How to tackle the AI inequality

[Music]

what would you do

if i give you one trillion dollar

right now and i guess

buy an island you would still have 990

billion dollars

during the pandemic of 2020 tech

billionaires guy

get richer by around 2 trillion

while 200 million people lost

their jobs but how is this large sum of

money

related to ai

in 2012 the ai wave of deep learning

started when a deep network called

alexnet revolutionized how to tackle

a core problem of image classification

or enabling the computer to identify the

content of an image automatically

this was a massive leap forward for ai

it was followed by

several leaps in the following years in

2015

alphago an ai developed by google

deepmind

defeated the world champion in go a game

like chess

by observing and learning from previous

games

in 2020 alpha fold another ai developed

by google deep mind made a breakthrough

in a 50 year old problem in biology to

predict

protein folding nowadays

the ai that run self-driving vehicles

face filters language to language

translation

even play chess or rely on deep learning

i witnessed this huge development in ai

myself

doing graduate studies focused in deep

learning here at kaust

after four years of hard work that paid

off in 2020 when i received national and

international hours in ai

i realized there is a big problem in the

ai field

there is huge ai inequality

from what i observe it can be

categorized into three main

aspects international industrial

and lack of diversity

on the international aspect we see this

ai inequality manifested in the gap

between countries

in terms of ai investment and ai

development

for example in 2019 the total investment

in ai companies were at 30 billion

dollar in the us

and china and only 5 billion in the rest

of the world

pwc projects that ai will add up to

f up to 15 trillion dollar

by 2030 to the world gdp that’s a lot of

money

but 10 trillion will be to the us and

china as predicted

and 5 trillion to the rest of the world

the same gap is observed in the science

wealth and money created by ai

across all different sectors

the future consequence is concentrated

wealth in these countries that dominate

ai today

the second aspect of ai inequality is

the industrial aspect we see i think

some tech companies that dominate

industries and create ai monopolies

these businesses are largely driven by

deep learning algorithms that are fed

enormous amount of data

the data are used to develop predictive

models

and train they call it in ai train

models and these giant models are used

to enhance digital products

the products are used to collect more

data

and then the data is fed back again to

the models for fine tuning

for better products so this is a vicious

cycle is created

and it is self-reinforcing once it’s

created

it keeps growing

companies that own large amount of data

from the beginning today

are in better position to benefit from

this vicious cycle

i keep calling it vicious cycle and not

virtuous

because ai is a wealth generation

machine

and this ai inequality leads to

economic inequality

and it is known like predicted that it

will add

a lot of wealth to very few who own the

data

and necessary tools for them for example

with its autopilot program and hundreds

of millions of recording hours by its

fleet of cars

the world top electric car company has

an advantage in the race to achieve

self-driving cars

another example is social media

platforms

that know more about us than we do know

about ourselves

and that’s why we see this huge

addiction to social social media

platforms

it fit into this vicious cycle the last

aspect

i would like to discuss today about ai

inequality

is lack of diversity in ai if you look

at the internet content

which is a major source of data used to

develop

ai tools today sixty percent of it is in

english

language compared to one percent in

arabic language

for example in comparison native arabic

speakers account for

five percent of the world population

that means

the data used to train and develop ai

tools

are actually is actually biased not only

that

but also the training and educational

content about these new technologies is

are inaccessible

to a large portion of these less

represented groups

this has a direct impact in both

technology development

and the diversity of the ai workforce

for example

not a single arabic speaking country is

among the top 20 countries

according to nature research index in ai

if you look at industry there are bigger

disparities in the number

with some diversity reports showing that

blacks arabs and other minorities

account for less than five percent

of the ai workforce in some of these big

companies

compared to let’s say 40 of asians and

40 of whites

this is lack of diversity in ai today

and you might ask me you might be asking

yourself

is there something that can be done on

the individual level to tackle this

i’m here to say yes we can do something

at least with this

last aspect in 2019

a group of mind and i started fahim dot

fihim is an arabic word it means

understanding

so ai or understanding ai is an online

platform

that helped the arab community a

disadvantaged group in the ai landscape

to learn and utilize the recent ai

technologies we felt that this arabic

region is deeply disconnected

from the recent technologies in ai and

usually ignored by big tech companies

when it comes to technology development

they are fond of this region when it

comes as a consumer

but not in terms of technology

development

so we started this initiative that

provides

free high quality ai and machine

learning educational content

in all levels in arabic language

the team have pushed through to publish