AI Newsletter

ISSUE NO. 1 | MAY 2023

Business Ethics
The CEO’s Guide EU AI Act to Target
to the Generative Open Source
AI Revolution

Research Learning Hub

DragGAN: ChatGPT Prompt
Interactive GAN Engineering for
Manipulation on Developers
Image Manifold

In The Open Public Figure

StarCoder: Prof. Geoffrey Hinton
The Ultimate Code
LLM Powerhouse

MoroccoAI Conference Videos

Towards a Promising National
AI Strategy for Morocco
Opening Letter
Greetings MoroccoAI Community,

We are delighted to present this new edition of MoroccoAI Newsletter,

a concise and captivating source of news summaries covering the
forefront of the AI landscape. The newsletter is designed to keep you
well-informed and up-to-date with the most significant advancements
in the field, catering to AI professionals, researchers, and students.

Our team has curated a collection of noteworthy news pieces, carefully

selected for their credibility and value. From groundbreaking research
breakthroughs to thought-provoking reports on AI ethics and impactful
business updates, the newsletter covers a diverse range of topics that we
hope will captivate your interest and deepen your understanding.

MoroccoAI is a non-profit organization committed to fostering AI

education and excellence in research in Morocco. We achieve this by
promoting dynamic collaborations within our community and strategic
partnerships with renowned institutions and industries. MoroccoAI
regularly hosts a variety of engaging events, including virtual and in-
person, including AI Webinars, Meetups, Workshops, and Conferences.
We are immensely grateful for the support of our incredible
community, which has grown to over 20,000 subscribers across our
MoroccoAI team

social media platforms. It is your enthusiasm and dedication that fuel

our drive to lead the future of AI in Morocco and beyond.

We invite you to follow us on Linkedin to stay connected and engaged.

Share this newsletter with your peers, colleagues, and anyone passionate
about the exciting world of AI.

Enjoy your read!

2
Content
Opening Letter
Ethics &
02 Responsible AI

11
AI Around
The World Learning Hub

12

Business
Public Figure
05
13

Research
MoroccoAI
07
Highlights
In the Open
MoroccoAI
09
Conference

15
Business
Anthropic introducing 100K context windows for Claude
Anthropic, founded by ex-OpenAI employees,
expanded Claude’s context window from 9K to
100K tokens, corresponding to around 75,000
words. Businesses can now analyze and retrieve
information from hundreds of pages with Claude,
and conversations can go on for hours or even
days. The 100K context window feature is
accessible via Claude's API.
read more Anthropic - May 11th

Microsoft reinventing productivity with Copilot

Microsoft is integrating a new AI-
powered Copilot for its Microsoft 365
apps and services, designed to help
organizations drive business
outcomes, improve operational
efficiency, and create exceptional
customer experiences.
read more Microsoft - May 9th

PwC $1 billion investment to scale its AI capabilities

PwC to invest $1 billion over the next three
years to expand its AI offerings. This follows a
new strategic alliance with an AI Startup,
called Harvey, to build on PwC’s ability to
bring human-led and tech-enabled solutions
to clients. In tandem, PwC US will invest in
upskilling its 65,000 people on AI tools and
capabilities.
read more PwC - April 26, 2023

5
 Business

OCP and UM6P invest $200 Million in Bidra Ventures

Bidra Innovation Ventures announced
commitment of $200M by UM6P and OCP
Group to expand its investment scope to
include innovations in energy, water, mining
and agriculture for Africa and beyond. The
VC firm is an active investor in AI and
Robotics agricultural startups like Niqo
Robotics, which focuses on AI-powered
sustainable spot spraying technology.
read more Business Wire - April 5th

300 million jobs could be lost globally to AI automation

Goldman Sachs predicts that AI has the potential
to significantly disrupt labor markets, exposing
300 million jobs globally to automation, but it
can also boost labor productivity over the
coming decades, leading to substantial
economic growth.
read more Goldman Sachs - March 26th

The CEO’s guide to the generative AI revolution

Immersed in the transformative era of the
Generative AI revolution, CEOs should
identify their company's "golden" use cases,
which could offer genuine competitive
advantage. BCG crafted a report laying out
strategic options for CEOs keen on harnessing
the potential of generative AI to unlock
hidden value and boost efficiency across
their enterprises.
read more BCG - March 7th

6
Research
Interactive GAN manipulation on image manifold
DragGAN controls GANs in image synthesis. It
allows to interactively manipulate images by
"dragging" points to specific positions. It
provides precise control over pose, shape,
expression, and layout of objects, resulting in
realistic outputs. DragGAN outperforms
previous approaches in image manipulation
and point tracking tasks. It can even
manipulate real images through GAN inversion.
read more Hugging Face - May. 18th

Google announced PaLM 2, its next-gen LLM

Google announced its latest large language model
PaLM 2 with improved multilingual, reasoning and
coding capabilities. The system is available in four
sizes and has been fine-tuned on domain-specific
data to perform certain tasks for enterprise
customers. It is already being used to power Google
Workspace apps and products, including Bard.
read more Google- May 10th

Meta AI model understands multi-sensory content

ImageBind uses images to connect different
modalities and learn a single joint embedding
space without needing paired data for every
combination of modalities. The approach brings
machines one step closer to humans’ ability to
learn simultaneously, holistically, and directly
from many different forms of information,
without the need for explicit supervision.
read more Meta AI - May. 9th

7
 Research

OP3-Soccer: Autonomous soccer playing robotic system

Deepmind presents OP3-Soccer, a robotic
system designed for autonomous soccer
playing. The system combines computer
vision, motion planning, and control
algorithms to enable the robot to perceive
the environment, make decisions, and
execute complex actions on the soccer field.
read more Deepmind - April. 26th

Computer vision models with self-supervised learning

Meta AI open-sourced DINOv2, the first
self-supervised learning method to build
high-performance computer vision
models. It can be applied to a wide
variety of applications and capable of
achieving high-quality visual features
without the need for fine-tuning.
read more Meta AI - April. 17th

Segment Anything: Advancing instance segmentation

Segment Anything Model (SAM) can "cut out"
any object, in any image, with a single click.
SAM can be used to generate high quality
masks for all objects in an image. It is a
promptable segmentation system with zero-
shot generalization to unfamiliar objects and
images. Authors released their open source
dataset of 11MM images and over 1 billion
masks, the largest mask corpus to date.
read more Meta AI - April. 5th

8
In The Open
StarCoder: The ultimate code LLM powerhouse
StarCoder is a powerful LMM for Code trained
on more that 80 different programming
languages as well as text extracted from
GitHub issues and commits and from
notebooks. It outperforms other open Code
LLMs on several programming benchmarks.
Designed to assist programmers, StarCoder
facilitates the creation of high-quality,
efficient code in shorter time periods.
read more Huggingface - May. 4th

Effortless fine-tuning of LLMs with H2O LLM Studio

Large Language Models (LLMs) have
revolutionized NLP, but fine-tuning them
for specific tasks can be complex. H2O
LLM Studio simplifies the process with a
user-friendly GUI, advanced metrics, and
easy model sharing, empowering NLP
practitioners without coding experience
to customize models effortlessly.

read more H2O.ai - May. 1st

Mojo — a new programming language

A new programming language has emerged. It
tackles the speed limitations of Python, offering
unparalleled performance that can reshape the
AI field. Mojo enables developers to write fast,
small and easily-deployed applications that take
advantage of all available cores and
accelerators. It was designed with full
compatibility with Python ecosystem.
read more Modular - April 22th

9
 In The Open

Auto-GPT: The rise of autonomous agents

Auto-GPT is a tool that transforms ChatGPT into an
autonomous AI assistant, handling tasks and
projects without need for user prompts. It is able
to carry out more complex multi-step procedures
than existing LLM-powered applications by
creating its own prompts and feeding them back
to itself, creating a loop. Auto-GPT is capable of
browsing the internet and incorporating relevant
information into its computations and results.
read more Github - April 22th

Stability AI reveals StableLM and SDXL

Stability AI introduced two innovative AI models,
StableLM, which purports to achieve similar
performance to OpenAI's benchmark GPT-3 model
while using far fewer parameters, and Stable
Diffusion XL, its latest text-to-image. Stable
Diffusion kickstarted an era of rapid development
in image-synthesis technology, and its move into
language models could inspire similar results.

read more Stability.ai - April. 19th

Open Assistant: Empowering language innovation for all

OpenAssistant is a chat-based assistant that
understands tasks, can interact with third-party
systems, and retrieve information dynamically.
In an effort to democratize research on large-
scale alignment, they released a
conversational dataset comprising 160K
messages in 35 languages, a remarkable
worldwide crowd-sourcing effort.
read more Github - April 15th

10
Ethics & Responsible AI
EU AI Act to target Open-Source
The EU's Amended AI Act aims to regulate the
use and distribution of AI systems. The Act
mandates extensive licensing requirements and
expensive risk testing on AI projects. Open-
source developers and software distributors
could face substantial financial penalties if
unlicensed generative models become
accessible in Europe. Critics argue that the Act
could encourage unsafe AI practices.
read more europarl - May. 16th

A Policymaker’s Guide to Foundation Models

This IBM article advises against overestimating
the capabilities of foundation models,
spotlighting the inherent risks tied to training
data, outputs, and governance. It underscores the
diverse potential for harm that can arise from
various AI applications. The article suggests that
policymakers can best address these concerns by
implementing risk-based regulatory obligations
and an AI policy framework tailored towards
those deploying AI systems.
read more IBM - May. 1st

How can we build human values into AI?

There is no consensus on a single set of human
values or preferences to govern AI. DeepMind
research proposes that the impartial nature of
the Veil of Ignorance may provide a starting
point for the selection of principles with which
to align AI with our diverse human values and
opinions.
read more ScienceDirect- April. 24th

11
Learning Hub
Stanford Institute released its 2023 AI Index Report
The AI Index tracks and evaluates AI
progress through a wide range of
perspectives, looking at trends in
research and development, technical
performance, ethics, economics, policy,
public opinion, and education. This year’s
report included new analysis on
foundation models and noted how AI
Learn More industry is advancing over academia.

Practical Deep Learning for Coders at Fast.ai

Fast.ai is an online platform and community
that offers practical, coding-focused courses
in deep learning. The courses are designed for
beginners while also offering advanced
material for experienced learners. Its emphasis
on hands-on, applied learning equips students
with the skills and knowledge to undertake
real-world projects.
Learn More

ChatGPT Prompt Engineering for Developers

A new free course on Prompt Engineering is
offered by Deeplearning.ai in collaboration
with OpenAI. It outlines best practices in
prompt engineering, which is the art of
skillfully communicating with an LLM. The
course gives participants hands-on practice
with the OpenAI API and it is designed to be
accessible to everyone with a fundamental
Learn More grasp of Python.

12
Public Figure
Prof. Geoffrey Everest Hinton
Geoffrey Hinton, a distinguished British-
Canadian cognitive psychologist and computer
scientist, is an Emeritus Professor at the
university of Toronto. He has been honored
with numerous accolades, including the
prestigious ACM Turing Award in 2018, widely
regarded as the Nobel Prize of computing.

Hinton is recognized as the Godfather of Deep

Learning for his pioneering work on algorithms
that enable artificial Neural Networks (NNs) to
learn distributed representations. In 1980s,
Hinton demonstrated the immense potential of
backpropagation, a central mechanism by At the latest MoroccoAI
which NNs learn and forms the bedrock of this conference, Hinton presented
technology today. Hinton has made significant a new learning algorithm for
other contributions to the field which have neural networks called the
paved the way for current AI systems. "forward-forward algorithm",
inspired by our knowledge of
neural activations in the brain.

Earlier this month, Hinton

resigned from Google to
openly discuss the risks
associated with AI. Since then,
he has publicly addressed the
potential of AI to both benefit
and harm society, He also
cautioned that autonomous AI
systems may pose a serious
threat to humanity and that
the kind of intelligence we're
developing is very different
from our intelligence.

13
Highlights
Conference Highlights

AI should not be restricted

to academic elite.
It should be accessible to all
citizens
H.E. | Dr. Aawatif Hayar

WATCH NOW

Al is changing the way

Research is about effort..
humans behave and we would
It is no coincidence. You have
like this change to happen in
to experiment. You have to
the good way for what we
work hard
value in our society

Prof. Rachid Guerraoui Prof. Malik Ghallab

WATCH NOW WATCH NOW

Data scientists and factory leads

We often overlook the
have different backgrounds.
crucial role of computing
Developing a common language
is key to succeed in platforms in deploying
building an Al bridge Al at large and at edge

Dr. Hicham Zmarrou Prof. Said Hamdioui

WATCH NOW WATCH NOW

15
 Conference Highlights

Once you get close enough..

suddenly, all sorts of things
start making sense

Prof. Geoffrey Hinton

WATCH NOW

In the future, AI will be distributed, Wherever there is data being

and we will need a network that collected, there lies a significant
interconnects all the resulting opportunity for AI to address various
systems challenges and issues

Prof. Merouane Debbah Dr. Bouchra Bouqata

WATCH NOW WATCH NOW

When developing AI strategies, it is AI development in Morocco

important to take into account both cannot be dissociated from
the potential of positive and a global strategy on
negative climate impacts of AI science and technology

Dr. Priya L. Donti Prof. Mohamed Najim

WATCH NOW
WATCH NOW

16
 CURRENT AVAILABLE OPENINGS
AT MOROCCOAI

Join Us Today!

Social Media &

Community Engagement

AI Newsletter

Creative Design

Website Development

Study Groups &

Journal
Study Club& Journal Club
Groups

Apply now !
 AI Meetup

Webinars Conferences

Promoting excellence in AI education & research

and connecting today's and tomorrow's AI leaders

ISSUE NO. 1 | MAY 2023