Dayana

 Start by saying no to everything. I say no to everything. I don't have a calendar.

 So when people say, how about such and such time? I'm like, hmm, well, 

 would have to either set an alarm for it or I would have to remember it.

 badly want to do it, I just can't even do it.

 So the maximum number of meetings I can have per day is one, 

 because that's all I can remember or all I'm going to set an alarm for.

 Tom, you bring up a very good example of how it's okay to constrain your 

 own freedom as long as you understand what contracts you're getting into.

 Hey, I'm Sejal, one of the engineers working on AirChat.

 I'm the only major on the planet, they can't do much about it.

 I just want the trash, made of this real, real, real.

 All my pimples, nose, falling off, just dripping back.

 Can't slow it, get the tempo, just a couple of kicks.

 Bitches tight, make them tight, mama said I'm kicking.

 Bitches talking, can't you make it quick? Mama said I'm kicking, bitches talking.

 living in a dorm building with like 20 floors and there would be lots of fire

 alarms you can imagine college kids, college kids drunk, you know drunk lots of

 false alarms and we'd have to go go downstairs on the stairs and then wait

 outside until it was you know all clear 

 and then go back in and I still remember

 the lines that would be there for the elevators and there would be kids lined

 up waiting like a half hour to take the elevators up just cursing and bitching

 the run up the stairs and get to my place

 how sad it was that you got these 18, 19,

 like a few flat of stairs or five flat of

 stairs and they'll just wait in line for the elevator and I never forgot that

 the airport or wherever if I ever have a

 choice between an escalator or stairs or an elevator and I'm not in a rush I

 always take the stairs heck a lot of times the stairs are faster because

 you're the only one and it's less about the stairs it's more for mindset thing

 our bodies are meant to flourish you know bodies are meant to move and express

 themselves to movement and it's great for our health so use this mindset and

 move the body whenever given the option to so wherever you are given the choice

 between stairs and escalators always take the stairs

 One is try to make your tweets timeless so they don't expire.

 A second one is make each tweet stand alone so it can be understood and

 re-communicated as a single standalone capsule and doesn't depend on all the others.

 A third is don't be afraid to say controversial things.

 A fourth is don't be afraid to mix disciplines like science and spirituality and

 wealth creation or business. In other words, don't care 

 about what the audience thinks and don't stay in your lane.

 But ultimately it was really about just creative freedom.

 I didn't think that, oh boy, I crossed the threshold of 30 tweets 

 and so now Twitter is going to collapse my tweet storm, which it did.

 I didn't think, oh, I better make sure I don't 

 offend people or go for this audience or that audience.

 It was really just written in an hour on my iPad in bed 

 the middle of the night or the early hours in the morning.

 And it was just I had a bunch of tweets floating around, just kind of 

 put them together and then I corrected some errors and spelling mistakes.

 In fact, even as I was tweeting it, I was 

 oh shoot, I got the grammar wrong on that one.

 So I would go back and I would delete a tweet and I would have to 

 a bunch of tweets and then rewrite the whole thread from that point on.

 But it was as close to spontaneous as you can get for a Twitter thread.

 I'm going to stop treating my time as my most valuable resource

 and I'm going to start treating my attention as my most valuable resource

 I think proof of the value of your attention as the most precious, most valuable asset,

 currency or belonging that you have is number one, the battle for our attention that is

 displayed every day with ads, new buy, buy, buy and desires created for us artificially.

 And number two, the fact that simply with your attention you can create worlds.

 What I mean by this is that if you were left 

 a desert island with nothing else but infinite

 amounts of time, but you were deprived of your attention, you would be poor.

 But if you were in a place where you only 

 had a few seconds, but you had infinite access

 to your attention, you would be a rich individual.

 Some notes from OpenAI's head of GTM and Solution Engineering at a recent AI day,

 which has not been recorded or tweeted or published anyway.

 So there's a lot of conversation about the use of the OpenAI internal workflows,

 building new products and services for their customers,

 and how they're integrating into existing products,

 highlighting just a couple of things that were talked about, including

 three CEO, obviously looking for guidance, whether you're a startup or a larger,

 running a larger business. What you should think 

 about in your business that your vendors cannot build,

 should not build, or if your competitors did build, you would lose.

 What are the business internal and external experiences that are most powerful to use OpenAI?

 There's a lot of conversation about what the art of the possible now,

 asking a banking app, how am I doing this 

 versus looking at a bunch of charts and numbers,

 planning a trip, you really like the beach, I don't really care where I go,

 I want to spend this much money, throwing out your product strategy,

 reimagining the things you promised to your customers a long time ago,

 and if you could inject this magic into your experiences that could have shown up

 on every product roadmap, what would that be?

 So, you know, looking at most CEOs don't actually want to tell their head of product

 to show out to years of development, and then we talked a bit about ML teams

 and what are the biggest blockers of all.

 It's a perfect inverse correlation of adoption of OpenAI

 to the size and influence of the enterprise ML team, which is pretty true.

 Really important to call out what the ML teams have been promising their CEOs.

 No one really lays off the ML teams, it's very 

 expensive and sends a negative signal to the market.

 Looking now at the future of what the ML teams will now have to answer,

 that OpenAI models are much better than the thing that they were building.

 The future of product will be better, the models will become more integrated,

 ML engineers job now may be to help find alpha in those experiences and fine-tune them,

 Obviously there's edge cases like robotics companies or Tesla

 their own ML team and invest in the GPUs, etc.

 But if you're not a domain expert and it's not an edge case, then probably not.

 Anyway, some other key interesting items. What 

 are OpenAI building and where does it go from here?

 We all know about the multimodal future, building models that can do everything a human can.

 So everything will come together in a single model, text, audio, speech, images and video,

 continuing to invest in their crown draw, which is actually language.

 Chat, GPT-5 will be even better than GPT-4.

 Obviously the probably industry may actually look like the amount of competition will shrink.

 So the world will look similar to what tech companies are.

 No one really comes out with a start-up to say,

 hey, we want to build a cloud computing company now.

 Or a lot of VCs won't give money to that idea.

 Looking at humanizing OpenAI and influencing the way the world thinks,

 so the sooner the enterprise adopts it, the sooner the users adopt it,

 the sooner the value is shown, the sooner the public policy will catch up.

 But perhaps this may or may not come to pass.

 Some audience questions involved the proprietary enterprise data,

 unless it's the open AI, open source data issue that has been on these chits,

 and is there a way to bring these together and ship product?

 And the response was they're creating a space where the 

 enterprise data can live in a single tenant environment,

 and you can retrieve that data and print it at 

 chat GPT for the internal users of that product,

 but avoided some of the other questions around the open source data,

 What are the interesting items? Oh yeah, the training of the models on public data sources.

 One of the biggest problems is that GPT is still trained on 2021 data.

 Much of the time at OpenAI is spent on alignment, product policy,

 figuring out what they should reuse and how to answer questions carefully.

 Just remember when you hire employees that you're also trained on the internet

 and they are also capable of doing and saying weird things.

 What else? Some other points about hallucinations and the plan to tackle this.

 It's one of the most important problems that OpenAI are focused on.

 Next few weeks there will be some developments that they're releasing,

 because the biggest limitation on the models right now is, as we mentioned before,

 trained on 2021 data, and in future they're going to site sources, so that's interesting.

 Where is OpenAI getting rid of their ML team?

 No, that's the answer. Noting that they only have 375 

 employees at OpenAI versus Alexa, which has 10,000 employees.

 The research team, where GPT-3, 3.5 for 4, are looking at optimizing it, making it faster.

 For example, the Turbo version, if anyone's used to it, is faster than the Primus version.

 The optimizations are making it obviously run on less compute,

 and they're confident that they're going to continue to decrease the costs.

 As you've seen, GPT-4 costs have come down a lot in the future.

 It's a quick summary of some of the conversations to share with you on this chip.

 yet, but I did fool around with it tonight

 for answering questions that I was genuinely interested in, and I found it basically worthless.

 And the reason is because it always gave me 

 the mid-wit answer, and I don't want the mid-wit

 answer. The smartest people, the most creative 

 people, the best people are now available to me on

 Twitter and YouTube and online and blogs and 

 essays, and I can just seek out the best sources.

 epistemology when David Deutsch and Karl Popper are there.

 I don't need the mid-wit version. I don't need the mid-wit 

 version on diet when some of the best nutritionists and trainers

 on the planet are available. I don't need the mid-wit take 

 on geopolitics when some of the best historians on the planet

 are available. So the problem in all of these 

 is I think chat GPT is replacing bullchit make-work

 that should have never existed in the first place.

 Biurequratic checklists, high school essays, tests, people basically putting up fake tests

 for other people. It's not actually solving new 

 problems, and it's not giving me the best answers, and

 what I wanted the best answers, and they're already available.

 So on the coding front, it looks real from afar.

 On the self-driving front, these things look real.

 But when it comes to just like generating essays and 

 text, I think it's a novelty. I just don't see the value.

 One other thing I'd like to point out is your comment about the linearity of podcasting.

 Our entire system design is centered around the concept of taking chits,

 chits are our atomic unit, chits, sort of a time 

 series of chits, equal a chat, right, a conversation,

 and unlocking the value of a particular chit.

 How do we do that? We do that through stitching and through rechitting.

 And the concept here, again, is to take that one 

 piece of the conversation that matters most to you,

 that's most thought-provoking or potentially is most combative or where the knowledge is the densest,

 and to be able to fork that and turn that into another conversation.

 And that is really the power of this system. It is not linear in any respect.

 If you find something interesting, you can grab it.

 You can pull it into your chat. You can pull it into a private chat.

 You can push it to the feed, or you can start an entire 

 new conversation around it. That's the power of AirChat.

 why you don't want to import a social graph.

 or 100 people that I've imported, that's enough.

 Now show me something good. We can't show you something good

 because most of those people are not in their chat.

 And even if they were, most of them are typers, not talkers.

 So it would create this false sense of job done.

 Whereas now when you come in, if you don't see much,

 pretending we've done the job. Instead, we're failing you very explicitly saying,

 we haven't yet found people for you to talk to,

 I love this so much. We were recently talking about that 

 lot of people these days are passive consumers of content

 There's a tendency to Watch more than like or 

 comment and I think you what you're saying here

 Is that built into our chat is this invitation to be an active participant?

 to stitch and build on that chit and make it a different conversation or

 right now and also to to really go and seek

 People that you want to know more about 