Cassie Kozyrkov
@decisionleader
Followers
45K
Following
5K
Media
334
Statuses
4K
CEO at Data Scientific. Former Chief Decision Scientist, Google. ❤️ Stats, AI, puns, decisions.
New York, USA
Joined November 2009
Who needs an #AIPC? 🤖 Trick question: everyone!. AI-powered apps are for all of us, not just the AI pros. From Zoom blurs to productivity apps, these @IntelBusiness machines handle it all—no more slooow performance and tragicomic battery life for your toaster!. Thanks, Intel,.
7
2
6
Arriving tomorrow! Your fun + intuitive guide to these #AI algorithms:.* k-Means.* k-NN.* Perceptron.* Maximal Margin Classifier.* SVMs.* Decision Trees.* Boosting.* Random Forests.* Ensemble Models.* Naive Bayes.* Linear Regression.* Logistic Regression.* Neural Networks
20
208
2K
For those who missed my personal news last month, I've quit Google. I'm doing a bit of blogging to answer all the questions I've been getting about why I left, what it was like, and how things are going now that I'm on my own.
39
106
930
A free #AI course that puts the fun in fundamentals!. Sign up here: - the first 90min arrives in a few hours!. #DataScience #ML #MachineLearning #RStats #Statistics #Analytics #ArtificialIntelligence #DeepLearning #MFML
27
312
884
A huge thank you to all of you for encouraging me to write! I just noticed that on Medium, my community of followers is 70% the size of Barack Obama’s. Whoa!. I've published 180 articles, so it's time to break them up into learning paths. Here they are:.
14
88
573
Why #data wrangling is 10% skill, 90% anger management. #DataScience #statistics #rstats #AI #analytics. Image via @dr_keithmcnulty.
26
157
559
Personal news! Following a wonderful decade, I'm now officially an ex-Googler. It's bittersweet to leave my colleagues and my work identity as Google's Chief Decision Scientist behind, but that's how we keep learning and growing. Goodbye, Google. Hello world!. Let's find all.
50
19
534
Now in a single video: Making Friends with #MachineLearning (#MFML). Enjoy all 6.5 hours of Google’s legendary #AI course designed to enlighten AI beginners, grow technology leaders, inform better citizens, and amuse AI experts!. Don't forget to RT 💖.
22
154
491
#TensorFlow just went full #Keras! (!!!!!) Here's why that's an earthquake for #AI and #DataScience. ⬇️. @fchollet @TensorFlow @tensorflow_devp .
12
162
500
The best #MachineLearning and #AI course you've ever watched is now available to everyone:. * Designed from a tech leadership perspective. * Wonderful intro for total beginners. * Cheeky comedy for total experts. * Free on YouTube ⬇️.
2
98
460
Today, I want to demystify ChatGPT — a fascinating new application of GANs (Generative Adversarial Networks) that has been generating a lot of buzz in the AI community. Read this for an intro to ChatGPT, awesome demos, and how to try it out yourself:.
13
93
425
Decision tree cocktails, anyone? A behind-the-scenes look at the difference between support vector machines (SVMs) and decision trees, explained with mixology.
3
19
354
😂 #unexpected.🇬🇧 UK Border control: Where are you travelling from?.Me: New York.🇬🇧 UK Border control: How long are you staying for?.Me: The weekend.🇬🇧 UK Border control: Are you going to run for prime minister?.Me: Should I?.UK Border control: . well, everyone else is doing it.
14
12
369
I'm excited to announce that Google's Dataset Search is now out of beta!. If you aren't familiar with this #DataScience game-changer, I've written about it and explained how it works here: #statistics #analytics #rstats #AI #data.
9
201
344
You can’t do data science if there’s no data. If you get hired to be head of data science in an organization where there’s no data and no data engineering, guess who’s going to be the data engineer…? You!.
20
35
323
Starting out in #DataScience? Things to keep in mind:. - Useful is worth more than complicated. - Data quality is worth more than method quality. - Communication skills are worth more than yet another programming language.
3
81
315
Want data? Here's 20 million datasets at your fingertips…. #data #DataScience #AI #statistics #rstats #ML #MachineLearning #analytics.
3
153
304
A preview of @Google's Making Friends with #MachineLearning. a course for everyone! It's beginner-friendly + designed to amuse experts too. Sign up here: - the first 90min arrives this Friday!. #AI #ML #RStats #ArtificialIntelligence #DataScience
6
89
284
💌 🧠📊 I'm starting a #newsletter. who wants in?. Data-driven decision-making, plus a dash of humor!. #DataScience #Statistics #rstats #analytics #AI #MachineLearning #puns #humor #psychology #leadership.
34
80
290
It's here!! The most fun you'll ever have in a #MachineLearning course!. Making Friends with Machine Learning:.Part 1 - Part 2 - And here's all the course info: RT for your human friends. 💗. #MFML.
6
88
281
👍 Just wait till everyone sees how hard AI is to maintain in production. .
Personally, I think AI will lead to more software engineering jobs, not less. ¯\_(ツ)_/¯.
19
41
273
A few more definitions to the Stats Gist List added today: Enjoy scedasticity explained as sausages vs trumpets, plus a bunch more. 🌭🎺
1
49
251
"It's okay, we don't need a data engineer.".
5
63
244
0/ Essential philosophy for #DataScience, a thread of 32 questions. Grab a friend (virtually) and tackle these 32 essential questions (all with more than one reasonable answer) that every serious #data professional should answer for themselves. #philosophy #rstats #statistics.
11
80
240
After 2 years of writing articles for your amusement, I finally decided to try my hand at short vlogs. Here's a sneak peek into my makeshift home studio. Hope you like it!. My channel:
11
25
228
Some of your best skills don't have a name yet. Don't forget to be proud of them.
8
20
212
Machines are better than me at whatever they're for. that's the point of tools. A calculator is better than me at 238÷182 and a bucket is better than me at holding water (and far better at holding water than headlines spreading fear of machines that are "better than us").
4
77
220
First time I heard a math class I was in described as "male-dominated" I remember being confused: how can it be male-dominated if I'm the one dominating it?. Quality, not quantity. 😉.
4
26
206
What if I told you I can show you the difference between Bayesian and Frequentist #statistics with one single coin toss?
12
21
197
Ever been tempted to learn some #Statistics? . Here's my mini-course with the complete set of my statistical musings (blog, podcast, videos), organized for you by theme! . RT if you like it. 💙. #DataScience #RStats #Analytics #DecisionIntelligence.
4
93
199
We're conditioned to believe that if we leave a job/relationship without another lined up, it’s the end of the world, but it might be the smarter thing to do:.- Most of us wait too long to quit anyway.- You won't discover your best options while you're trapped in the daily grind.
14
15
195
Numbers won't help you get a good answer to a bad question.
4
35
191
It's finally here! Your cheeky and fun guide to #AI algorithms: Happy Thanksgiving, everyone! 🦃. #MachineLearning #MFML #DataScience.
4
34
185
Leaders! Stop dropping the ball on #machinelearning projects! . This thread is a manual for the leader's role in ML/AI projects, with short guides to each bit. 1/🧵.
4
58
186
"GPT is autocomplete on steroids but it gives the illusion that it's more than that.". -@GaryMarcus, chatting with Noam Chomsky here at Web Summit.
8
31
187
🤖🤗🤖I made the friendliest #AI course for you: Making Friends with #MachineLearning!. The first 9 min will take you from zero (Is AI magic?!) to knowing where to apply it and even having a sketch of how GANs work. Enjoy!. #DataScience #rstats #ML #data.
9
56
178
When you keep finding bugs in your salad.
12
26
180
In statistics: . Type I error is changing your mind when you shouldn’t. Type II error is NOT changing your mind when you should. Type III error is using all the right math to correctly change your mind about the wrong question.
5
37
175
Is the emperor wearing any clothes? Let's strip way the code, math, and marketing to see the bare ideas behind how #MachineLearning algorithms work.
9
59
171
When you need to learn something quickly, read. When you need to clarify something, write. When you need to prove something, build.
8
36
175
#AI algorithms explained so simply that humans of all stripes and all career backgrounds can understand them! Coming on Thanksgiving Thursday. Who's excited?
7
16
170
Unlike data scientists, data engineers tend not to spend much time looking at data. Instead, they look at and work with the infrastructure that holds the data. Data scientists are the data-wranglers, while data engineers are the data-pipeline-wranglers.
8
16
174
"When will #AI surpass humans?" . 🤔At what, though?. Pen and paper has had a better memory than you this whole time. That's the point of tools. We don't use a tool unless it's better than us at something. All tools are better than us, that's how this works. AI is a tool too.
13
55
166
A reader asked me to cover these topics in my blog:.* data types.* classification vs regression. The letter was cute so I'm tempted - should I do it? Would you read it?.
28
5
168
622.6% of data projects need unit testing 364.3% of the time.
8
19
164
Course: Making Friends with Machine Learning.Section: Part 1 - Introduction.Style: Powerful analogies + humor to help you understand deeply.Length: 90min.Designed for: Everyone.Cost: Free.When: Shared on Friday.Where: Available via newsletter (link below).
2
34
164
If data science is the discipline of making data useful, then data engineering is the discipline of making data usable.
1
41
163
It is really hard to think of everything alone - let's face it, we humans are creative but not *that* creative. The more identical your teammate is to you, the less ground your ideas will cover together. Diversity helps creativity flourish.
5
32
152
Where can I get me one of these make-me-look-like-I'm-AI-generated accessories? 😂👇
10
10
150
My top tip for learning to be a great technical presenter, even if your content is very equations-heavy:. Open a cookbook at random and meditate on how you would present it in a way that doesn't put your audience to sleep. Yeah, I'm not joking. Why do this? 🧵.
5
25
149
All jobs are made up primarily of boring bits no one tells you about when they're summarizing 5 years in 15 minutes.
8
20
144
Sometimes I wish I had the absolute shamelessness that it takes to call yourself an AI influencer when you only got into it after ChatGPT. My home is too full of mirrors, though, it would be inconvenient not to be able to look myself in the eye. But uggggh, how do I bulk block?.
18
12
140
As every engineer knows, regular drawers that make optimal use of space won't get you funded/promoted. 😂.
Engineering.
4
12
139
Data is only useful if you act on it. Don't just collect data. Interpret it. Experiment with it. Use it to inform better decisions.
7
19
139
A decision scientist's 2 tips for new year's resolutions involving diet and exercise, a thread. 1/🧵.
7
32
138
Since people are talking about #SimpsonsParadox in #COVID19 data, this illustration might be useful.
3
26
138
As a statistician, I’m painfully aware of how easy it is to lie with numbers.
6
21
138
As AI gets better, decision-making is going to become one of the most important skills in the world.
13
18
124
In case you missed me recently, I've been battling pneumonia these past few weeks. It's good to be back from the dead! 🧟♀️.
21
0
127
People: "We demand answers." .Statisticians: "The data/tools to answer to your question don't exist. No, wait, don't use that. That's not what it's for! What are you d-- No, stop that. Ow! Hey--". #statistics #rstats #DataScience.
5
30
122
I once asked ~20 colleagues in #AI research at Google which topic gets too little attention in #MachineLearning courses. The most popular answer by far? Validation. So, here's a video to get us started: what's the difference between validation & testing?.
3
32
126
What makes an analyst excellent? Speed! But that's not as simple as it sounds. They need 6 kinds of speed. 1/🧵. #DataScience #Analytics #rstats #DataAnalytics.
3
27
120
Someone asked me today for advice about, essentially, how to be okay with not knowing all the things about all the thing. Especially when there's so much new stuff to know. The pace of all the tech developments is making ambitious people panicky.
7
27
121
Indecision is a decision. It’s a choice to do nothing, facilitated by thought loops and fear.
14
23
120
When not to use #MachineLearning / #AI:.*you're not automating.*no data.*you can look up the answer.*there's no pattern that connects inputs & outputs.*you can find a recipe today but it won't apply tomorrow (nonstationarity). #ML #rstats #DataScience .
9
45
118
Are you Bayesian or Frequentist? Here's a handy way to self-diagnose.
2
38
114
Can I tell you the big ideas in #statistics in just 12 minutes? Challenge accepted!. Podcast episode: Statistics for people in a hurry!. #rstats #DataScience .
3
24
115
🙃In the unlikely event that I ever own a boat, I shall name her Sunk Cost.
6
4
116
Happy 2022! 🥂 As always, some resolutions might not go as planned. But this year, try to focus more on your learning rate than your winning rate.
4
9
118
Sorry, buddy. If you can explain what a confidence interval *is* in one short sentence, then you did not understand it. Same goes for p-value. One sentence is enough for what they *do* but not what they *are.*. #DataScience #Statistics #rstats #Analytics.
5
11
113
#Statistics:."statistically speaking. "."almost significant"."accept the null"."we reran the test until. ". #ML:."the AI decided. "."after exploring the data, we split it. "."test performance was even better than our training performance". #Analytics:."we conclude"."insights".
What phrase signals that the person using it doesn't understand your field?. Example: "Computer error.".
5
29
113
Tip: Resist naming your metrics until the very end. (+ other advice for measuring the abstract, the impossible, and the gooey human stuff). If you liked it, put a RT on it. #DataScience #psychology #Statistics #rstats #MachineLearning #UX.
5
74
100
How many of these can you explain confidently?. CDF, Distribution, Estimate, Expected Value, Histogram, Kurtosis, MAD, Mean, Median, Mode, Moment, Probability, PDF, Random Variable, Skewness, Standard Deviation, Tails, Variance. Got gaps? I've got you. 👇.
2
26
106
This should be obvious, people.
4
26
107
You call them "soft" skills, I call them "hardest to automate" - soft skills are not soft!. But where are the programs to train great data leaders? Who teaches these skills?. 🤫 Little secret: I'll be releasing a new course for you soon! Who's excited?.
4
17
105
If you’re considering taking a job as a head of data science, your first question should always be, “Who is responsible for making sure my team has data?” If the answer is YOU, well, at least you’ll know what you’re signing up for.
4
10
102
Data quality seems to be exactly the kind of “everybody’s job” that ends up being nobody’s job. 1/🧵.
4
17
108
There are many different kinds of data professionals. It would be a tragedy if we forgot that and started yelling at people for having different skillsets. Then they'd all only be useful for one thing - one tiny thing in the huge universe of data. Different is good! 1/2.
3
19
105
So honored to be selected as a Top Voice for 2020 in #DataScience and #AI on LinkedIn. Thanks for cheering me on! .
14
7
104
Welcome to #Statistics, where The Answer is p = 0.042 but you don’t know what the question was!.
8
21
100
Data professionals take note: having conversations is an important part of #analytics and #DataScience. There's much more to this career than just code and numbers, so working on your people skills is key!.
6
18
98
Not to be morbid or anything, but if you're going to be obsessed with mortality rates, then use the right denominator. The wrong one:. Rate != (Deaths TODAY) / (Infections TODAY) . This would only be correct if death were instant. Adjusting for virus lag, rate estimate is larger.
3
21
101
🤦♀️When science/math/engineering folk use "artsy" as a derogatory term, I have to wonder whether they're immune to irony. 1/🧵.
5
14
102
10 differences between amateurs and professional analysts! 🧵.
3
19
98
If I ever came into possession of a yacht, I would name her Sunk Cost.
11
8
101