The 1,000x AI Engineer: Swyx

00:00:00.000 | A few logistical things.

00:00:16.520 | One, I'm carrying a magic trackpad because everyone has clickers.

00:00:20.280 | What if we had multiple dimensions?

00:00:22.280 | We're going to experiment with this today.

00:00:24.280 | And two, I'm using AI, like fancy new everything, right?

00:00:29.560 | So this is Tome, and we're going to go two-dimensional with our slides as well.

00:00:34.420 | So I'm here to talk about the AI engineer.

00:00:36.720 | You're all here because you believe that there's some value to this idea.

00:00:40.260 | And then I just put like a ridiculous 1000x on this.

00:00:42.520 | But I do think there is some meaning towards thinking about higher orders of magnitude towards

00:00:48.520 | raising your ambitions.

00:00:49.560 | And that's what I would like all of you to do today and to do with your friends back home.

00:00:55.240 | So and obviously a lot of AI-generated art because, I mean, it's an AI conference.

00:00:59.120 | We've got to do it.

00:01:00.120 | We've got to do that.

00:01:00.920 | First of all, I want to congratulate you on being here.

00:01:03.980 | I'm not talking about here location-wise, physically.

00:01:07.880 | I'm talking about here in terms of the point in time.

00:01:11.820 | Imagine if you were a mathematician.

00:01:13.080 | When was the best time to be born?

00:01:14.980 | I would propose around about 600 AD, this dude, Brahma Gupta, he invented zero.

00:01:23.740 | Pretty novel invention that took us only 4000 years to do that.

00:01:28.180 | But there's certain times where if you're in that field, you have to be there.

00:01:32.280 | That's the thing.

00:01:33.280 | When was the best time to be born?

00:01:34.280 | There's a right answer.

00:01:35.280 | 1905, 1927.

00:01:36.280 | And this conference kind of is inspired by the Solvay Conference.

00:01:45.280 | That's Albert Einstein, Marie Curie, and a lot of people that you just saw in the Oppenheimer

00:01:51.140 | movie.

00:01:52.140 | Same thing.

00:01:53.140 | If you made cars, there was a right time, 1900 to 1930.

00:01:58.560 | If you made personal computing products, 1980 to 2010.

00:02:03.200 | If you're a millennial, if you're very online, you ever get these memes like you're born too

00:02:07.020 | late to explore the earth, born too early to explore the stars, you're not too late.

00:02:13.540 | We are here.

00:02:14.900 | This is based on demographics and history, the approximate timeline of all of humanity.

00:02:19.580 | We know that we're roughly about 73% of all concurrent intelligences if we don't expand

00:02:24.820 | our own intelligences or go to other planets.

00:02:29.200 | So my argument and my message to you today is that you are just in time, and the timing

00:02:33.060 | is right to 1000x.

00:02:37.080 | I think a lot of my technology and industrial organization thinking is informed by Carlotta

00:02:42.480 | Perez, one of the most influential thinkers on tech revolutions.

00:02:46.780 | She wrote this book about the installation and deployment periods of tech cycles.

00:02:51.460 | And we're definitely going through one today.

00:02:53.020 | A lot of you on your mind here, I know you're here, but also mentally you're back home thinking,

00:02:57.620 | how much of this is a fad?

00:02:58.600 | How much of this is Web 3 again?

00:03:04.740 | And we've seen this over and over.

00:03:06.020 | the historians greater than us have explored this over the industrial revolution, the age

00:03:12.320 | of railways, age of heavy engineering and steel, oil, and most recently the tech revolution.

00:03:18.860 | Funny enough, they all roughly span between 50 and 70 years, and if you're around in that

00:03:25.260 | time, that's the field to be pursuing.

00:03:28.460 | So when did the AI revolution start?

00:03:30.940 | We're very lucky.

00:03:31.940 | It's very hard historically to place a start point on something that changes human civilization.

00:03:36.820 | We have a moment, 2012, AlexNet.

00:03:38.820 | We're roughly 10 years on.

00:03:42.920 | And we can put numbers to it, right?

00:03:45.020 | So most of the time these curves are sort of theoretical.

00:03:48.080 | They're just kind of like, bleh, like y-axis is just bleh.

00:03:52.960 | Here we can actually just put the amount of compute we're using towards training in models.

00:03:56.820 | There's a huge inflection, that's AlexNet, right on the blue dot over there.

00:04:00.460 | That's a huge inflection where we realized, gradually realized, it took too long to realize,

00:04:05.100 | but scale is starting to work.

00:04:06.700 | And if you actually take this out, a lot of people have been taking this out, and I want

00:04:12.740 | you to take scaling seriously.

00:04:14.560 | There's three reasons why six is a magic number.

00:04:17.200 | There's a very famous investor who I shall not name, and says, "Imagine roughly 690

00:04:21.560 | in moral compute by the end of the decade and plan for that."

00:04:24.140 | So there is more of this coming, linear projection-wise, and you can plan on a lot more investment

00:04:30.080 | in language models.

00:04:32.020 | John Carmack says there's six key insights towards AGI.

00:04:35.600 | And lastly, George Hotz has these really nice analogies.

00:04:39.040 | GPT-3 took about one-person year of compute.

00:04:42.040 | GPT-4 took about 100-person years of compute.

00:04:44.440 | You stretched it out to GPT-10, the difference between GPT-4 and GPT-10, again under the six-fold

00:04:51.380 | increments in GPT advancements.

00:04:53.200 | And that would be more compute than the equivalent compute of every human ever who ever lived.

00:04:58.840 | So just being in the right moment, you will get to live on top of these mega, mega trends

00:05:03.960 | that is greater than any single one of us.

00:05:06.380 | And I think you're all here thinking about the AI engineer.

00:05:09.320 | And I put it in a very, very small sort of local context of, hey, what's the org chart?

00:05:15.140 | Where do the ML engineers sit, where the ML researchers sit, where the software engineers sit?

00:05:19.080 | And what's the gap that's opening?

00:05:21.080 | It's the AI engineer.

00:05:23.900 | It's very much of a demand and supply argument.

00:05:25.900 | There's something like 100,000 card-carrying data science machine learning engineers.

00:05:30.400 | And GitHub claims to have 100 million registered developers.

00:05:34.080 | I don't know what the real number is.

00:05:36.180 | You can debate 40 to 50 million to 100 million.

00:05:38.940 | It's orders of magnitude more.

00:05:41.420 | So we think there's going to be much more AI engineers than ML engineers.

00:05:46.520 | There's all these reasons why.

00:05:47.860 | Same reasons that I mentioned in the blog post that you've all read.

00:05:51.860 | And also, why engineering and not just prompting is because LLMs themselves are not AGIs yet.

00:05:58.100 | We actually have to coordinate them in systems of software.

00:06:00.840 | We have to write code around them and orchestrate them with code in order to do something useful.

00:06:05.660 | And we already know how to do code.

00:06:08.280 | So I want to spread it out a little bit more.

00:06:10.160 | I think that the conversation on AI engineer has a vague discrepancy.

00:06:17.460 | And I want to basically split it out into three areas of AI engineer.

00:06:20.540 | Software engineer enhanced by AI tooling, like Co-Pilot.

00:06:24.140 | Software engineer building AI products, like Mid-Journey.

00:06:26.800 | AI product that replaces human engineer, potentially like Auto GBT and maybe Replica Ghost Rider.

00:06:33.560 | So let's give these guys a name.

00:06:35.120 | And in case you're wondering, enhanced by versus replaces, I think about it very much like the

00:06:40.620 | self-driving car terms.

00:06:41.620 | Like level two, level three.

00:06:42.620 | There's a difference between whether humans in the loop were there or humans as to fall back.

00:06:48.120 | So let's name it, three major types of AI engineer.

00:06:51.000 | The AI enhanced engineer for people who are enhanced by AI.

00:06:55.020 | People who build AI products, AI products engineer.

00:06:56.900 | And then the AI engineer agents who is not human.

00:07:01.780 | And naturally, of course, if you're interested in sort of progressing up the career ladder,

00:07:05.320 | AI enhanced engineer, then product engineer and engineering agent.

00:07:09.380 | So this talk was really inspired by actually Amjad, who is speaking next, where he did a recent

00:07:15.280 | talk with the XTNCD podcast, and Sam Altman, who actually sees 1,000 X engineers in OpenAI

00:07:21.280 | every day.

00:07:22.280 | And it's really a set of stackable 10 by 10 by 10 improvements.

00:07:26.620 | Over the course of the next two days, I think you'll be seeing a lot of the speakers will

00:07:30.320 | be working on different parts of this stack.

00:07:33.260 | So I really encourage you to think about where in your life this AI movement can improve

00:07:39.000 | and increase your productivity.

00:07:43.440 | I'm very, very honored to have drawn from all over the world the leading eight lights of the

00:07:50.720 | AI engineering movement.

00:07:52.040 | We are a very small room today.

00:07:53.200 | I do think we can 100 and 1,000 X from here.

00:07:56.940 | And it's not just about tools and speakers, it's also about you.

00:07:59.040 | So I highly encourage you to take part in all the opportunities that we have for you to

00:08:03.240 | mix and mingle with each other, with the speakers and with the sponsors as well.

00:08:07.980 | So there's that.

00:08:08.980 | The final word I do want to offer you is effectively what I think, in terms of non-technical

00:08:15.720 | terms, the 1,000 X engineer could offer.

00:08:19.320 | My favorite advice for what a 10X engineer could look like is an engineer that teaches 10 other

00:08:24.720 | people what they know.

00:08:26.720 | networks, that's not a technical term, but it is very useful.

00:08:31.460 | And there's all these scaling laws for networks, which I really keep in mind.

00:08:34.460 | So you can go from O of N to O of N squared to O to the power of N. But really what O of N

00:08:41.200 | is, is you attending all the talks and consume all the content and letting people in with

00:08:45.940 | your Pac-Man rule.

00:08:47.940 | O and Squared is helping others learn.

00:08:49.940 | My very first blog post was at exactly a conference like this where I was encouraged to write something.

00:08:53.940 | And of course it was on machine learning.

00:08:56.940 | And finally going home and then building your own networks of AI engineers and helping to

00:09:03.100 | grow networks of learning as well.

00:09:06.100 | So I hope you take that with you in your AI engineer journey.

00:09:09.220 | I hope that over the next few days you get a sense of what it's like to be at the start

00:09:12.820 | of an industry.

00:09:14.120 | And I'm just glad to be here with you.

00:09:15.680 | Thanks so much.

00:09:16.260 | Thank you.

00:09:17.260 | Thank you.

00:09:18.260 | Thank you.

00:09:19.260 | Thank you.

00:09:20.260 | Thank you.

00:09:21.260 | Thank you.

00:09:22.260 | I'll see you next time.

The 1,000x AI Engineer: Swyx

Chapters