Developing your AI BS detector

# Developing your AI BS detector
## Rational AI: Demystifying Machine Intelligence In The Enterprise
### MaRS DD 20180405

Adam Drake

@aadrake

http://aadrake.com

---

# @aadrake #hacker

# @aadrake #thoughtleader

---

# Overview

1. Background

1. AI for fraud detection

1. Wheat and chaff

---

# Background

* 20+ years in tech

* Advisor to growth-stage startups

* Management consulting and high-performance tech/ML

---

# AI for fraud detection

* Problem domain

* Founding team

* AI solution

---

# Our AI

* 100% proprietary AI framework (fused knowledge base/inference engine architecture)

* Accurate - Proven savings with real-world customers

* Real-world - Developed via on-the-ground work and tunable/parameterized for different industries and companies

* Scalable:
    * Written in Go (very fast!)
    * Standard HTTP/JSON API
    * Devops from day 1 (microservices architecture, single Docker image to large Kuberneties cluster)
    * Insanely high performance (27,000 transactions per second on a small AWS instance = 2.3 BILLION transactions per day)
    * ALL VisaNet traffic (~57k RPS) for ~$20/mo

---

```
func isFraud(ccIsForeign bool, amount float64, transactionsToday int) bool {
    if ccIsForeign && amount > 1000 && transactionsToday > 5 {
        return true
    }
    return false
}
```

---

# AI is a **PURSUIT**, not a **THING**.

---

# AI is forever a set of tools that **don't exist YET**, for solving problems that **aren't solved YET**.

---

# AI is done, to the extent it is *done* at all, by **research teams** not **product teams**.

---

* Executive Information Systems

* Business Intelligence

* Data Mining

* Data Science/Big Data/Machine Learning

* Deep Learning

* AI

* ...?

---

# An anecdote on fraud, AI, and rules-based systems

---

# Some BS Detector questions

---

# What specific problem are you solving?

---

# What is the most naive solution?

# Did you try that?

---

# What about the next-most-naive?

---

# Why is what you're doing considered AI?

---

# Are you using a framework?

#### If so, why is your solution defensible?

#### If not, explain why you aren't wasting time.

---

# When you hear AI, **be suspicious**.

---

# @aadrake

# http://aadrake.com

# Questions?

---

```
package main

import (
    "fmt"
    "net/http"
    "strconv"
)

func isFraud(ccIsForeign bool, amount float64, transactionsToday int) bool {
    if ccIsForeign && amount > 1000 && transactionsToday > 5 {
        return true
    }
    return false
}

func main() {
    http.HandleFunc("/", func(w http.ResponseWriter, r *http.Request) {
        cc, err := strconv.ParseBool(r.URL.Query().Get("foreigncc"))
        amt, err := strconv.ParseFloat(r.URL.Query().Get("amt"), 64)
        numTrans, err := strconv.Atoi(r.URL.Query().Get("numtrans"))
        if err != nil {
            w.WriteHeader(http.StatusBadRequest)
            return
        }
        fmt.Fprintf(w, "{\"fraud\": \"%v\"}", isFraud(cc, amt, numTrans))
    })
    http.ListenAndServe(":8080", nil)
}
```