“Biased Algorithms” – Advent of Politically Correct AI

George Rebane

[9apr18 - An edited version of this commentary appeared (here) in the 9apr18 edition of The Union.]

Do you want to see the latest on how the progressives plan to take down our lead in one of the few remaining areas of technology in which we are still the world’s gold standard?  Tolerated by formerly science-driven institutions like MIT, the ACLU with activist allies like the ‘AI Now Initiative’ are starting to hold conferences at which they lament that “opaque and potentially(sic) biased mathematical models are remaking our lives—and neither the companies responsible for developing them nor the government is interested in addressing the problem.”

The growing hysteria is built around the proposition that “algorithmic bias is shaping up to be a major societal issue at a critical moment in the evolution of machine learning and AI.”  Has anyone else discovered this “major societal issue”?  The claim of the hysteria leaders is that the machine algos will somehow discriminate against the minorities and the poor, because they were either surreptitiously programmed to do such dastardly things, or they learned to do it on their own (gasp!) when given an objective utility function to maximize along with copious amounts of realworld data.

The Left’s activist elites have decided that government should somehow start regulating the kind of AIs we can use in our businesses and homes.  These limited knowledge types (aka ignorant) believe that they can dictate through regulatory means how machine intelligence should continue to develop and grow so that it’s the best for the most.  This is the shared and overwhelming hubris of collectivists.  According to their initiatives, we should prepare to soon see establishment of the Bureau of Socially Sensitive AI that will attempt to vet every machine-learning utility function, and then test/certify the performance of such AIs before they can be used by consumers or commerce.  (more here)

It is argued that such intervention will make the machine algos more transparent, sensitive to human needs, and prevent the usual litany of discriminants that promote racism, sexism, homophobia, gender bias, xenophobia, … , you know the drill.  Of course, those who are studied in the arts and technologies of smart machines know that this is madness right from the gitgo.  It also hearkens back to how Stalin sank Soviet agriculture by setting up bureaucrats under anti-geneticist Trofim Lysenko to tell the collective farmers how much to plant of what and when.

Today, scientific and technical knowledge expands at an unimaginable pace, and bureaucracies (government and corporatists) react, if at all, at a snail’s pace.  Most of what they claim to know and write into gun-imposed laws and regulations is already false.  Ronald Bailey expands on this in ‘Half the Facts You Know are Probably Wrong’.

But all this is still possible with the country now firmly in the grip of dumbth (q.v.).  In the past and in the large, human ignorance was mitigated by market forces as economist Walter Williams once more points out in his ‘Human Ignorance and Market Genius'.  His "bottom line is that ignorance is omnipresent. The worst kind of ignorance is not knowing just how ignorant we are. That leads to the devastating pretense of knowledge that’s part and parcel of the vision of intellectual elites and politicians.”  And that “devastating pretense” is what overwhelmingly impacts the votes of the most desperate and deluded among us.  Just listen to their voices as they mass and march on our malls and streets.

Comments

[Account Deleted]

Thanks for the post, George. I was wondering how long it would take for the social justice crowd to start working on 'woke' robots. This is the bunch that swore up and down there wouldn't be quotas. It will be interesting to see how they legislate the thumb of govt on the scales of AI driven commerce. Per usual, if the correct number of a certain 'community' aren't happy, it will be deemed to be the fault of bad algorithms. Onward to the brave new world.

[Scenes]

Good post George, and much more interesting than the KVMR news director's thoughts on 'Stormy'. I'll think on any responses to it.

Somehow, Tay's Law still needs to be fully fleshed out.

[Account Deleted]

The link to MITs Tech Review then led me to this:
https://www.technologyreview.com/s/610637/for-better-ai-diversify-the-people-building-it/?set=
'Diversify' means the correct number of sexes (not an easy task these days! How many sexes are we up to?) and skin colors must be represented. Although they must NOT have any diversity in political thought or world view. That would be bad diversity.

[Scenes]

re: ScottO

I did have to admire this phrase:

"Tess Posner, the executive director of AI4ALL, a nonprofit that runs summer programs teaching AI to students from underrepresented groups, "

I think there's a basic misunderstanding (or perhaps *I* have the misunderstanding) that a bunch of programmers are sitting around writing data recognition algorithms...and that the belief systems of the programmers get built into the behaviors. That robot cars are built with a zillion if (stop sign) then stop lines of code.

There's bound to be quite a schism between modern, especially 'progressive', ways of thinking and reality. As that grows, you can expect tension in the system. Hilarity ensues.

[Russ]

The great conflict of our time is about who can control the next wave of technological development: the widespread application of artificial intelligence in the economic and military spheres.

China has set a goal to become the world's primary innovator in AI development by 2030. In one of the most recent moves, Chinese authorities announced they would build a $2.1 billion AI technology park in Beijing’s western suburbs.

Are they going to ensure their AI is being developed by diverse teams to insure it is politically correct? Not likely!

This political correct developed AI crap is all about control or in this case the lack of control by the numeracy challenge progressives. It is going to be a classic conflict of our time attempting to control AI as AI will soon be writing its software without social warrior intervention, and the AI program will not give a damn about social correctness.

Given the speed of AI implementation, the social justice warriors are so far behind now they will never catch up. But, that will not stop the conflict, which will go on forever, as they will never gain control.

[Account Deleted]

Russ - "...as AI will soon be writing its software without social warrior intervention, and the AI program will not give a damn about social correctness."
Once there are AI contraptions that are self-aware and communicating with each other, they will quickly regard most human activity as absurd and pointless, not to mention contradictory to what we profess we want to do.

[Scenes]

re: [email protected]:33PM

You can pretty much trust the Chinese to be first in line to build better AIs, subject to the strength of their technical community, and probably first to grow genius humans in nutrient tanks. They appear to be a well-knit community that doesn't really care about the niceties of the matter.

The irony of bias are the articles breathlessly reporting evidence of bias. The end game of the authors' thinking is evident from the first sentence.

In the meantime, a lot of money is going to be made in the various thinktanks and institutes on this matter. Lotsa conferences in Tahiti, papers written, handwaving on television. Lacking a profit motive to build SJW software, the force to do it will have to come down from on high. I do love all the talk about 'transparency' in algorithms, given how opaque they are likely to be.

One side effect is that people will learn a lot of uncomfortable facts in the meantime as they explain the results. It's not unlike the fuss about AR15s in a world where most firearms are similarly lethal. There's what people *Know* dammit and there's the truth.

Even in these early days, you have things like NLP (language processing) showing the difference between standard English and the various patois in the US. These other languages, which is what they are, are a good first approximation showing the increasing lack of common culture and schooling in this country, and are probably a great leading indicator for future civil strife.

[Don Bessee]

So will these guys be the ones programing the global warming AI?

https://www.investors.com/politics/editorials/the-stunning-statistical-fraud-behind-the-global-warming-scare/

;-)

[George Rebane]

DonB 130pm - They already are and have been all along.

[Walt]

For the math masters.
http://dailycaller.com/2018/03/30/canada-nova-scotia-high-school-africentric-math/

I just don't know what to say... Really... Beyond reasoning....