us anything!

Sam Altman

Brad Lightcap

We’re doubling down on AI for email - writing / replying, prioritization, NLP-enabled querying (eg what time is my flight on Monday). What do you guys think?

Any opportunities for others to get involved who aren’t currently working on an idea with a cofounder? Perhaps a list of individuals that are open to teaming up with groups that are accepted into the program?

For those of you running AI/ML departments at startups, how do you handle the tension between wanting to be very transparent about problems/how you’re solving them to attract talent and not wanting to just hand over the recipe every cool thing you’ve built to your competitors?

👋 hi there from <a target="_blank" rel="noopener noreferrer nofollow" class="css-4rbku5 css-18t94o4 css-901oao css-16my406 r-1cvl2hr r-1loqt21 r-poiln3 r-bcqeeo r-qvutc0" href="https://twitter.com/coqui_ai">@coqui_ai</a> what are the terms of the investment?

10% but they are experimenting with it

other terms? e.g. are you looking for board seats? @sama

no board seats! no other terms of note

the shortest TS in the history of the valley 😂

Does this mean OpenAI is giving up on the goal of AGI, or that the company doesn't know how to spend the extra $10 million, so its investing it in other AI companies instead?

we will remain very focused on AGI, but we think there will be a _gigantic_ amount of value unlocked for the world along the way. we want to enable startups to go after it.

No doubt that there's going to be a lot of valuable AI companies, but why OpenAI, if it doesn't fulfill the original mission of AGI?

I would love to fine-tune the "text-davinci-002" model because of the higher token limit. Do you know roughly when that will be possible?

still have a lot to figure out, but we definitely want to let people do more and better fine-tuning

Why not make these ML-specific sessions open to the public on YouTube the way YC did with Startip School?

we very likely will in the future! want to really understand what works for AI startups first; i think we are entering a pretty new era

What happens after the program ends? YC Like demo day? help in raising future rounds? More workshops and office hours?

no--this is an investment and some ML-specific sessions, not an accelerator-like thing

Some time ago I started working on generating complete programs using Codex (example output: https://github.com/tom-doerr/show_sunset_time/blob/main/main.py). I'm now able to generate complete projects, including configuration files and folders.However that requires executing a large amount of potentially dangerous code in a docker environment. Is that something that could be turned into a startup or would OpenAI not allow using Codex for this?

how could you make it safe? with a great answer to that, we'd definitely be open

What are the differences between the OpenAI fund and YC?

we will write you a check, give you preview access to new models, and try to help with ML issues and strategy.YC is YC.you could happily do both!

nothing to announce there yet, but the latest models available in our API are pretty good! check them out and let us know what you think.

Yeah they are amazing, hard to imagine they could be improved a lot

Yes, question! I How is your program related to YC. Are you a separate entity? Are you still active advising YC founders?

not related to YC. YC is the best accelerator in the world and i expect it to remain that way for a long time; this is very specifically to help startups rapidly come up to speed on what's about to happen with AI.

I see. Thanks for clarifying. I definitely agree about YC being the best! Do you accept mentees?Additionally, i love where your AI project is going. The future is brilliant! All the best! 😀

@sama can one attend both? You overlap by merely a couple of weeks.

Does your program have any interest in sponsoring companies that are not doing machine learning using deep learning?

focused on DL but curious to hear what you have in mind

My company <a href='https://twitter.com/symbolica_ai'>@symbolica_ai</a> is building a fully unsupervised end to end symbolic machine learning pipeline. We can currently hit SOTA on some basic CV benchmarks without any labels, loss function, or gradient descent. You feed data in and get features out. Working on language soon.

What insights come from the 60-sec videos that offer real predictive value for which companies are likely to succeed?If the enabling tech becomes commoditized, does it come down to who can identify transformative applications?How heavily would you weigh actual implementation?

clear, concise storytelling is a critical skill for founders, essential for recruiting, selling, fundraising, etc

This looks amazing. I'm not sure I understand the time commitment of 'approximately 6 hours per week', is that in training/meetings of some sort, or maybe a typo?

will vary by week but it's meant to be intensive. but may not be a fit for everyone.

(perhaps a stupid question but you said Anything!)Should the form on <a href='http://openai.fund'>http://openai.fund</a> send out confirmation emails upon submission?Didn't receive anything, would hate to miss an such an opportunity due to a technical issue.

check spam. confirming we received yours!

Could be my new .video TLD or mx records, thanks for taking a look. 🫡

What are common characteristics of problems that are best suited to be solved by AI?

will change over time, but today models work well in areas where there is a lot of repetitive or drudge work, or where high info load or info complexity lead to bad user experience with current tools. also enhancing human creativity.

We are not allowed to use openai products from Morocco yet.

morocco should be live! shoot us a message if you're having trouble

How will you judge pre-idea founding teams? Is it going to affect the chances of getting accepted?

large vision, clear thinking, strong foundersbeing pre-idea does not affect chances

What are you preferences? Companies using your tools (Jasper-like) or companies working on their own AI tech unrelated to current OpenAI offering?

optimize for what you need to build a great product and a large company

Hi! @sama and @bradlightcap. With respect to the application, would it help to provide a clear idea of how the start-up intends to use OpenAI computing libraries?

convince us you have a shot at building a $1T company, nothing else matters

when would you allow embedding creations using longer documents? This is gpt question.

actively working on this for future models

just applied. can you please see you got an application fro <a href='http://iWish.ai'>http://iWish.ai</a>? We are a paying OpenAI customer too.

Is previous experience as a founder required to be considered?

How much equity will teams have to give up?

10%. however, we are still iterating on this and open to experimentation.

now that I have this hot mic I’m at loss. Let me consult gpt3 and get back to you

We’re the research team working on Meta AI’s Universal Speech Translator project — Ask us anything!

Sam Altman and Brad Lightcap answer questions about AI and their new AI Fund