OpenAI CPO Reveals Coding Will Be Automated THIS YEAR, Future Jobs, 2025 AI Predictions & More!

all right we just wrapped up a conversation with Kevin from open AI uh I as the resident dodo uh got most of it but I'm pretty sure I missed I missed some I'm pretty sure warun grasped a lot more warun what do you think of it and what stood out for you so Kevin is the CPO of openi he's the one that led deep research and basically all their products dude the thing that stood out for me is somebody from anthropic had told me that code is going to be totally automated by 2027 Kevin's like

hold up it's probably going to be done this year or the next so so so very interesting insights how they think about product where's the commoditization he actually argues against commoditization he says sure they're all commoditized but there is gaps in capabilities and those gaps keep changing over time I mean there's so much interesting to take away if you're a user there so much to take away about timelines and predictions on what will happen over the next year the year after that and so on and so forth there are a bunch of business ideas right

t i mind out some business ideas that Kevin thinks will work in this world of AI I mean this is a very fascinating conversation I don't want to spoil it for you so please watch it and yeah can and yeah sub subscribe subscribe it's cuz you guys watch And subscribe that we get to have these fun conversations and uh roll the intro today's episode waron is going to be a going to be a special one CU people get to see uh people get to see uh somebody outside of the same two mugs that they see

every week on the show uh we have Kevin well uh from open joining us uh Kevin for folks for folks who might not know could you tell us uh could you tell us a little bit about uh about uh uh what you do at open and why are you so special why am I so special uh well I I have the privilege to be the the chief product officer at openai which is probably the most fun I've ever had in a job uh you know most I I've been fortunate to work a bunch of places

in my career I was an engineer uh in the early days of Twitter when it was about 40 people up through when it was 4,000 or so um spent most of my time there as head of product was head of product at Instagram was co-creator of Libra the crypto project out of meta um spent a bunch of years building satellites at Planet which is also an incredible experience but you know every every other place that I've been outside of open aai you kind of have a a basic sense of what computers can do you know

and you're Building Products you're thinking about who are you building for what problems you're trying to solve but technology is technology you know databases improveed but not that fast like your database this year is you know 5% faster than your database last year but it kind of does the same things the the crazy thing about open AI is that like every two months we make computers do something computers have never been able to do before in the history of the world and so like your conception of what technology can do changes and it means that

you have to almost like rethink your product every two or three months which is I mean it's intense but it's super fun all right so guys the reason why Kevin is special is because his resume is bigger than my Wikipedia page that's the main reason Kevin Kevin thank you so much for joining us yeah uh Von I know you have a you have a ton of questions so why why don't why don't you start awesome Kevin Kevin thanks thanks for joining us can you tell us um you know just how it's been at openi how

has it been sort of you know working on product because openi is two companies right you're a research firm and at the same time you're a product company and a lot of people have spoken about you know the bottles they going to keep getting better and research you know there's great research happening here there's great research happening in other companies but open eyes has also captured the world's imagination as a product chat GPT just chat GPT it is now a verb I use it as a verb so what's it like sort of you know what

what does your day-to-day look like yeah it's been awesome uh I mean when you if you go back open AI started what like 10 years ago now and was originally a research company it was started with you know a handful of at the time crazy people who said you know what we can build AGI we can build artificial general intelligence and at the time nobody really believed them they were kind of off off you know in a corner but they were right they they figured something out they started to build some amazing uh demos you

know robot hands that could solve Rubik's Cubes things like that and then they happened on these scaling laws that showed that the more compute the more data you appli these models started getting really good and uh and as you said they open AI was very quickly a worldclass research company but now as with chat GPT with our Enterprise Products our API products other things that we're thinking about we can't just be a world-class research company we have to be a world-class research company and a world-class Product Company and actually increasingly I don't think we can

be sort of these two things separately we have to work really well together because the like I said before the every two months your conception of what computers can do changes and if you're operating as two separate companies then the best thing you can do is sort of throw you know some improvements over the wall at each other and that's not going to be how we build the best products the best products and we're seeing this with things like deep research where it was really an integrated product research enge design team that they happen when

you can bring together what what people want and you know a sense of the problems you're trying to solve from the product and engine enging side with the incredible breakthroughs that are happening uh with the models and their capabilities and when you bring those two together early on then you get magic if you're throwing things over the wall at each other you know you could build cool stuff but I don't think you're going to build the best stuff and the market is so competitive we have you know great uh great companies building great models right

left and center the only way that we're going to win is if we really uh operate as one unit and that's increasingly how we're setting ourselves up so your day-to-day is mostly thinking about where the puck is going with research and then saying well here's where product needs to accelerate to meet it and here's what the entire whole thing looks like in 6 months from now or a year from now well it's both you know um because it's one of the one of the interesting things about working with models is the models these days are

so incredible they're not really intelligence limited they're they're in some sense like teach limited like you you can teach them to do anything you want them to do but you need to have the data the RL the environments that teach it to do something great and so there's actually an increasing amount of product work that goes into what are your evals and how are you what are you going to teach the model and there's Nuance to it so you know if you teach the model on again you you train it against a bunch of competitive

coding contests then it's going to get really good at competitive coding but competitive coding is not exactly the same thing as like real life you know coding a frontend app um and so you really need to get into what problems are you solving very specifically and make sure that you're teaching the model the right thing so that's a part where like actually the product goes backwards into research and then like you said there's all sorts of use cases where the the research goes forward into product there's some new capability and we're like okay how can

we how can we build something awesome with this uh and it's why you need both groups working together uh you get the best of both worlds and you build something amazing I see they spend so much time between research on product there's what ends up happening is you have very little time to end up naming the products that so Kevin just just just for a little context like vun varun's an engineer by by you know he he trained to be an engineer and I I'm also a comedian on the side I just want to give

context that I'm not just a random person showing up to make jokes all right go on no know it's uh we deserve all the the crap we get from people about our naming it's it's absolutely horrendous um but but we know it and we lean into it um you know we do we do Reddit amas every once in a while and I was doing one with Sam and someone's question was can you guys improve your naming and we were like oh man how do we respond to this you know we're sitting in a room going

back and forth and in the end I just responded with two letters no like we're trying to make AGI happen and the biggest consumer feedback is the names are kind of confusing you should just name everything agi1 AGI 2 ai3 it's just easier yeah I I'm I'm excited though uh you know Sam talked about this a little bit as we bring the models Back Together We we've had this uh we've had this bifurcation where historically we had all the GPT models right you're doing you have GPT three three and a half four now now 4.5

and these models are fed by bigger and bigger pre-training runs and then we we had this breakthrough last year with reasoning and because the models were incredible you 01 preview 01 you know 03 that's coming soon can do unbelievable things can I mean the benchmarks they're off the charts they can do things that you know by all the scaling laws you would have had to pre-rain a model that was 100,000 times the size to get there um so they're incredible in what they can do but at the time especially they didn't have all the bells

and whistles that our GPT models did they they couldn't search the web they couldn't you know handle file uploads Etc they're just really good at reasoning um and because of that we gave them a new name and we sort of branched the series uh because we couldn't really combine them they didn't have the same set of capabilities but there were times when you really wanted to use one and times when you really wanted to use the other and you know I think that was the right decision even though it was confusing because it helped help

us get these things out into the world and now people are making advances with them they're doing amazing things with the models whether it's like coding or Frontier science or other things um and so you know in in in most things we optimize for velocity we optimize for putting these new tools in the hands of people so they can do more cool things faster but we bifurcated and now as we've sort of gotten familiarity with these reasoning models we're teaching them to use all the tools that the GPT series can do so they can handle

file uploads and they can search the web and they can use python as they reason which is kind of crazy by the way like as their reasoning to solve a problem they can write code to help them like understand it better um super cool when it happens but uh so now the O Series and the GPT series are starting to have the same sort of capabilities and it gives us an opportunity to bring them back together and that's what we're planning to do with gp5 so we we are working on simplifying even though we are

ethically terrible at naming do you have a do you have a this is a hard question but do you have a timeline for GPD 5 I just I had to ask um it it's I mean soon enough that we're talking about it right so there's a little bit of uncertainty because we do want to to actually simplify as we do this which means all of the tools and everything that that people are used to need to work really well so we are having to teach the reasoning models a bunch of new skills and it's research

so you you know some things have error bars on them so I won't give you a time but it's soon enough we're we're like we're out talking about it we're very serious about it people are working on it as I speak Kevin uh you know I I I'm I'm not an engineer right I'm I'm I'm I'm a I'm a I'm a user of the products and I I I love them by the way I I I think you guys are doing a great job uh you know as someone who's making the products right like as

time has gone by you know we've seen competitors Rising right and people almost say that okay the the models are could get commoditized so the the actual product in the use cases become that much more important and now you guys also must have noticed that okay everyone's going to sort of like if you know if open AI leads the way then the others are sort of going to come come out with similar products so as as someone who's making the products do you guys are you guys now conscious of you know what the competition is

doing while you're making the product and how do you guys think about you know continuing to maintain this sort of uh first mover advantage and a lead uh as someone who works on the products like how do you guys think about that it's a great question and you said I'll sort of disagree slightly with the framing of the question a little bit which is I actually I actually don't think that models are becoming commoditized I think they're moving really quickly and they're getting better fast uh I mean honestly the rate at which at Which models

are gaining intelligence is unbelievable Sam was talking um a couple weeks ago and said that if you take constant intelligence take take a certain level of intelligence as your benchmark the cost of that intelligence is coming down by a factor of 10 every year you think about Mo's law yeah which is you know basically defined our age for the last 60 years it's doubling of the number of chips on of transistors on a chip every 18 months and so that's twice every 18 months this is a factor of 10 every 12 months so I mean

massively steeper curve right and of course it's not intelligence isn't saying constant it's it's dramatically increasing at the same time as costs are dropping I mean just like a series of trends that are unbelievable um but uh I I don't think that these models are commoditizing they're just all like because we're on such a steep trajectory people catch the lead fast but that doesn't mean that the lead isn't valuable so I think even a three to six month lead is super important because it means that you're the first to new capabilities which means you can

be the first to launch new products that meet those capabilities and since these capabilities are new to the world these products have never existed before things like deep research right and you're the first to do those you are sort of leading the way and everyone else is following and I think that's a valuable place to be so um so you know we're we're going to do our absolute best to to keep the lead that we have other people do great research and they're aspects where where other labs are in the lead and like more power

to them but uh it's we are hyper competitive and we definitely watch what other folks do um but we want to lead the way we want to we want to like our mission is to uh to ensure that AGI benefits all of humanity we do that by trying to put AGI or put AI in everybody's hands you know both in the first-party products that we build with chat GPT and others and by making an API that you know three million developers are using on a regular basis uh to embed AI in like every tool and

C and company and product out there so like we're going to do everything we possibly can to bring AI to as many people as we possibly can yeah I have a question here right which is actually it's more like a personal observation which is look I was about to turn I'm just being frank I was about to turn off my chat GPT subscription and move to Claude for a while what yeah what this is this is this is this is like a short while ago but then deep research came out and I was I was

on the regular regular plan uh the plus plan and then deep research came out and I heard it was a Pro Plan and I was like you know what it's it's probably not going to be that good but then I saw a lot of tweets about it so I saw one tweet people saying this is phenomenal then I saw another one I was getting I was starting to get four more right and I saw third one and I saw fifth one I said you know Screw let me let me let me go get uh o1

Pro right let me get the Pro Plan and to be honest it's been a long time since I had the kind of holy moment that I had with deep research like deep research is great it's like you know perplexity has a deep research your you know Gro has a deep research nothing comes close to the Deep research you have how much of that is you know the underlying models how much of that is your work on the product Because deep research is phenomenal I use it every day um you know you you were resource constraint

at some point right you can only do a certain number of searches I I made another account and upgraded so I can do more I use it for the simplest of quaries because you know it's going all through Reddit and like a bunch of blog posts that nobody's ever seen in the the world collecting all that it hallucinates very very little it's it's a phenomenal in my opinion that's it feels like AGI because because AI always felt like I was getting content out but it was generic in nature but with deep research it's now filled

with insights nuggets of insights like if you ask it how many people on average does it take to make a AAA game now it the earlier versions of AI would be very flaky about it but now I get specific examples saying this one did this many this one did this many and it's just so useful how much of that was you in product and how much of that is the research underneath it's the longest compliment Kevin's received by the way no no you you can keep going though you can just keep talking about deep research

you're doing a great job um yeah it's I'll be honest actually I think deep research is the best product that we've launched since cat GPT itself um it's it's just magical and it's one of the first internally we talk about these like feel the AGI moments when you just you just get goosebumps and you start to you know you you intellectually you you see the the trends in in Ai and models increasing in intelligence and all these things and it's sometimes kind of hard to actually reason out like you know what is this going to

feel like when we really have AGI around us at all times but every once in a while you get a hint of it and deep research was one for me because you just I mean it's so many of the things that I use chat GPT for it's saving me like 5 10 minutes at a time and that's awesome right I'll take all the time I can get I've got like job three kids you know family all these things going on um but in a lot of them there there are things that I could have done

without cat GPT it's just cat gbt makes them better faster Etc deep research was one of the first places where it's just I could not have done these things you know I was like I'm a physics nerd so uh I I was like trying to learn about uh muon colliders and and I like had it go do this big thing for me and in 20 minutes it came back with like a 15-page report on neon colliders there's just I I just wouldn't have done that you know ditto like doing some medical research for some stuff

with my son I don't have the capability to do that kind of research um and it went off and did it for me and provided me a bunch of peace of mind um and so there's just it wasn't just saving a little bit of time it was fundamentally doing something for me that I could not have done myself and that's that's it's it's eye opening right and back to your question I think uh I mean I think there's a credit starts with the with the research team they did an unbelievable job but it it is

one of the examples I was telling you where we have a research team and a product and engineering and design team all working together from the beginning making sure that the the problems we're trying to solve for people match up with the ways that we're like Ealing our model and as we're training to get the model to like get better at certain skills those are directly tied to the product we're trying to create and it really is the magic of bringing these teams together so um I think it's an example of something we couldn't have

created if we were operating you know as a research team over here and a product team over here interesting and and one follow-up question on deep research and and I guess not on deep research but you know for example I've tried 4.5 recently and 4.5 has this very distinct big model feel like I don't know how to how to explain it but it's almost like when I use a smaller model like let's say or 3 minut I can tell it's it's it's taking me too literally I don't I don't know how to explain it but

it's just one of those Vibe things I get with a model but like 4.5 writes really well it's able to pick up Nuance in what I've said it's able to pick up edge cases in what I've said and I from my understanding 4.5 doesn't have any reasoning this is just a lot larger of a pre-training set are you seeing are you are you now going to see with this level of new pre-training with this level new cus combined with reasoning we're going to have something new that the world's never seen is that do do you

expect a huge capabilities jump or does reasoning not matter if the model is already that large no no I think reasoning definitely matters and and you can see it in the benchmarks right like if you're looking at some of the more academic sciency benchmarks whether it's like GP QA or Frontier math or rcgi these kinds of things 4.5 doesn't hold a candle to like 01 let alone 03 are are you know forthcoming Frontier reasoning model but there's also like you said there are these sort of softer evals these things that are squishier and more human

and like softer that 4.5 is just unbelievable at in ways that are very hard to quantify um but but you can certainly see it in like human feedback you can see it in you know AB tests where you compare 4.5 next to other models and ask people which one they prefer and you see a massive preference for 4.5 so I I think it goes to the fact that there really are there are two ways that we know of today to scale the intelligence of models you can do a bigger and bigger pre-train and you can

do more and more RL on top of it to teach it to reason and it the answer in the end isn't one or the other it's to continue to do both so these bigger and bigger pre-trained they get the soft skills they have better World Knowledge they're more interesting to talk to they feel somehow more um you know human or Alive isn't exactly the word I'm looking for but they feel more real to you to me do you have a theory white um I think that I think you can encode a lot more subtlety at

the end of the day in a bigger model the human world is is full of Dimension and subtlety and nuance and when you take a like if think about going the opposite direction if you take like a 40 model and you distill it down into say 40 mini you get a model that depending on how you distill it can keep most of the like the skills on a particular Dimension that you care about so if if you want to build a really good small coding model you can totally do it you can take 40 distill

it um or take 01 and distill it and and produce a really good coding model and it will you know maybe not exactly be at the same benchmarks on coding but but like not very far and it will be a much smaller model but if you then try and talk to it it's not a super fun model to talk to it sort of lost its personality in some ways yeah and like if it's a coding model who cares that's not why you're using it so I mean small models are awesome in a lot of ways

but there's some there's some sense in like which as dimensionality collapses you losing like you're losing subtlety and nuance and and there's something about that that is that makes things very like Pleasant to talk to and I think the same happens when you're going the opposite way and expanding it there's just a lot more room if you will in the in the added dimensionality to like encode more Nuance um you know that's a little bit viby but I think there's there's there's some science behind it as well uh but it seems to be true in

everything that we see and 4.5 is just I mean if you're if you're asking a model to represent a particular voice if you're asking a model for relationship advice if you're using it for like more of the soft human things 4.5 is the best thing out there by far and you you you won't know it until you try it and then you you try it and you're like oh I totally get it right I mean it sounds like that was your experience yeah it's a better writer it just there's something about the writing on 4.5

that just just feels human it's it's it's it just feels human and it's almost like my two or 3 years ago I would have said maybe reasoning is the better way to write better right maybe you're logically you're saying this happened then this happened and this happened that should help you write better but it almost feels like just you know this subtlety like you said is is what leads to an improvement writing skills which I wouldn't have expected so I've I've done a course correction in my head right but I have a question here and

I want to ask questions on specific skills right I want to start with let's say coding because you know three years ago if you said you know coding is going to be automated nobody would believe you they'll be like you know the gpd2 can't do you know it's not even competent this with GB3 sure it outputs some code but then you know this you want to make a small button fine it can do that two years ago when we started seeing I mean a year and a half ago let's say when we started gbd when

we started seeing gbd4 it started getting used every day right and now it's gotten very good to the point where comparative coding has gotten very good but also you have this specific I'm I'm let's say I'm trying to build front end for something or I'm trying to build the infra for something on the back end it's able to do a much better job of it now and I was talking to somebody from anthropic right a few weeks ago and I said does anthropic have a timeline for when code gets like 99% automated I'm talking about

actual functional code that you write for front end and back end uh they said yeah 2027 right do you have timelines on that do you have a theory on where code goes because you write this on a daily basis and you work on re you work with the research teams and you build product do you have a timeline in mind or at least a thesis on how this plays out I I mean at the rate that we're going I would be surprised if it's 2027 I think it's going to be sooner and topic said 2027

we're gonna say 2026 I just I look uh so when we launched GPT 40 like you said very good coding model like back end of GitHub co-pilot and all these other things people are using it at scale all over the world but it doesn't really compare with like when we launched 01 preview it was a much better coding model because reasoning matters when you're writing code right whether you're anything hard you're doing if you're doing a crossword puzzle or a Sudoku you know you're writing code you want some level of reasoning some ability to like

break down a problem into smaller problems form hypotheses validate or refute those hypotheses and like that's what reasoning is doing so 01 preview was I think the best like the millionth best uh competitive programmer in the world you know when you when you play it against a bunch of you sort of replay a bunch of programming competitions it would come in at about a millionth which you know doesn't sound that great but there are I don't know 30 40 million programmers in the world so you're like top two three% 01 that was 01 preview 01

when it launched our first like real launch of a coding model was something like the thousandth best engineer in the world uh at competitive coding contest 03 which is coming soon according to the same benchmarks is the 175th best competitive coder in the world and as we are starting to train you know the successor models they're already better and so I I think this is the year that at least by the competitive coding Benchmark this is the year that that AI becomes better than humans at competitive code forever right in the same way that computers

passed humans at like multiplying 70 years ago and and they passed AI passed humans at chess like 15 years ago this is the year that AI gets better than humans at programming like forever right and there's no there's no going back um and and you know we're putting a lot of focus into this anthropics putting a lot of focus into this Google's putting a lot of focus into this so it of of all the things this is going to move really fast um and I think the world's going to be different because of it I

think a lot better you I mean imagine all the things that you can do if you don't need to be an engineer to create software like software if you could write software I mean the reason that it's actually AI passing humans at software is way more important than AI passing humans at like chess is that with software you can create anything you want almost and so what a democratizing effect this can have on the world if everybody can create software I was talking to some people um they were relaying uh you know back in the

covid days they were trying to create for their local like City they were trying to create a website that allowed them to like track various things related to covid data and there were no Engineers free and they didn't have the skills themselves and they just couldn't do it and as a result like they just didn't have that information I mean you could do that today there would be no problem today for any of the top models let alone when these models are able to produce sort of arbitrary amounts of of great software so I'm excited

for this future it's going to be here very soon Kevin but do you really think Engineers having more free time is good for the world cuz I have I have many engineering friends and I don't need that competitive energy to be channeled into things that I like to do I mean you know I don't I don't know about you I spend some like non-trivial fraction of my day doing stuff that I wish I did not have to do you know I'm like whether it's work stuff where it's kind of like busy worksh or I'm like

filling out forms for my kids soccer team or something you know like there's a big chunk of my day that I just wish could be automated right and it should be and similarly I I was an engineer for a lot of years and you you get to a point some problems are really hard and you need to like put all of your time and attention into them for a long time other times you're like okay I know how this is going to work I just have a lot of typing ahead of me and it's not

super differentiated I just have a bunch of typing the R smiling You Know It uh uh and like you know you that that that stuff should be automated and it will be Kevin as as a content creator right like there's an ongoing discussion around just how much AI generated content is going to is going to come online like Von varun's Instagram is basically a lie it's not vun at all like 9 95% of his videos are not him uh and it's someone clicking three buttons to make it seem like it's V like where do you

think this is headed like it feels like authenticity is diminishing at a rapid pace and like do you guys had open ey is this is this a concern to you guys is this is this something that you guys think about actively about you know what happens to the internet when everything can be generated automatically yeah so I'll give you my personal view on this which is um I I think that humans with most like human creation there is a sense of like a proof of work that's valued um so I if it becomes super easy

and I think it will to for to create a bunch of like AI based content you can already do with images all the time right it it it's not there's people value something that they know took a lot of time and energy to create they don't necessarily value the the the thing that you can make in five seconds and so you know as it relates to like take Sora right you can't go to Sora today sora's our our video model for folks that haven't heard of it you can't go to Sora today and be like

Sora make me a movie right it just doesn't work yeah um but it can if you give it like really thoughtful detailed prompts you can have it you can make it create amazing amazing things and we have a we have a Creator internally um who's like worked in the industry before he can do things with Sora that I can't believe like they're well beyond anything I can do we have the same tool but he can still do amazing things with it I was talking to a director the other day uh who's looking at using Sora

um and he was saying look before he was he like does sci-fi stuff and he's like okay so imagine you've got this sci-fi scene you've got like you know a plane in outer space kind of like zooming into a planet like maybe some sort of Death Star type planet and then you need to cut to a scene where the the the thing is like zooming along the surface of this like technological Planet how do you make that cut from one thing to the other there's a bunch of different ways you could do it and what

he was saying was look today I would go to a special effects studio and I would give them like $100,000 and they would make me two different versions of the cutscene and it would take them a month and I would select from one of those two because that's all I really have with Sora he can make in the space of an afternoon he can make like 40 different versions of that same cutscene exploring a whole bunch of different Avenues just like letting his creativity run wild and you know sort of collaborating with the AI to

do even more and then at the end of the day he might still go to that same like special effect Studio to do the one that he really likes but he's going there with a whole bunch of different ideas in his head he's chosen one out of 50 and then he ends up with something that he likes way better because he collaborated with the AI like I think there's a world like that it doesn't mean that you just like snap your fingers and everyone creates amazing stuff it's going to take a huge amount of work

to do great work you know the bar probably stays the same for how much effort and expertise it takes to do great work it's just the output with AI can be that much better and maybe that much more accessible to the rest of the world and faster I think if if you're if you're able to see outcomes you're not rate limited on somebody else so you you're you're sort of spending the same number of man hours but you're saying well if a VFX studio is going to take a month to do this I get the

outputs immediately but I was anyway not putting in the work because I was sort of you know Outsourcing it to them I was talking to you know another person from the industry and they said that well the future of creative work or coding is sort of going to turn into management because this ability that we have now with AI was always available to people with capital it's called hiring people right uh I don't write as much nearly as much code as I used to write maybe a year ago because now I can hire a bunch

of people and say Hey I want to try this experiment and it's probably going to take them a week to try it out because they're using AI do you do do you if you were a creative either a coder or let's say if you were somebody who's writing content or making ad scripts How would you pivot to this new world because there's lots of people in India that and across the world I'm sure that value intelligence over everything right saying that look I I built this from scratch and it's so hard to fight that because

sometimes you know we speak to people and they're like you know I wanted to build this from scratch I know the absolute nties of this and we're like good but that should reflect in the outcomes it's like but today interviews tests for do you know can you build it from scratch yeah well I think the world I mean technology is an ever increasing uh Step Up the Ladder of abstractions right we move faster because we're not having to design our own circuits and then once you have your circuits you're not having to like write machine

code and then you have assembler but you don't have to write you're not writing machine code anymore you're writing assembler and then you're writing C and then you're writing Python and and then you have web Frameworks that do things for you and you're not worried about the underlying operating system because you're living in a browser you know like the the world is a series of increasing abstractions and those abstractions help you move faster and AI is one more abstraction that helps you move faster so yeah do you do you if if you use AI a

lot to write the to write a bunch of your code and if if you're if you don't ultimately care about the code that much what you care is the output of the product do you know the internals of the code as well as if you'd done it yourself probably not yeah but you know do I know the inter my dad was an engineer uh worked on compilers at Microsoft back in the day you know do I know the internals of the thing that's compiling my code to the degree that my dad did definitely not but

does it Ma I mean I I'm also not writing like Fortran I'm you know writing python in a web browser and deploying it in real time so like these things are tradeoffs there's nothing um like you lose maybe a little bit of that but you gain speed and you gain uh the ability to bring this not just to like the subset of people that happen to know compilers and Fortran but to anybody in the world that can use chat GPT and that's amazing right it doesn't preclude you from ever going way down into the details

if you want to right there are still people that make chips and they they do an incredible service to the world but the majority of people don't need to learn that level of the stack and they can like operate faster and do more things as a result but in the shortterm if you've made your identity around intelligence and let's say coding is one good symbol of intelligence and that's getting commoditized if you were an engineer who is not at open AI maybe starting out your career how would you deal with the status hit that's incoming

because you can already signs of it on Twitter where every day there's a conversation around this there's people in denial there's people in grief all all all stages of of of of this and there also people saying well nothing's going to happen because the effort required to create an outcome is the same and that part I I agree with but you know if everyone has access to it now you're sort of competing and that's the problem with status right it's all about competition now you're competing with so many other people who can write English who

don't know as much about the the underlying stack is you but are still now in the market competing for that same job or that same title how do you deal with the status effects of this I I mean I I think so what will matter in a world where I mean first of all like Jeff Dean is a better engineer than I am and I'm willing to bet you that Jeff Dean plus AI is a better engineer than me plus AI you know so like the the expertise and the experience right just not just like

the raw intelligence but understanding what problems to solve where to focus your work uh you know where the Leverage is th those kinds of things are still going to matter um and so I don't think that like everything just completely equalizes across the board and and expertise no longer matters also though then you know you you start to think what what matters at that point and if you can create anything then it puts a huge emphasis on knowing what to create you know so who has a feel for what are the most important problems to

solve and how you would go about solving them and if more people have access to the ability to solve those problems the world is going to be a better place because more problems are going to get solved and like we will move ahead faster as a society and I like I'm excited for that it it almost it it feels like okay so the one one thought is okay High competency plus AI uh is continuing to going to have an advantage right but the other school of thought is that okay AI is also going to create

new types of jobs they almost seem kind of contradictory right which is that okay existing jobs will exist but AI will be a part of it uh Kevin I I I wonder I I don't think I've got I've heard a great answer to this which is what kind of new jobs do you see existing like is is is there anything concrete that you've been working on a product and you're like ah this is definitely going to be a thing in the next few years like a new type of job yeah I you know it's always

hard to say um I'm not sure I have the the answer that is going to make you happy here either um but I I do think that just about every job will be complemented by AI you know either you're going to be using it day in and day out to augment yourself in your job or like Veron was saying people are going to increasingly be sort of managers of these AI employees that will do a lot of the the basic work for them and can you know maybe we'll leave the the sort of so what

to humans like they'll do a lot of the what and then humans will still take responsibility for the so what in in in the same way that like you often do as a manager of people um but I don't know I I I think one thing that I'm confident in you know a lot is going to change uh over the next few years a lot is going to change one thing I'm confident in is people inherently I think most of us all of us want to be a part of something bigger than ourselves so I

don't really believe in a world where like we all get you know Universal basic income and like lie on the couch and do art I think people want to create they want to help people they want to make the world a better place and AI will change things it will help us do that in ways that maybe we can't now but I think that fundamental desire is is is like very innate to people so I don't see a world where we're all like you know relaxing and doing nothing because AI does all the work for

us I think we will still strive we will still try and leave the world a better place we're just going to have different tools to do it with okay maybe maybe not this but I'm sure while building products like you guys have uh just these many priorities you're building your own products but while working you're like Ah that's a startup uh it's it's like you guys are creating intelligence and you're like oh that that that that's got to be a company uh we of course are not going to spend time building that cuz we got

other priorities but has there been a couple of ideas in your head thinking oh someone should make that someone should use what we are building and then make that t is mining for startup ideas yeah I'm mining ideas uh you're looking for something to do after the podcast game um you know I know English I can build when this is automated we need something to do yeah yeah drop us some couple of couple of ideas Kevin ideally if they they could be worth a billion dollars that would be useful yeah you know here's so I'll

give you a general one and then I'll give you a specific one uh the general the general thing is I was saying earlier these models are not intelligence limited they're like teaching limit they're they can learn anything and it's about what you have to teach them so so I think there are a huge range of big companies waiting to be built leveraging private data right the models are trained off of the internet and public data but the majority of the world's data is not public data it's private data that's locked up behind the walls of

like Enterprises institutions governments Etc and you can use that to make the models incredible and so you know you're already seeing this in a bunch of places and legal and others where they're taking private data and using that to augment the models and make them amazing at like function specifically and like really dig in and learn legal workflows I think you can do that across a whole bunch of different things um the specific one is and this is super relevant for I mean for the whole world for India very very particularly I really want to

see I want every kid to grow up with a personalized tutor and I don't know why it doesn't exist yet today like my kids don't have one I don't know if you guys you know know anybody that has I like for some this feels like one of the most value added things that you could do to meaningfully inflect the pace of change in the world and like the the quality of life all over the world because every result I've ever seen every study I've ever seen done says that kids with uh you know normal Education

Plus personalized tutor are like standard deviations above kids that just get a normal education and then obviously there's kids all over the world that don't get what we would you know look at as a basic education even to begin with but they probably have access to like uh you know phones and chat gbt is free and like I I just want to see the world where every kid grows up with a personalized tutor that can teach them anything they want to know push them as hard as they want to put be pushed and like you

know grow as fast as their as their innate intelligence will let them grow the world will be a better place and like the AI can do it today so it it it's one of those things where it's like we're not waiting for some breakthrough four years from now like the AI is ready for it today and I I want to see somebody build this and get it to like three billion kids in on the planet that's because AI is too kind warun Indian Indian students aren't used to really kind tutors I need my teacher to

yell at me if I underperform you can get the unhinged mode I was gonna say Tom we'll train you we'll train you a model that'll yell at you hey I I have a question here right which is um let's say you know a kid has a lot of wife questions and the AI can sort of answer the why questions with you but I saw a tweet recently by somebody that said that you know unfortunately all the wise that the that the that the that it can answer for you it can also do by itself right

um is that is that a fear that that by the time the kid grows up maybe 20 25 years a lot of the education might not be useful but maybe internalizing that education helps you make better decisions like how do you think about you have three kids how do you think about how they would be schooled and what their life would look like 10 15 years later I know I'm asking you to to paint a very very Longway picture maybe it's not the open ey road map but in general where you see the world going

now that we've unlocked this I mean I think education has to change in the same way that you have calculators all of a sudden you shouldn't be teaching people how to do like super long division as a skill anymore like you just don't need that education should change and evolve with AI but um yeah there's I think there's value in being in a classroom and like in a social setting having a teacher that can guide you and there's value in having a one-on-one experience with an infinitely patient AI that can go whatever Pace you want

to go you can ask at anything you don't have to feel silly um and like it's the two of these things together I think will I mean you could imagine 16year olds today versus 16y olds in the future and they they're five grades ahead because of the the tutoring that they've gotten and if you could do that the world would just it's just a fundamentally better place like we could do so much better than the education system that we have so um this is like my personal thing I really want to see someone uh take

the World by storm with an amazing AI tutor um I I think we at open AI would go to the ends of the Earth to try and support somebody doing this at scale so um it's just like one of the most tangible ways that we can improve the world like I think I think yeah tutors could get way better but so do video games so it's like the the problems will still exist you know hey has has has voice mode picked up because I feel like I thought that when we have you know something like

the movie her uh we'll use voice mode a lot and we'll talk to the AI all the time but I don't find myself doing that I find myself texting the AI more often I don't use any voice mode on anyi which is so weird because I always thought it would be the opposite any any guesses because you you had producted openi right any guesses why it hasn't picked up or maybe you know I've got the data wrong no I think it'll get there we have a lot of folks that that are using it all the

time there's a bunch of power users for sure that are you know I know people that like walk home from work and as they're walking you know whether to their car to the bus whatever they uh they actually just talk to chat GPT the whole time and they take it through their day and like go through the meetings they had talk about the to-dos and at the end of the day they they come back with like a set of to-dos from chat GPT having like debrief their entire day so there's a bunch of really cool

use cases for voice I think it can be also way better than it is um you know it's still if you leave a gap in the conversation today voice mode will jump in because it thinks you're done talking humans don't do that you know they take cues and and in ways that the AI doesn't today and you you still have this like you start talking and then it stops and then it starts like like we we haven't you know as humans we kind of learn to talk over one another a little bit and then you

kind of learn the cues for when to back off and who continues I me know we've been doing it all day in in in this podcast um and so there are those kinds of cues like the subtleties that will make this feel really authentic like we care a lot about this I think we're going to have uh we're GNA you're going to see a lot of improvements real soon so like voice mode is amazing to me and also not quite where it needs to be and I'm very confident that we're going to get there and

that that will uh unlock just like even more right you want to be able to speak to chat GPT in every way that you interact with another human being sometimes that's video sometimes that's voice sometimes that's like typing uh but you need to be able to do all of it Kevin sometimes you look at look at the work you're doing and you think yourself maybe maybe the answer to this is a physical product like is is that something that you've thought of yeah the we're starting to look at robotics uh you know TVD whether that's

because we're going to go do a big robotics thing ourselves or we just need to have enough uh Real World experience with it um to build great like Vision models and real world understanding models because no matter what we know that robotics is going to be a big thing and we want to power it for lots of other companies doing amazing stuff with robots so like to me that's the next like first you get the digital world right and then after you have you know AGI that's that's helping us do all kinds of things in

the digital world the the next obvious place for that is Robotics and real world impact so uh you know if we want to be like the company that we want to be we've got to be able to play them both all right Kevin thank you so much for giving us your time uh I know you got to run so please do your thing and all the best and keep doing the great work that you guys continue to do and thank you for deep research yeah thank you for deep research hey we've got a lot more

coming please please keep uh you know I want all the feedback so tell me tell me when things are working well but especially tell me when things aren't working well because nothing inspires our team more than someone saying you know what I just wanted it to do this thing and it couldn't quite you know we'll turn that around like a few weeks later we'll be able to do that thing for you thanks Kevin thank you thank you awesome thanks guys bye bye bye