Introduction

[00:00:00]
>> Will Sentance: We are live. Hello and welcome to artificial intelligence and machine learning for software engineers, prediction and neural networks on Frontend Masters. Let's have an enormous thank you to Frontend Masters.
>> [APPLAUSE]
>> Will Sentance: As you can hear, we have hundreds of people here in the room, but we're gonna select just seven of them to introduce themselves.

[00:00:22]
Here we go, machine learning, artificial intelligence, and neural networks. I'm just delighted to be able to share this with everybody, because prediction in software engineering changes what we can build with code. We're gonna be able to do things like predicting fraud, okay? Maybe we've heard of this before, maybe a fraud detection model.

[00:00:43]
We're gonna build one. We're gonna build a DoorDash refund predictor that's gonna allow us to automatically give a refund, improve the product experience on DoorDash, or Uber Eats, or other delivery services. But predicting fraud, it's gonna extend to predicting the content of images. It's gonna turn out to be no fundamental difference to predicting fraud.

[00:01:09]
We're then gonna discover you can extend that to predicting responses to questions. All of this prediction, this ability to see the future, I never said that phrase before, that may be a bad way of describing it. Is gonna enable not only bridge and fraud, image content, responses to questions.

[00:01:28]
But also emergent phenomena, and maybe even intelligence, if all we think there is in the universe is distributions of data. We'll see what all of this means for these terms, will all come out. But at the core of all this grand set of goals is one thing, the ability to develop rules to match patterns in known data, a sample.

[00:01:55]
Again, see all these terms in the coming slides. And then generalize to a population, from a sample to a population, where we have unknown data that we want to, instead of know, we want to predict. These transformative things are gonna be powered by enormous amount of compute. That is the processing power that it takes to do the work of prediction or to do the work of generating the tools to do prediction, is remarkably heavy lift.

[00:02:34]
This is why Nvidia is worth, how much today? Is it in the billions? Well, I guess definitionally it's in the millions, I guess, definitionally. But is it in the billions? Is it in the trillions?
>> Speaker 2: Yes.
>> Will Sentance: Uh-huh, how many [LAUGH] trillions?
>> Speaker 3: 3.5.
>> Will Sentance: Three and half trillion dollars, all because of the demand for compute.

[00:03:00]
It also requires a deep understanding of the data and exploratory introspection into the nature of the data. And phenomenal creative scientific research, two of the Nobel Prize winners this year, two of the Nobel Prizes, sorry, more than two of the Nobel Prize winners. Two of the Nobel Prizes this year were from breakthroughs in scientific research that backs up all this work.

[00:03:26]
One by Geoffrey Hinton, who also popularized an algorithm that we'll work with today, or at least work towards, known as backpropagation, a key part of building out neural networks. Enormous amount of scientific research has gone into all this work, or to find the best fit rules, known as models, to capture the patterns in data.

[00:03:45]
Software engineers, all of us, are at the heart of turning these machine learning AI models into products. Things you can use, things that can, quote, change the world. In collaboration with machine learning engineers and data scientists.