Data Leakage - Illustration 2¶

You might be thinking "But, that last page seems contrived."

It isn't.

Read the attached article (it's short!) online here or the pdf here. Ask yourself:

What does the article imply about the value of ML?
What is the implication for how you react to new ML developments?
How does the experience of others impact what you think we should learn next?

In [1]:

import requests

class Tweet(object):
    def __init__(self, s, embed_str=False):
        if not embed_str:
            # Use Twitter's oEmbed API
            # https://dev.twitter.com/web/embedded-tweets
            api = 'https://publish.twitter.com/oembed?url={}'.format(s)
            response = requests.get(api)
            self.text = response.json()["html"]
        else:
            self.text = s

    def _repr_html_(self):
        return self.text

Tweet("https://twitter.com/cristobalyoung5/status/1557665261695643649")

Out[1]:

Sometimes machine learning *dramatically improves* predictive accuracy - performing much better than simple regression models

Turns out, those cases are all errors

Here's the mistake they are making... (1/3)https://t.co/fSETMYH6tj
— Cristobal Young (@cristobalyoung5) August 11, 2022