Sam Altman On Miyazaki’s thoughts on art, Design Jobs, Indian AI, Is Prompt Engineering A Job?

ladies and gentlemen I'm with Sam Alman<br>and Sam uh I've seen a bunch of your<br>tweets over the last few days they're<br>called posts now they're called X's now<br>i don't know what they're called anymore<br>i don't know what they're called i still<br>call them tweets you've been tweeting a<br>lot about India you've also been<br>tweeting a lot about the new image gen<br>firstly I want to say I've had early<br>access to the the photo image gen thing<br>it's been great everything from mood<br>boarding making images UI it's just been<br>phenomenal in fact it's also very smart<br>so we've been we have this very clever<br>use case where we're generating the<br>front view of a 3D model and then asking<br>it for the left side view the right side<br>view the view behind and then we're<br>throwing that into you know another<br>model that generates the mesh and then<br>reproing these textures on top and then<br>also asking it to generate height maps<br>which is just the insanity of this<br>existing where it didn't exist a couple<br>of months ago is just it it's it blows<br>my mind what's the response been like<br>any numbers for us i I don't know if I<br>can share an exact number but like way<br>more images than I thought the internet<br>had uh like appetite to to do like the<br>um we will cross a billion total images<br>in the notistant future i can say that<br>oh man that that's a huge number and<br>what's the response like any any<br>creative use cases that you've seen on<br>on images so far apart from you know<br>animating yourself yeah I I mean the the<br>thing that's been exciting to me is just<br>the breadth of creative use cases<br>obviously turning yourself into anime<br>was the one that kind of first lit up on<br>the internet but but it's been amazing<br>to see the sort of creative expression<br>people making just really new kinds of<br>images I had never thought about before<br>um there's obviously a lot of great<br>commercial use cases too where people<br>are you know using this in their small<br>business for new logos or graphic design<br>or whatever but but really the creative<br>spirit of people just doing things that<br>you know images kinds of images that<br>I've never seen exist before that's<br>that's been to me quite exciting and and<br>has India contributed i mean I think<br>India has been a big user of this from<br>your tweets at least india I I believe<br>is our fastest growing market now um<br>India's been been very special to us for<br>a long time um India was one of the<br>first markets outside the US that really<br>jumped on AI in a huge huge way and<br>since this moment has happened it is now<br>our I'm pretty sure it's our fastest<br>growing market wow that's that's that's<br>so it's been very cool to see that's<br>crazy too and and have you seen Indians<br>also use things outside of imagen have<br>you seen them use deep research are they<br>a paying market uh I was in India a<br>couple of months ago and the that was<br>obviously before image genen and you<br>know it was like non-stop stories from<br>people about what they were using chatbt<br>for um what what they were using chatbt<br>to do not catch 40 um but but it was<br>really it's like quite incredible the<br>degree to which it has permeated Indian<br>society for for all sorts of things you<br>know certainly we get feedback that our<br>price point is too high for mass<br>adoption in the Indian market and we're<br>looking at things to do there<br>unfortunately our compute costs are<br>still just quite high but we are hard at<br>work on more efficient models and I'm<br>optimistic we'll be able to bring cost<br>down over time amazing and I want to<br>touch upon the other side of this right<br>which is the UI and design jobs for a<br>second you know we thought when these<br>tools come out came out a year ago that<br>well we're going to have you know these<br>tools do most of the work for us but you<br>know in our own company we have a bunch<br>of people who are video editors we have<br>people who are designers and it turns<br>out we still need a person to run the<br>tool because humans still need<br>accountability someone maintaining<br>context and it's almost like the manager<br>of the tool in some senses right uh is<br>this how is this how jobs are going to<br>evolve is have you seen something<br>similar or do you have a do you have a<br>path for how jobs will evolve now that<br>these sort of tools are here I think<br>it's going to be uneven and it'll be<br>different for different kinds of jobs<br>there will be some jobs that are that<br>totally go away where the AI just does<br>them end to And mostly I think it'll be<br>a case of a new tool where people are<br>just much more productive and can do<br>work at a higher quality so there will<br>still be people that are in charge of<br>you know making a website look great but<br>the expectations of how good a website<br>looks just go up a lot and we all get<br>better stuff uh in fact maybe we get<br>more maybe there's more demand for that<br>in a world like that that that I think<br>can cut many different ways um and then<br>there will be totally new classes of<br>jobs that we just haven't seen before at<br>all yeah yeah do you have examples so<br>far there are some kinds of customer<br>support where I think an AI can just do<br>that end to end and that'd be in the<br>first category in the second since we're<br>on this example of graphic designers you<br>know taste still really matters and like<br>the<br>the you know when we went from graphic<br>designers had to do everything with like<br>uh paper and pencil or you know pens and<br>paper whatever you want to call it to<br>having computer tools they were able to<br>do more and do better but we still<br>needed a lot of good graphic design in<br>the world um we can now have a new tool<br>but and and no doubt it's going to<br>change the workflow of a graphic<br>designer and some things that some<br>people were paid for they won't be paid<br>for those tasks but what it appears to<br>me will happen is you will still need<br>some sort of job which is you know make<br>this website look beautiful um<br>and maybe there are fewer people that do<br>that but they make much more money maybe<br>there are more people that do it because<br>we just have a you know explosion of how<br>many websites we can get like may maybe<br>it turns out there was way more demand<br>for graphic design in the world than<br>than we could sort of like afford to<br>fill i think we just don't know on that<br>yet um and then there's other and then<br>in terms of like entirely new jobs you<br>know the first example we saw of that<br>was prompt<br>engineer before AI came along there was<br>no such thing as a prompt engineer not<br>only that there was no it would have<br>been hard to conceptualize that there<br>would be such a job um there's there's<br>many more but that's the example I give<br>is the sort of near and dear to my heart<br>one and it's so funny when whenever some<br>of these new jobs come in they're always<br>ridiculed at first right like even<br>building rappers and we'll come to that<br>in a bit but even with prompt engineer a<br>lot of people still don't take it<br>seriously well all you're going to do is<br>write English do you have a do you have<br>a defense for that do you feel there's<br>there's some way to you know make the<br>people who are now going to be prompt<br>engineers feel a little more you know<br>like they're not losing status for doing<br>it when I was young um I knew I wanted<br>to go study computer programming<br>and the adults in my life would say "Oh<br>that's like that's a hobby job these<br>computers aren't going to go anywhere<br>you need to go be a doctor or a lawyer<br>like you need to get a real stable job."<br>You know this is like a this is a<br>dangerous path to go down um and I<br>thought I understood something that they<br>didn't and I think I really encourage<br>people to have their own conviction on<br>what things are going to be like and<br>just because something is not<br>a just because something is not a<br>historically valuable or high status job<br>doesn't mean it won't be in the future<br>um this is this is like the way of<br>technological progress you know for<br>people that are watching there's a post<br>by Sam Alman on this which is the<br>strength to be misunderstood i love that<br>that blog post by the way so thank you<br>so much for that you wrote it many years<br>ago i think it was it was very<br>insightful when I decided to become a<br>content creator and it was a low status<br>job back then i was going to say about<br>your job you know I<br>remember you know like 10 years ago or<br>something uh I would like I would meet<br>people I I was you know I would go talk<br>about doing startups or whatever and I I<br>would like go meet young people who<br>wanted to start startups and I always<br>kind of like asked people what they<br>wanted to be or what they thought the<br>interesting industries were because I<br>think if you kind of like<br>ask young people that you get a you get<br>a perspective that is interesting and<br>different and usually right and every<br>not everybody many people said they<br>wanted to be a YouTuber at the time i<br>was like what but I I it hit me then<br>that I'm like ah I cannot dismiss this<br>like people dismissed computer<br>programmer when I was young yeah i I<br>want to touch upon something else that<br>the internet's been sort of debating<br>over the last few days which is you know<br>there's this video of Miyazaki so of<br>course it's an out of context video<br>where he's looking at a model from you<br>know many years ago something some<br>creature thrashing around and he said<br>that well you know I don't like this<br>essentially that was the the you know<br>the summary of what he said you know as<br>an artist who's been making videos for n<br>9 years I feel some of the craft and the<br>you know the effort I'd have to put in<br>to make an image or even to make a video<br>has dropped over the years and it is a<br>little bit upsetting it it feels like<br>it's a little bit of a bummer as an<br>artist right but as an entrepreneur as<br>somebody now that has a team to manage<br>and worry about their salaries and<br>payroll and whatnot it is very good that<br>this technology exists because it helps<br>me go a little bit faster right it helps<br>me you know make sure that I'm I'm<br>dealing with a lot a lot less stress do<br>you have a view on um you know how the<br>internet should think about this<br>evolving nature of art i think what it<br>was like to try to um do what you do 30<br>years ago it was impossible or very hard<br>i mean you'd need like a camcorder and<br>like VHS tapes and then you'd have to<br>like edit it in some complicated way and<br>then even if you did that you would have<br>to like distribute tapes because there<br>was no internet and no YouTube or there<br>wasn't internet but it wasn't very good<br>and now<br>like anybody on earth can get a<br>smartphone and record themselves and put<br>it up and if they have something<br>interesting to<br>say they get it out there and the world<br>benefits from that<br>um I think the democratization of<br>creating content has been a big netwin<br>for society it has not been a complete<br>win there are negative things about it<br>for sure and certainly it did something<br>about<br>the the art form but I think on the<br>whole it's been a win and this is<br>something that I very deeply believe<br>about the power of technology which<br>is giving everyone more tools making<br>things easier lowering the barriers to<br>entry um does significantly increase the<br>number of people that can contribute to<br>society and we all benefit from that<br>overall doesn't mean that it doesn't<br>like cause some job loss and some people<br>who had a sort of differential ability<br>to do something now have a lot more<br>competition but overall I think it's a<br>real benefit to society<br>um my own experience of this was<br>watching the barrier to entry to<br>starting a company really change um this<br>happened when it got cheaper to start a<br>company because of a whole bunch of<br>things but AWS was one of the biggest<br>and most important moments access to the<br>internet in general was was another big<br>one and there was a shift that Y<br>Combinator I think really helped drive<br>or popularize whatever you want to call<br>it uh starting in 2005 where young<br>technical founders could all of a sudden<br>much more easily start companies and<br>that led to an explosion of great new<br>products and technologies that the world<br>got to benefit from now some of the<br>people that those companies competed<br>with probably wish that hadn't happened<br>but as a citizen of the world there was<br>a lot more stuff openai itself I think<br>was an example of a company<br>that only got to happen because the<br>barriers to entry to a bunch of<br>different pieces of technology stack got<br>significantly low and this ragtag bunch<br>of us were able to just do something we<br>had no right to do yeah I think I agree<br>with you i think especially after Stripe<br>and in India after UPI there's just been<br>an explosion of new companies new new<br>types of things that were just not<br>possible because the amount of red tape<br>that it required to even get started i<br>assume now it's sort of become the same<br>in design in fact on the note of YC<br>right I actually you know I want to ask<br>you to put on your 20-year-old when you<br>were a 20-year-old hat of you know going<br>to YC again if you knew everything you<br>knew right now and the debate was you<br>should build a rapper or no you<br>shouldn't build a rapper OpenAI would<br>probably kill your startup or on the<br>other side foundational models are<br>worthless rappers are the only ones that<br>are making money and able to raise money<br>and whatnot um what would you do knowing<br>everything you knew but also with the<br>ability to shut out the noise i think<br>the whole like oh your startup is just a<br>rapper or oh like people used to say<br>that about other categories too you know<br>there was a time when people say well<br>you're just a rapper on AWS probably um<br>the people are building absolutely<br>incredible new companies based off of AI<br>most of them will fail or not do that<br>well but some of them will find a really<br>enduring business and generate a ton of<br>value and that's that's always the case<br>that's just how the world work like you<br>know most startups fail and some really<br>succeed<br>and it's easy to sound smart and to just<br>dismiss everything as a rapper or some<br>thin thing that's not going to you<br>know create much value and I remember<br>when people used to say very similar<br>things about open AAI yeah what was it<br>what was the they used to say<br>you're just you know you're just like a<br>little research lab and Google's going<br>to crush you and like you're just<br>putting this like thin layer on or<br>Nvidia is going to get all the value or<br>you're going to like you're doing this<br>like little thing and it's impossible to<br>compete with the big tech companies and<br>it was not long ago they were saying<br>this this is a couple years ago so I<br>don't know<br>i it sounds very cool and sophisticated<br>to always say why things aren't going to<br>work and you will be right most of the<br>time um and kind of the the most<br>important lesson I think you can learn<br>as an entrepreneur is that that or an<br>investor is that that is not how great<br>companies get built or invested in<br>so I think if I were 20 in NYC right now<br>and someone was calling my company a GPT<br>rapper I'd be like "Ah I'm I must be you<br>know in a promising direction." Do you<br>do you have any specific ideas that you<br>would be working on if you were 20 but<br>you were not an open AI you didn't have<br>the resources you had 500K as a matter<br>of principle I try to never answer that<br>question because I I the thing I care<br>about most is that people don't start<br>someone else's company like you<br>you the ideas that I have for companies<br>are probably mid at<br>best and the idea that some 20-year-old<br>doing YC right now that has a company<br>that has an idea for a company that<br>sounds totally crazy and bad to me<br>that's probably the next open AI and I<br>never want to discourage anyone from<br>that by saying some like safe middle of<br>the road like out of touch idea so go do<br>your crazy idea it's my advice that<br>makes sense it's almost like you need to<br>do what you're passionate about<br>otherwise you never end up building a<br>Snapchat because everyone would think it<br>was a bad idea on day one um that's<br>amazing Sam you know we did a we did a<br>podcast with Kevin uh your CPO the other<br>day and he gave me a timeline for when<br>code is going to be automated he said<br>2025 anthropic said 2027 he said "No we<br>can do better we'll do 2025 or maybe<br>2026." Um now that we know everything we<br>know do you have a timeline for when you<br>know a lot of writing code becomes<br>English<br>or is that not going to happen i think<br>it's the degree of automation that<br>matters to get to 100 truly 100%<br>automation you know you can make a<br>complex thing and never touch<br>code that that's one thing<br>but I'm less interested in that question<br>than when a coder becomes 10 times more<br>productive and I think that could happen<br>this year next year something like that<br>so a coder is 10x more productive i<br>think a coder is already several times<br>more productive i think there's already<br>like at least in you know the way people<br>around me have started working the good<br>ones who've adopted the yeah people<br>maybe say 2x 3x but I haven't heard<br>anyone I haven't heard many people I<br>have not heard a consensus for 10x yet<br>but you think that by next year it's<br>going to be a a 10x step function look<br>predictions especially predictions with<br>the timeline attached are always<br>difficult but I think we can aim at that<br>yeah interesting you know there's there<br>is a negative side to this right which<br>is that you know Daniel Grath had this<br>this one article where he said that $250<br>billion of India's exports are actually<br>IT services and if you think from that<br>angle that could just be GBD4 tokens<br>right or GDP5 tokens at some point for<br>India that $250 billion is very<br>important and not all of those engineers<br>are building are are elite engineers who<br>will benefit from this technology and go<br>10x right some of them are still<br>building simple crab crowd apps and<br>there is competition from the 10x<br>engineer maybe not from AI itself but an<br>engineer using the AI and sort of<br>outpacing them and It's they're able to<br>do this at much cheaper right because<br>the tokens are of course cheaper do you<br>have something that India can do uh to<br>actually make sure we absorb some of<br>that we we continue to be relevant there<br>if there's one area that I think I think<br>this is an important general question<br>but if there's one area where I think<br>that the world just has so much more<br>demand than we can currently supply it's<br>for<br>codewritten my my belief is that the<br>world just wants way more software and<br>is about to get it and if you give<br>people these tools to write software<br>much more<br>effectively I at least think for a while<br>what will happen is someone who today<br>can write code at some value will be<br>able to write code or create software<br>let's say at some much higher value now<br>they have to pay the AI a lot so they<br>won't capture all of that um and also<br>the market price for a piece of code<br>will go down but I I would bet this is<br>an example everyone seems obsessed with<br>Jevans paradox all of a sudden and I<br>would bet this will be an example of it<br>so a lot more software being built so<br>this it might it might the cost might<br>drop a certain amount but then also<br>demand goes up by that amount i think so<br>interesting that's very reassuring to<br>hear actually i don't think it's going<br>to be like that i don't I don't want to<br>like dodge the question because I don't<br>think it'll be like that in every<br>industry but code is one where I would<br>guess it's like that and and do you see<br>a lot of these I mean we've been seeing<br>sort of early agents right we've seen<br>deep research d research has been great<br>but DP is also a passive agent it's not<br>taking my card and doing something with<br>it so there's there's low DP research is<br>low risk um I've seen that you have been<br>very bullish about agents there are lots<br>of businesses that have been very<br>bullish about lots of other companies<br>that have been bullish about agents um<br>do you have do you have something in<br>mind for what agents will become over<br>the end of this year or by by next year<br>and some use cases that you think we'll<br>see on an everyday basis with agents<br>just to stick with the code example<br>because we're on that i I think this<br>idea that you can like say to a coding<br>agent um I would like this new feature<br>built and it goes off and pings you back<br>you know some period of time later and<br>says okay here's the pull request ready<br>to<br>go that I think will feel very crazy to<br>people there's other categories where it<br>kind of goes off and does stuff for you<br>on the internet operator points at that<br>a little bit um but I would expect the<br>coding one to be a big a big<br>anything on a consumer use case like<br>something like booking your your food or<br>something ordering a cab or something<br>like that that's the kind of thing that<br>I very much expect to happen and that<br>that'll feel like a very agentic task<br>let me ask you another question which is<br>if you were an employee at a job but you<br>were working at a legacy company let's<br>say they were slow to adopt all this<br>technology they're not even aware that<br>you know maybe they've tried it in<br>passing they've seen a GPD3 and they've<br>become dismissive about and lots of<br>legacy companies are like this if you<br>were sitting in that company and you<br>were unable to convince your company to<br>use these new tools because of<br>bureaucracy or whatever reason would you<br>what would your career plan be then<br>would you move to a faster moving<br>company but also you know take on the<br>risk of that company probably I would do<br>that<br>yeah both because I think companies that<br>don't adopt these things will be in a<br>tough place most I mean there's some<br>categories of companies that of course<br>be fine but also because I'd want to be<br>responsible for my own career success<br>and I'd want to make sure I was learning<br>and you know if I felt like I was not<br>getting exposure to these tools it would<br>seem super broken that makes sense okay<br>one last question and then we should be<br>good um can you give me a<br>counterintuitive insight you've had<br>about AI worked here for a very long<br>time of your life and you've done a<br>bunch of things you've done YC where<br>you've invested in companies you've<br>you've you've been around the valley for<br>a very long period of time what is<br>something you thought about AI but you<br>have a totally different experience now<br>i somehow thought it was going to be a<br>bigger change to society than it's<br>turned out to be um you know I I I I had<br>this sort of like sci-fi like vision of<br>what AI was going to be like and maybe<br>it'll still get there but right now it's<br>just this like crazy powerful tool and<br>life goes on mostly the same and and I'm<br>actually very happy about that i think<br>that's good i think I was naive before i<br>also think that would have been bad but<br>it's it it's very different than I kind<br>of imagined it 10 years<br>ago but I think I think that's wonderful<br>yeah I think that's wonderful too i<br>think you know having godlike technology<br>being able to solve at some point<br>medicine and help us do less of the work<br>we hate but more of the work we enjoy I<br>think is is is net good and I hope that<br>you know everyone benefits from it at<br>some point um thank you Sam uh this was<br>this was a wonderful interview thank you<br>so much for being support and and doing<br>this you know on such short notice and I<br>hope uh I hope you know India continues<br>to use this technology and I I have no<br>doubt I I I what is happening with this<br>technology in India I I have uh I've<br>never seen adoption like this of<br>anything anywhere and it's very cool<br>yeah I really hope we do more of that<br>and I hope you know we put it all over<br>the timeline and and you know you see<br>some of the things we you created thank<br>you awesome thanks Sam bye