, 22 tweets, 58 min read
The @StanfordHAI (Human-centered #AI) Conference on #AiEthics, Policy, & Governance is getting started, live stream below. Speaker list is fantastic. Will be tweeting.

#machinelearning #data4good
@StanfordHAI "Silicon Valley, we have a problem" opens @erikbryn

Speaks to the 'great decoupling' - where productivity and GDP expand with technology, but don't take household incomes with them.

Part of the problem is lower-paid occupations have more automate-able tasks.
@StanfordHAI @erikbryn Both @Susan_Athey and @erikbryn speaking to ML adoption in private sector and the important distinction between labor replacing vs. labor augmenting #automation.

The "Suitability for Machine Learning" or SML scores from this
@MIT_IDE is useful here: ide.mit.edu/sites/default/… #ai
@StanfordHAI @erikbryn @Susan_Athey @mit_ide Moderator @kendenman asks about the #technology talent pipeline - there's a good recent report on this problem from @CSETGeorgetown: cset.georgetown.edu/wp-content/upl…
@StanfordHAI @erikbryn @Susan_Athey @mit_ide @kendenman @CSETGeorgetown Now onto regulating big tech - @ericschmidt starts by noting upcoming release of reports from (1) DoD on #AiEthics and (2) National Security Commission on Artificial Intelligence (Congressionally formed commission on national #AI strategy)

#StanfordHumanAI #TechPolicy
@StanfordHAI @erikbryn @Susan_Athey @mit_ide @kendenman @CSETGeorgetown @ericschmidt This is the Defense Innovation Public Board Meeting that @ericschmidt is referring to, at @georgetownsfs:

It's "an independent federal committee advising the defense secretary on issues that focus on people and culture, technology and capabilities, practices & operations"
@StanfordHAI @erikbryn @Susan_Athey @mit_ide @kendenman @CSETGeorgetown @ericschmidt @georgetownsfs "If we had imagenet for medical data, we'd see tremendous progress in curing diseases" - paraphrasing @ericschmidt (he does note the #privacy issues).

This... is probably true. I need to read the proposed CMS data interoperability standards: edit.cms.gov/newsroom/fact-… #AiEthics
@StanfordHAI @erikbryn @Susan_Athey @mit_ide @kendenman @CSETGeorgetown @ericschmidt @georgetownsfs "You don't have to yell that there is bias in the data, there is always bias in the data, the real question is what do we do about it" -@ericschmidt

I get his point but this is slightly divorced from the reality - we still need to yell about bias.

@StanfordHAI @erikbryn @Susan_Athey @mit_ide @kendenman @CSETGeorgetown @ericschmidt @georgetownsfs On #deepfakes: "Fake video is a hard problem at social level and technical level... this is a battle that the researchers need to win."

I am sadly skeptical that this is a winnable battle for algorithms, but glad it came up. #disinfo
@StanfordHAI @erikbryn @Susan_Athey @mit_ide @kendenman @CSETGeorgetown @ericschmidt @georgetownsfs "I start from the premise that this technology is extraordinarily beneficial" and "We are just starting to see the capacity of Ai in #science" -@ericschmidt has cited applications in chemistry & protein folding.

IMO, AI+sciences does seem like something to be optimistic about.
@StanfordHAI @erikbryn @Susan_Athey @mit_ide @kendenman @CSETGeorgetown @ericschmidt @georgetownsfs Now @MarietjeSchaake, Member of European Parliament and International Policy Director of @Stanford_Cyber:

"We must consider what problems should be public, and not in a market."

#TechPolicy #StanfordHumanAI
@StanfordHAI @erikbryn @Susan_Athey @mit_ide @kendenman @CSETGeorgetown @ericschmidt @georgetownsfs @MarietjeSchaake @Stanford_Cyber A much more somber speech, @MarietjeSchaake asks

If AI will make undemocratic/authoritative governments more powerful, then why do we have such high expectations of how this technology will benefit us all?
@StanfordHAI @erikbryn @Susan_Athey @mit_ide @kendenman @CSETGeorgetown @ericschmidt @georgetownsfs @MarietjeSchaake @Stanford_Cyber "I heard Zuckerberg warning capitol hill that #China is the alternative to the Libra cryptocurrency, but Facebook has data sharing agreements with four Chinese companies" -@MarietjeSchaake

She has heard many times from #tech's lobbyists: "Do not regulate us, otherwise China..."
@StanfordHAI @erikbryn @Susan_Athey @mit_ide @kendenman @CSETGeorgetown @ericschmidt @georgetownsfs @MarietjeSchaake @Stanford_Cyber "But inaction to regulate by democracies has not slowed China..."

👏 "We need more institutions and processes, more than just good intentions or 'scouts honor' from #bigtech" 👏

"I don't know if 'do no evil' is Google's motto anymore..."

@StanfordHAI @erikbryn @Susan_Athey @mit_ide @kendenman @CSETGeorgetown @ericschmidt @georgetownsfs @MarietjeSchaake @Stanford_Cyber "Digitization means privatization - the outsourcing of governance to tech companies." - @MarietjeSchaake

"The shift to private & opaque governance through tech standards is one of the most important consequences of #AI."

Everyone should watch this speech. #StanfordHumanAI
@StanfordHAI @erikbryn @Susan_Athey @mit_ide @kendenman @CSETGeorgetown @ericschmidt @georgetownsfs @MarietjeSchaake @Stanford_Cyber .@math_rachel makes this v. important point more clear than I did, so to +1:
-@jovialjoy up next
-@GoogleAI researchers just put out a paper on #AI bias against people w/ disabilities
-new @nature article on racial bias in health ML affecting 200m people

@StanfordHAI @erikbryn @Susan_Athey @mit_ide @kendenman @CSETGeorgetown @ericschmidt @georgetownsfs @MarietjeSchaake @Stanford_Cyber @math_rachel @jovialjoy @GoogleAI @nature "Data is destiny" and the facial datasets on which so much new #ai relies are mostly "pale and male" argues @jovialjoy.

Cites new @IBMWatson / @IBMDataScience paper, I think this: arxiv.org/pdf/1901.10436…

I'm genuinely curious if @ericschmidt is still there.
@StanfordHAI @erikbryn @Susan_Athey @mit_ide @kendenman @CSETGeorgetown @ericschmidt @georgetownsfs @MarietjeSchaake @Stanford_Cyber @math_rachel @jovialjoy @GoogleAI @nature @IBMWatson @IBMDataScience Also relevant, evidence of "substantial skin type and gender bias in #AI services from companies like Microsoft, IBM, and Amazon."

From @jovialjoy's testimony to House Science Committee: science.house.gov/imo/media/doc/… #StanfordHumanAI #DeepLearning #FacialRecognition
@StanfordHAI @erikbryn @Susan_Athey @mit_ide @kendenman @CSETGeorgetown @ericschmidt @georgetownsfs @MarietjeSchaake @Stanford_Cyber @math_rachel @jovialjoy @GoogleAI @nature @IBMWatson @IBMDataScience For ~binary~ gender classification, in a ~~commercially~~ sold model, the error rates were nearly _50%_ for people w/ "dark brown to black skin, brownish black eyes, naturally black hair."

This from @jovialjoy's PhD dissertation, 'Gender Shades' proceedings.mlr.press/v81/buolamwini… #aiethics
@StanfordHAI @erikbryn @Susan_Athey @mit_ide @kendenman @CSETGeorgetown @ericschmidt @georgetownsfs @MarietjeSchaake @Stanford_Cyber @math_rachel @jovialjoy @GoogleAI @nature @IBMWatson @IBMDataScience Joy says the original research was well received, and that companies released far better performing models.

This is a good argument for some form of independent #machinelearning auditing - the models certainly can be improved if there is pressure to do so.
Missing some Tweet in this thread? You can try to force a refresh.

Enjoying this thread?

Keep Current with Alex Engler

Profile picture

Stay in touch and get notified when new unrolls are available from this author!

Read all threads

This Thread may be Removed Anytime!

Twitter may remove this content at anytime, convert it as a PDF, save and print for later use!

Try unrolling a thread yourself!

how to unroll video

1) Follow Thread Reader App on Twitter so you can easily mention us!

2) Go to a Twitter thread (series of Tweets by the same owner) and mention us with a keyword "unroll" @threadreaderapp unroll

You can practice here first or read more on our help page!

Follow Us on Twitter!

Did Thread Reader help you today?

Support us! We are indie developers!

This site is made by just three indie developers on a laptop doing marketing, support and development! Read more about the story.

Become a Premium Member ($3.00/month or $30.00/year) and get exclusive features!

Become Premium

Too expensive? Make a small donation by buying us coffee ($5) or help with server cost ($10)

Donate via Paypal Become our Patreon

Thank you for your support!