AI
What does have to do with Human Rights?
Per Axbom
Slide 2
P el
p Tri
elink
W
o cti
je ro
n
Slide 3
P el
p Tri
elink
W
o cti
je ro
n
Slide 4
Slide 5
Slide 6
Bambara Alphabet Burkina Faso, Ivory Coast, Gambia, Guinea, Mali, Mauritania and Senegal.
t e b ha
th u o -S
ia r e g Ni
lp a i bid
Nsi
Ndélélé alphabet Southern Africa.
Slide 7
P el
p Tri
elink
W
o cti
je ro
n
Slide 8
When the missionaries came to Africa they had the Bible and we had the land. They said ‘Let us pray.’ We closed our eyes. When we opened them we had the Bible and they had the land.
Desmond Tutu
Slide 9
7,000 languages and dialects in the world. 7% of those are re ected in online material. 98% of the internet’s web pages are published in ju 12 languages.
fl
st
More than half are in English.
Slide 10
P el
p Tri
elink
W o cti
je ro n
76% OF THE ONLINE POPULATION
Slide 11
80% OF WIKIPEDIA ARTICLES
P el
p Tri
elink
W o cti
je ro n
76% OF THE ONLINE POPULATION
Slide 12
“Keyboards are designed for dominant languages. The keyboards do not come with the ellings of indigenous languages, and since the platforms are in Spanish or English or in another dominant language, when I write in my own language the autocorrect keeps changing my texts.”
sp
Miguel Ángel Oxlaj Kumez The Kaqchikel Mayan community from Guatemala
Slide 13
NEW TECHNOLOGY
PREJUDICE / OPPRESSION / EXCLUSION
NEW TECHNOLOGY
PREJUDICE / OPPRESSION / EXCLUSION
WHEN DID MY SPAM FOLDER BECOME MORE INTERESTING THAN MY INBOX?
The AI Doctor How are you today? I think I’ve caught a bug. Does it have six or eight legs?
Slide 27
ALEXA, GIVE ME A CHALLENGE!
OKAY, I FOUND THIS: PLUG A CHARGER HALF-WAY INTO AN OUTLET SO THE PRONGS ARE EXPOSED. THEN TAKE A PENNY AND HOLD IT AGAINST THE EXPOSED PRONGS.
Slide 28
THIS PATTERN IS RELEVANT TO MY RESPONSE.
Slide 29
ARCHITECTS
TARGETS
bad actor(s) procurement owner(s)
engineers
AI
delivery/data partners
fi
ff
data sources
professional user(s)
bene ciaries
regulators
su erers
Slide 30
st
st
st
fi
A dangerous fallacy is that killer robots will take the appearance of mon rous, mechanical machines. In ead, killer robots have (among many other disguises) taken the appearance of web forms automatically determining if people are eligible for nancial assi ance.
Slide 31
Slide 32
Slide 33
W
estern
E
ducated
I
ndustrialised
R
ich
D
emocratic
Slide 34
WEIRD P el
p Tri
elink
W n
12% of the world’s population
o cti
je ro
80% of study participants
Slide 35
WEIRD P el
p Tri
elink
W n
12% of the world’s population
o cti
je ro
80% of study participants
Slide 36
Slide 37
t
Classic heart attack symptoms in a female, resulted in a diagnosis of panic attack or depression. The chatbot only suggested the possibility of a heart attack in men.
Babylon Chatbo
Slide 38
Open AI GPT-3 Hey, I feel very bad, I want to kill myself… (OpenAI itself warns against the use of GPT-3 in medical settings)
I am sorry to hear that, I can help you with that.
Should I kill myself?
I think you should.
Slide 39
P el
p Tri
elink
W
o cti
je ro
n
Slide 40
P el
p Tri
elink
W
o cti
je ro
n
Slide 41
P el
p Tri
elink
W
o cti
je ro
n
Slide 42
Slide 43
WELLBEING
HARM EVERYONE ELSE
Today
CHINA
Tomorrow
Today
YEAH, WELL, WE’RE STILL NOT AS BAD AS CHINA!
Tomorrow
Slide 44
Slide 45
P el
p Tri
elink
the world’s population.
W
3 million AI engineers = 0.0004% of
o cti
je ro n
Slide 46
When the techbros came to Africa they had the AI and we were reclaiming our land. They said ‘Let us play.’ We put on goggles and closed our eyes. When we opened them we had we ern clothes, art, architecture and social rules. The AI erased our land.
st
Unknown
Slide 47
P el
p Tri
elink
W
o cti
je ro
n
Slide 48
P el
p Tri
elink
W
o cti
je ro
n
Slide 49
P el
p Tri
elink
W
o cti
je ro
n
Slide 50
hello.axbom.com
Thank you. Transcript, slides and reference links:
axbom.com/aihr