Wei Lee Woon

Senior data scientist at Expedia

[Jimmy Peng @jimmychpeng]

Can residents in a community #microgrid enjoy greater electricity quotas during blackouts by paying more? Study shows US residents support market-based mechanisms even in life-and-death situations. #EnergyJustice #SocialResilience #ClimateChange Read: https://t.co/OwEyXW2Ymw

[Wei Lee Woon @karipuf]

@HakThye This is how it's done!!🩷 reporting to interpol etc just cedes the high ground. Also, how about responding in kind, like a reaction video... @MrNigelNg maybe?

[Wei Lee Woon @karipuf]

@1stRadiant @BigscienceW Yeah.. getting the same... guess some wider issue, hope resolved soon
@huggingface

[Wei Lee Woon @karipuf]

@DJGagneDos Douglas Adams foresaw this >40 years ago 🤣

[NWS Seattle @NWSSeattle]

An upper low looks like it is probably centered around the Oregon, Idaho and Nevada border triple point. You can see the morning low clouds over Western Washington too. #wawx

[Wei Lee Woon @karipuf]

@_RobToews @ForbesTech .. that threshold (i) it will be hard to spot at first (ii) rudimentary versions will be likely be error prone and glitchy. Has GPT-3 finally cracked it? Don't know, but I'm not ruling it out, and certainly the fact that simpler systems failed to, is not evidence either way (2/2)

[Wei Lee Woon @karipuf]

@_RobToews @ForbesTech But, my other point is that semantics and meaning are emergent properties found when systems reach a certain level of complexity. A single LSTM layer won't have it, and indeed, GPT-2 and current object detection NNs probably don't, either, but when we do eventually cross.. (1/2)

[Wei Lee Woon @karipuf]

@_RobToews @ForbesTech Some might say, that's what we do :-)

[Wei Lee Woon @karipuf]

@_RobToews @ForbesTech LoL not a good look I agree :-) but finding pathological cases like this doesn't mean that the network has no coherent internal representation (e.g. in computer vision there are "one-pixel" attacks where changing a single pixel leads to error. Humans have blind spots, too)

[Wei Lee Woon @karipuf]

yup.. https://t.co/p1IG7ryPxE #MachineLearning #explainableAI

[Wei Lee Woon @karipuf]

It's interesting that Monopoly, the board game that in many ways epitomizes capitalism, includes a form of UBI ($200 every time you pass go!) - strangely this "free money" doesn't make the players lazy, or any less competitive :-) #UBI @AndrewYang @HumanityForward

[Wei Lee Woon @karipuf]

@syaffers @alhanaei I think there's something deeper going on here... it's not just *tolerated* that humans are black boxes, it's actually *celebrated*. With AI, while people may eventually accept that some decisions simply can't be explained, I doubt they will celebrate it in the same way

[Wei Lee Woon @karipuf]

When humans make instinctive decisions, the reception is often positive, e.g. "finding the answer deep down", or "flash of inspiration".. but when it's an AI, people complain of "black boxes", or that "explainable AI" is needed, etc. Double standards? #ExplainableAI #DeepLearning

[Wei Lee Woon @karipuf]

@comcastcares hi comcast/xfinity - I am unable to connect - the modem is flashing orange/red - I have tried tightening cables/restarting etc, and nothing works, and there are no reported outages. I just can't through to a human on the chat line. Can I please get some help?

[Chris Bail @chris_bail]

1/n Can machine learning solve social science problems? This unprecedented study led by @msalganik and dozens of co-authors is a #kaggle-style competition to predict outcomes of children from vulnerable families using held out data from a multiwave study: https://t.co/0rctz0YFYG

[Wei Lee Woon @karipuf]

If someone says: "it's just a couple lines of code", my response is: coding is NP - given valid code, I can quickly check if it works. What takes time isn't typing 1 line of code - it's searching the universe of >100000's of possible lines of code, and finding the correct one.

[Wei Lee Woon @karipuf]

@alhanaei @brilliantorg Lol 😋

[Wei Lee Woon @karipuf]

@alhanaei @brilliantorg But only integers allowed 🙁