Becoming an AI engineering company

9 months agoNovember 19, 2024
46:01
377 views
12 likes
c

code.talks (ehem. Developer Conference)

Channel

Interviewed Person

Malte Ubl

Description

by Malte Ubl (CTO @ Vercel) AI represents the biggest disruption to software development since the release of the iPhone and maybe even the web itself. What makes the AI revolution unique is that it transforms how we write software, and what software we write, and that these dual disruptions compound: AI makes it easier to write software, and smarter AI models make it easier to use AI in software. Yet beyond the initial excitement lies a deeper change to software development that requires us to step out of our comfort zone as legacy processes, rooted in specifications, and automated testing confront a world of non-determinism and free-form human computer interaction. In this talk we’ll dive into Vercel’s own journey from being a traditional software company to becoming an AI company, and how you can apply some of the same lessons to your own business.

Transcript

Group by:

[Music] I'm so happy to be here in H B I used to live here I don't anymore I live in San

Francisco California it's been like 12 years but I'm glad to be back I'm also back for like 12 hours with a little bit jetl I did get some sleep but yeah always always good to be home in in my hometown uh awesome to have this event here all right so again my name is Mal I'm the CTO for sale I heard there was people wondering what weel is or not I don't know we make the we help people ship the best products so this is not a talk about verell just so we're clear uh

but it is kind of I like giving meta talks so it's a talk about our journey and that as an example for maybe the journey that some of your companies can do at the same time and maybe we can help you along the way what we do is we make JavaScript Frameworks which is exciting and then we built the infrastructure to help you operate the applications that you buil with

them all right again that's for The Meta context uh I think it's fair to say that AI is going to change software very broadly right and that's both true for developers and for end users so I think many of you are probably be using I don't know co-pilot maybe you're like super hip and using cursor you probably used jbt and you had that moment yourself where you're like wow this is actually great right like

like my my my world has changed right I was there in 2007 when the iPhone was was released and it felt very similar but but if you actually think back that was so slow like no one had a phone everyone had to buy one it was expensive and Android didn't even really launch until 3 years later so it was like this super slow transformation now we have this big transformation where chbt gets like 100 ion users in 3 months and it's accelerating at the same

time so that was the GPD 3 era couple years ago we then came to the gbd4 era and that's really interesting in that we now have multiple Frontier models froner models are the best models out there it's not just open AI we have Google with Gemini we have anthropic building CLA and super relevant is llama 3.1 from meta it's the only open weight model open weight means you can download

it you can put it on your own $100,000 GPU and run it yourself or you know I don't know ask someone else but like but it's obviously exciting that it's just out there and you can do that the next phase just last week we had open a launch 01 it's a reasoning name model it's very slow but I actually give credit to the cloud model they have they innovated on this notion of Chain of Thought prompt where you where the model kind of thinks about what it should do and then it iterates through that and it gets the

better outcomes um and you can do that today and obviously things are only going to get accelerate further in the future and I think it's worth calling this is disruption right and it's worth thinking about what A disruption is so A disruption is when the world changes could be for the better for the worse it could be good for you it could be bad for you it could be invariant orthogonal to you who knows right but

92 segments (grouped from 423 original)7450 words~37 min readGrouped by 30s intervals

Video Details

Duration
46:01
Published
November 19, 2024
Channel
code.talks (ehem. Developer Conference)
Language
ENGLISH
Views
377
Likes
12

Related Videos