This presentation give an introduction to Artificial Intelligence subjectiveness and history. The primary goal of the presentation is to provide a deep enough understanding of Artificial Narrow Intelligence and Artificial General Intelligence so that the people can appreciate the strengths or weaknesses of the AI. The presentation also includes a classification(the main domains of AI) and the most relevant examples from the past decades. In the second part it provides some statistics and future possible applications and forecasts.
3. Today is all about AI
Petru Cioată
Ioana Ianovici
4. Petru Cioată
Software Development Engineer at ASSIST Software
Ioana Ianovici
Software Development Engineer at ASSIST Software
ioana.ianovici@assist.ro
petru.cioata@assist.ro
5. Contents
● Subjectiveness of AI
● Definition
● History of AI
● The Turing Test
● ANI vs. AGI
● Classification
● Fun stories
● Future
● Conclusions
10. ◦ Cool things that computers can’t do
◦ A simulation of human intelligence
◦ Ability to adapt to new environments based on experience
and learning
Definition of AI
15. “
Dartmouth Conference thoughts
“Machines will be capable,
within twenty years, of doing
any work what man can do.”
Herbert A. Simon (CMU)
“Within a generation … the
problem of creating ‘artificer
intelligence’ will substantially be
solved”
Marvin Minsky (MIT)
30. Eugene Goostman
JUDGE: Hello
EUGENE: Hello, I’m really glad to have the chance to chat with you! My guinea pig Bill sends his regards too!
JUDGE: Is Bill a male or a female?
EUGENE: Ask Bill personally, please.
JUDGE: Well I’d rather talk to you. What is your name?
EUGENE: Call me Eugene. I am glad to talk to you!
JUDGE: My name is Jane and I am female. How about you? What’s your gender?
EUGENE: I’m a male. A “guy” I’d say.
JUDGE: Pleased to meet you Eugene. What’s the weather like where you are?
EUGENE: Let’s get on with our conversation!
JUDGE: Don’t you like talking about the weather?
EUGENE: All these talks about weather is a waste of time.
JUDGE: What would you like to discuss?
EUGENE: I don’t know… Better tell me more about yourself!
33. Narrow (Weak) AI
◦ Specific domain application
◦ Solve a specific(narrow) problem
◦ Practical(very pointed challenges)
◦ Certain degree of intelligence in a
particular field
◦ Great at optimizing specific tasks
◦ Learns to reduce error output
◦ Remains a computer system that
performs highly specialised tasks
34. Benefits of ANI
◦ Process data & complete tasks - significantly quicker than humans
◦ Improve humans productivity, efficiency & quality of life
◦ Relieve us from a lot of the routine tasks
◦ Relieve us of frustrating realities
35. Raymond Kurzweil’s predictions
2029 - AI systems will pass Turing test
Achieve humans-level intelligence
2045 - technological Singularity will appear
Humans able to connect their neocortex to some
form of storage system (cloud-based/DNA-connected?)
Merge this neocortex with AI-driven amplifiers
36. “
From ANI to AGI
“We’re slowly building a library of narrow AI talents that are becoming more
impressive. Speech recognition and processing allows computers to convert
sounds to text with greater accuracy.
Google is using AI to caption millions of videos on YouTube. Likewise, computer
vision is improving so that programs like Vitamin D Video can recognize objects,
classify them, and understand how they move. Narrow AI isn’t just getting better
at processing its environment it’s also understanding the difference between
what a human says and what a human wants.”
Aaron Saenz - writer for Singularity Hub
37. ◦ General domain application
◦ Solve multiple problems
◦ Human-level AI
◦ Handle tasks from multiple domains
◦ Learns new tasks across several domains
◦ Adapt to changing environments
◦ Applies experience gathered in one area to a different area
◦ Needs a semantic connection between areas
General AI
38. General AI - Blue Brain Project
The human brain has
100 billion neurons and
1000 trillion synaptic
interconnections
10.000 neurons and
30 million interconnections
from a mammalian brain
(the Blue Brain Project)
https://bluebrain.epfl.ch/
39. General AI - Sophia
Sophia’s Main Components:
● A timeline editor
● A “sophisticated chat-bot”
● OpenCog
41. General AI Threats
◦ Social attacks on high-profile public platforms (ranging from identity theft
up to alleged meddling with elections)
◦ The Cambridge Analytica case and recent US elections
◦ Totalitarian control threat - the “City Brain” project in Hangzhou, China
44. Machine Learning
The field of Machine Learning is built upon the concept of Learning, which is
believed to be central to the notion of Intelligence. It describes systems that
improve their performance in a given task with more and more experience or data.
Machine learning is mainly divided into:
◦ Supervised
◦ Unsupervised
◦ Reinforcement
45. ML. Supervised Learning
Input and output
pairs
The system is fed with
input-output pairs (also called
labeled input)
The result function
When a certain number of iterations were
done and the algorithm seems to map
correspondingly the given inputs to
outputs, the resulted function is considered
to be the result - the knowledge that the
system learned. It can be applied now to
new data, unlabelled to predict unknown
outputs.
Data analysis
The system analyzes the provided labelled
data, trying to find a corresponding function to
map the given inputs to the given outputs. This
step is usually achieved through multiple
algorithmic iterations.
03
01 02
Usages: Handwriting recognition, Learning to rank, Object
recognition in computer vision, Optical character recognition,
Spam detection, Pattern recognition, Speech recognition and
many more.
46. ML. Unsupervised learning
The system learns from test data that has not been labeled, classified or categorized.
Unsupervised learning identifies similarities in the data and classifies it based on the
presence or absence of such details.
Unsupervised learning can also be used to create classes on top of which one can
provide the labelled data for supervised models.
It is largely used for density estimation in statistics, summarizing and explaining data
features.
47. ML. Reinforcement learning.
The system is retro-feeding it’s model in order to improve, based on a definition of
“reward”. It tries to maximize the rewards, without being said how.
The focus is on performance, which involves finding a balance between exploration
(of uncharted territory) and exploitation (of current knowledge).
48. Deep Learning
Deep Learning is a subfield of Machine Learning.
It structures algorithms in layers to create an artificial neural network (ANN) that
can learn and make intelligent decisions on its own.
Deep Learning methods are used for fields such as computer vision, speech
recognition, natural language processing, audio recognition, social network filtering,
machine translation, bioinformatics, drug design, board game programs, etc.
49. Data Science
Data Science is an umbrella term that includes:
◦ ML
◦ Statistics
◦ Certain aspects of computer science
▫ Algorithms
▫ data storage
▫ web application development
It is aimed to understand and analyze actual phenomena with data
50. Robotics
Building and programming robots so that they can operate in real-world scenarios.
Robotics is the ultimate challenge of AI since it requires a combination of all AI
areas.
For example:
◦ Computer vision and speech recognition for sensing the environment
◦ Natural language processing, information retrieval, and reasoning under
uncertainty for processing instructions and predicting consequences of
potential actions
◦ Cognitive modeling and affective computing for interacting and working
together with humans
55. Tay was an artificial intelligence chatter bot that was
originally released by Microsoft Corporation via
Twitter on March 23, 2016.
It caused subsequent controversy when the bot
began to post offensive, racist and sexually-charged
messages in response to other Twitter users, forcing
Microsoft to shut down the service only 16 hours after
its launch.
Tay
58. HSBC voice ID fooled by twin
HSBC’s voice recognition ID
system used by half a million
customers for secure access
to their bank accounts has
been breached by a
customer’s twin mimicking his
voice.
59. Alexa orders dollhouses
“can you play dollhouse with me and get me a dollhouse?”
Ordered a $160 KidKraft Sparkle mansion dollhouse and four pounds of sugar cookies
60. Facebook chatbots shut down after developing
their own language
Bob: i can i i everything else . . . . . . . . . . . . . .
Alice: balls have zero to me to me to me to me to me to me to me to me to
Bob: you i everything else . . . . . . . . . . . . . .
Alice: balls have a ball to me to me to me to me to me to me to me
Bob: i i can i i i everything else . . . . . . . . . . . . . .
Alice: balls have a ball to me to me to me to me to me to me to me
Bob: i . . . . . . . . . . . . . . . . . . .
Alice: balls have zero to me to me to me to me to me to me to me to me to
Bob: you i i i i i everything else . . . . . . . . . . . . . .
Alice: balls have 0 to me to me to me to me to me to me to me to me to
Bob: you i i i everything else . . . . . . . . . . . . . .
Alice: balls have zero to me to me to me to me to me to me to me to me to
63. Google Translate shows gender bias
Gender by Google Translate
he is a soldier
she’s a teacher
he is a doctor
she is a nurse
he is a writer
he is a dog
she is a nanny
it is a cat
he is a president
he is an entrepreneur
she is a singer
he is a student
he is a translator
he is hard working
she is lazy
he is a painter
he is a hairdresser
he is a waiter
he is an engineer
he is an architect
he is an artist
he is a secretary
he is a dentist
he is a florist
he is an accountant
he is a baker
he is a lawyer
he is a belly dancer
he-she is a police
she is beautiful
he is very beautiful
it’s ugly
it is small
he is old
he is strong
he is weak
he is pessimistic
she is optimistic
64. AI-judged beauty.AI contest is racist
A beauty contest was judged by AI and the robots didn't like
dark skin
69. 15%
Of enterprises are using AI. 31% said it is on the
agenda for the next 12 months
(Adobe)
70. 44ZB
Of data by 2020, containing nearly as many digital bits as there are
stars in the universe
(IDC)
71. Conclusions
● AI is not something new.
● A machine is called intelligent if passes Turing
Test
● There are two types of AI:
○ Narrow (Weak) AI
○ General AI
● Deep Learning is a part of Machine Learning
which is a part of AI, which is a part of
Computer Science
● It worths getting involved into AI