orasis

12 ms·

91.
by orasis 4y ago
The author argues that AI algorithms don’t have desires, yet this is precisely what reinforcement learning simulates. With reinforcement learning positive rewards are assigned for positive outcomes (a “carrot”) and negative rewards are assi
92.
by orasis 4y ago
“The Century of the Self” by Adam Curtis. https://youtu.be/eJ3RzGoQC4s It goes through each decade of the 1900s and explains how Freud’s psychology and the new field of marketing completely reshaped society. For the first t
93.
by orasis 4y ago
That is an extremely difficult film. For me it was on par with watching Schindler’s List.
94.
by orasis 4y ago
I didn't find the author's argument compelling. The "Hero's Journey" continues to be a useful model to understand both myth as well as an individual's growth into mature adulthood.
95.
by orasis 4y ago
What brings you joy on a daily basis? I’m a black belt in Jiu Jitsu and as my mastery continues to deepen it just becomes even more fun and rewarding and I now have a global network of extraordinary people. It’s been well with the thousand
96.
by orasis 4y ago
Many gym’s business models don’t work without subscriptions. Should they simply cease to exist or is it maybe a decent thing that customers subsidize each other?
97.
by orasis 4y ago
This is an interesting result from a real world data set. It’s certainly got me thinking about how budget constraints might be best used with https://improve.ai .
98.
by orasis 4y ago
For me, it was an intentional MDMA session focused on shame that spontaneously cracked open my self love. The result is extraordinary - I really don’t care what people think about me because I no longer need validation and “love” from stran
99.
by orasis 4y ago
Thank you. I’m surprised that anyone read that part. Thank you for noticing.
100.
by orasis 4y ago
Thanks so much for pointing this out. I’ll get that fixed!
101.
by orasis 4y ago
We use XGBoost as the core learner for reinforcement learning at https://improve.ai despite the popularity of neural networks in academia. With tabular or nested data a human has already done a lot of work to organize that data
102.
by orasis 4y ago
When I was burned out in my late 20s, I wish someone had made me aware of MDMA-assisted therapy. This book is a nice introduction: https://www.amazon.com/Dose-Hope-Story-MDMA-Assisted-Psychot...
103.
by orasis 4y ago
A JSON array doesn't allow easily appending a row. JSONLines is a bit better.
104.
by orasis 4y ago
async/await is FAR less error prone than explicit threaded concurrency.
105.
by orasis 4y ago
I think we've done a pretty good job on the scoring side with a fast and simple to use API that runs in-process: https://improve.ai
106.
by orasis 5y ago
It is now up on pypi: https://pypi.org/project/improveai/ We’ll be updating the docs with examples over the coming weeks. Are there any specific use cases you’re interested in?
107.
by orasis 5y ago
For our initial launch we decided to go with Python and Java. For the time being for other backend frameworks we figure people can fairly easily spin up a Python microservice. We are listening to the requests for other SDKs so thanks for m
108.
by orasis 5y ago
Wow, thank you so much. What a nice thing to say! The inspiration comes from way back in the early 2000s. My team invented HTTP Live Streaming and there were a lot of performance sensitive places in the code where I was essentially making g
109.
by orasis 5y ago
I'd love to learn more about what you're envisioning, you can e-mail me at justin@improve.ai or hit me up on Discord at https://discord.com/invite/mxtdJxfyRk
110.
by orasis 5y ago
I hear you on the AI vs ML thing. I doubt the debate will ever be resolved, but the domain name is .ai so I went with that. The reason to use a full machine learning model instead of a simple statistical model is three-fold: 1) The framewor
111.
by orasis 5y ago
Thanks! For updating variants remotely you can query them from a database or use something like Firebase Remote Config. We've intentionally separated updating variants from the decision of which variant to choose so that decisions don&
112.
by orasis 5y ago
Thanks so much! fixed!
113.
by orasis 5y ago
Hi there, Justin here, the creator of Improve AI. I would love any feedback and feel free to ask me anything. Thanks!
114.
Improve AI – Optimize Your App Easily with AI(improve.ai)
48 points by orasis 5y ago|19 comments
115.
by orasis 5y ago
What license is the MIT license itself released under? Can licenses be copied and modified willy-nilly or do you need to license a license?
116.
by orasis 5y ago
I’ve been caffeine free for 10 years. It has greatly improved my ability to fall asleep and stay sleeping for ~8-9 hours. When I’m underslept my mood is significantly worse. I am one of the most consistently happy people I know. If I need a
117.
by orasis 5y ago
When it comes to code, we’re all in over our heads. I can reason about significantly more complexity than most developers I know and it’s still laughable how little complexity I can manage.
118.
by orasis 5y ago
That’s not really true. Consumers have a fixed willingness to pay and professional app publishers are good at discovering that number, so there really isn’t much passing costs onto the consumer.
119.
by orasis 5y ago
If it is a safer, smoother driving experience it should be legal. The same with drifting across lane lines when appropriate.
120.
by orasis 5y ago
“noise” is integral to all ML systems. You can view this through many lenses, but generalization can be thought of as decoding a noisy signal.

More ›