Machine
Learning (ML) and Artificial Intelligence (AI): Cognitive Services and Robotics
– Part Three
by
Dr. RGS Asthana
Senior
Member IEEE
Figure 1: Image: Penn Sate/Flickr [11]
Summary
Machine
Learning (ML) and Artificial Intelligence (AI) – Part three covers inroads made
by cognitive systems and services in the recent past and also near future as
well as adoption of these by businesses.
The
role of AI and ML based robotics is also discussed including its usage and risk
factors. Ai and robotics really is a deadly combo. In the end we do discuss way forward for ML
and AI based cognitive services and robotics offered by key vendors and its effect
on businesses.
Keywords
Machine Learning (ML) Tools, Artificial Intelligence (AI), Neural Networks, Supervised and un-supervised Learning, Internet of Things (IoT), Robotics
Prelude
Machine learning [1-6] within the field of AI and along with cognitive
systems, is getting acceptance with time.
With this success, it may also soon have a huge influence on the
software business also.
A
quote from Steve Wozniac [11] is given who changed his mind on perils of ML and
AI [1] after his warning that artificially intelligent robots may
turn humans
into their pets.
‘This originally started as I was extrapolating
the ways that you can talk to your phone, and the ways it talks back. It’s
becoming more like a person. What if this trend continues and the AI develops
conscious-type thinking? That worried me, and I spoke about it for a couple
years, and was joined by smart people like Elon Musk and Stephen Hawking. But then I started thinking about a
lot of the issues that come along with making an AI. We don’t really even know
what intelligence is. You have a lot of people who study the brain, and all
they can say is some processes are governed in certain places. But they don’t
know how all those processes are wired together. They don’t even know where
memory is stored!’
Elon Musk is concerned about the developments
being made in AI research as he fears that AI will, one day, overtake
humanity. Musk wishes to take all precautionary measures to ensure that
AI advancements don’t turn humans into second class citizens.
In 2015 [13], Musk donated $10 million to
the Future of Life Institute. This
organization gives money to researchers who are working to ease the existential
risks facing humanity from advanced AI. A quote from Musk in this regard is
given below:
“It’s best to try to prevent a
negative circumstance from occurring than to wait for them to occur and then be
reactive.”
Musk is behind a brain-computer
interface venture – called, ‘Neuralink’ or initially, ‘neural lace’ – was still
in the initial stages of development. This
company’s goal is to develop a device implantable into the brain in order
to enhance human intelligence.
AI helps marketers evaluate data
and involve clients. AI is so useful today is the same technology which seemed
incredible only just a few short years ago.
The basic idea of AI based
automation or robot to handle a given scenario is possible today, though its implementation
may be difficult. First, the AI driven automation or robot [34] collects detailed
data about the scenario presented through its sensors or human input as the
case may be. The automation or robot then compares this information with the stored
data and finds what the information means. The automation or robot performs
actions and predicts which action will be most successful based on the
collected information. Of course, the automation or robot can manage only known
scenarios for which it's programmed until and unless it has multi-scenario
handling capability.
It
is felt that a brief on supervised and unsupervised learning is worth again so
it follows again even if it is a repeat from Part one.
Machine
Learning: Supervised vs. Unsupervised [7, 8]
In ML, such solutions are called target or output and
situations are called input or unleveled data. Situation and solution in
combination it is called leveled data.
Supervised: So, in Supervised learning [4]
one trains ML job for every input with corresponding target, so that it is able
to offer possible solution for any new input after being given sufficient
training. If the targets belong to some classes, it is a classification
problem. Otherwise, it is a regression problem if the target space is
continuous.
Unsupervised: In Unsupervised
Learning, you train ML job only with a set of inputs to discover the
structure or relationships between different inputs. Clustering is the common unsupervised learning
method. It creates different cluster of inputs and will be able to put any new
input in appropriate cluster. Other
unsupervised learning techniques are: anomaly detection, Hebbian
Learning and Latent variable models [4].
What is a cognitive system [10]?
A cognitive system one that does one or more activities linked
to perception and acting like it has understanding, planning capability, decision
making powers, problem solving,
analysis, synthesizing, assessment and judgment capability.
A definition of cognitive system [12] is given below:
‘Mental
system consisting of interrelated items of assumptions, beliefs, ideas, and
knowledge that an individual holds about anything concrete (person, group,
object, etc.) or abstract (thoughts, theory, information, etc.). It comprises
an individual's world view and determines how he or she abstracts, filters, and
structures information received from the world around.’
Cognitive
Services [23]
All companies use ML in the cloud in some form or other. Use of these services not only makes apps smarter [it has now
become more common word] but also one can build and
run apps that can understand, reason, and learn and can also use high-performance
cloud infrastructure. These services are offered in the areas of computer Vision
to extract rich information from images to categorize them and process visual
data — and protect users from unwanted content or personalize experiences with
emotion recognition or detect, identify, analyze, organize, and tag faces in
photos; languages: correction of spelling, capability to understand user
command; and speech: speech to text as well as speaker recognition. We describe
below cognitive services offered or planned to be offered by key players in the
field.
From Microsoft
Microsoft
offers Project Oxford, a set of curated high-level free APIs to cover machine vision,
speech recognition, and language & speech analysis. We give here[18] some detail of each cognitive
service in areas viz., vision: Image recognition and sentiment analysis from
given images; speech: Fine-tune speech recognition for anyone, anywhere or give
your app the ability to know who's talking or Convert speech to text and back
again, and understand its intent; language: Detect and correct spelling
mistakes within your app or Teach your apps to understand commands from your
users or Easily parse complex text with language analysis; knowledge: explore relationships among
academic papers, journals, and authors or contextually extend knowledge of
people, locations, and events or add interactive search over structured data,
or provide personalized product recommendations for your customers or distill
information into conversational, easy-to-navigate answers; and finally search:
intelligent autosuggest options for searches, or Bring advanced image and
metadata search or trending videos, detailed metadata, and rich results or Connect powerful search to your apps; which you can use in your app and put your output on the device of your
choice.
Microsoft recently announced
that H2O’s AI platform is available on Azure HDInsight Application Platform. The
H2O team and Azure HDInsight team will integrate and deliver best solutions to
businesses within open analytics framework. Using the Face Detection API, you
can upload an image (or point to an online URL), and the API will return
information about any faces located in the image. The face API with
moving picture Video API [24] can get or extract deeper understanding from
images e.g., gender and age estimate of a person in the
image.
From IBM [19]
IBM Watson offers cognitive services rolled out an array of ML - powered
services on its Bluemix PaaS including: weather prediction, systems for analyzing language, image
recognition, language translation and sentiment &
tone analysis, and so on through its cloud. In the area of Language: it has a free service to begin with.
This service will remain supported till Mar 7, 2018 only. Use the Watson
Natural Language Understanding service in your app to tap into the same
powerful text analytics and natural language processing capabilities,
conversation, dialog, document
conversion, language translation, Natural language understanding and
classification and personality insight and tone analysis; Vision: recognition;
Speech: speech to text as well as text to speech conversion.
From Apple [20, 21]
Apple acquired Emotient, a startup which
was in cognitive systems and was developing software for recognizing people’s
moods through the analysis of facial expressions. It may be worth noting that Emotient
was involved with Apple competitor Google on its Google Glass project.
The
move to provide ML and cognitive capabilities in iPhone and MAC computers means,
“made-for-business apps now have the ability to understand, reason and learn
based on deep data analytics,” according to IBM. Watson APIs are optimized to
work with iOS 10’s speech framework. Next move will be of exploiting contextual
awareness. As per IBM, such systems can
“interact naturally; learn from interactions and surface meaningful insights
from huge amounts of data.”
From Facebook
As
per Zuckerberg - the founder of Facebook, 'one of our goals for the next five
to 10 years, is to basically get better than human level at all of the primary
human senses: vision, hearing, language, general cognition.
Facebook
team is interested in developing AI or ML software that can analyze a photo and
answer questions about what it shows, or study a picture of toy blocks and
predict whether they will fall over.
From Google
Google [22] offers
cognitive services in the area of computer vision: Through Google’s Cloud
Vision API, developers can identify the content of an image by
encapsulating powerful ML models by using REST API which classifies
images into thousands of categories (e.g., "sailboat", "dog",
"Eiffel Tower" etc.), can do object and face detection, and
finds text and identifies language contained within images. It can also do
image sentiment analysis, detect inappropriate content and integrate your image
storage on Google Cloud Storage;
Google Cloud Platform offers: Google Translate (an API with Neural Machine
Translation technology [35]),
and Google
Prediction API. Google has also introduced this tech to Maps. With this, Google
will automatically give translations on Maps in user’s native language.
From Amazon
Amazon ML is similar to
Google Prediction API in that models can be trained against data and used to
make predictions. It's a deliberately simplified service,
either for the sake of appealing to developers who only want to solve a
specific, narrow problem or because Amazon wanted to test the markets first
before making a full-fledged commitment.
It is clear from above that Apple and Facebook are ready for
deep dive in this technology and may also have cognitive services soon. Google
and especially Amazon follow one guiding principle for cloud approaches and it's
"less is more."
Robotics
ML and AI based
robotics [31] is entering in areas including Machine Vision, Imitation learning,
self-supervise learning, Medical Technologies and multi – agent Learning.
Smart Robots will be flying, swimming and will be able to
walk and crawl based on the need of the situation. Tech pioneer Masayoshi Son’s
robot Pepper [32] has emotions also and works as a bank clerk (see figure 2).
In other scenario, AI robots answer customer service questions for UK mobile
network O2.
In
the Terminator movies we see a great AI that has taken over the world and uses
time travel to send back cyborgs to eliminate its human foes from the timeline.
People and the market [33] decide to manufacture more and more human-like AI
based products (or Robots) which are not only economic but innovative also.
Figure
2: Robot Pepper - works as bank clerk
[32]
Robot Dependence: Is it bad?
Case -1: More
than a quarter-million Americans turn 65 every month. In US very soon, millions of people will need
help as their health predictably becomes feeble with time. It costs around
$90,000 a year if you hire a Skilled Nursing Facility and around $25,000 per
year for hiring an Assisted Living Facility. In view of this the case
of elder-care robots [16] will play an important role, i.e., not all
dependence upon machines is necessarily bad. Robots could most notably help ill or disabled
people as they need someone say, friends or social workers if they go out. In
the future, these people could be free to leave the house and live fuller home
lives alongside robotic helpers.
Case 2: Rotimatic
[15] is the first AI based robotic kitchen appliance (see figure 3) which
automates the entire process of roti making.
Roti making is a time consuming job but it is an ideal diet for a north
Indian. Fresh puffed rotis within 90 seconds can however be made by only touch of
button by Rotimatic. One can, however, set the parameters for the flour type,
thickness, roast level and the oil content for each roti.
Figure
3: Rotimatic in action
Case 3: According
to Ricardo Campa [14], a professor of sociology at the University of Cracow the
problem could be bigger than many expect.
“If we start building humanoid home robots that are
beautiful and capable of interacting with humans in a pleasant and satisfactory
way, the danger is that we will interact less and less with other fellow
humans.” The problem is that robots not
only be really beautiful but also very polite irrespective of the owner’s
behavior. These robots can be programmed to do many house hold jobs with
perfection. The main fear, therefore, is
that owner may like the Robot so much that he may lose human touch altogether.
Case 4: Though Boston Dynamics — a Google Company [25] — has been
building impressive but frightening legged robots for decades. But recently, it
has been using its dog shaped robot for commercial purpose to deliver a package
strapped to its back (See Figure 4) to someone’s front door. Company displayed a legged Robot nicknamed, ‘Handle’
can jump over hurdles and land on its wheeled feet, lift a single leg while
moving, stroll in the snow and go down stairs, as well as carry up to 100
pounds. This robot may be ideal for delivery.
Case 5: To produce Tesla Car Model 3 in time, Tesla [26] has received a consignment
of hundreds of Kuka robots for its production line. Tesla has been using both Fanuc and Kuka
robots on its production lines at the Fremont factory and at the Giga factory
in Nevada. It is a regular case as robotics
has been used in automotive industry for a long time in past for doing welding
and painting jobs.
Figure 4: Robot Dog used for Delivery [25]
In Giga Factory of Tesla, Self-navigating Autonomous Indoor Vehicles (AIVs) are
not caged or follow floor magnets or navigational beacons but AIVs can move in
the factory without restrictions. These robots not only detect but also safely
avoid people and obstacles with their sensors and use a digital map for
navigation.
They are used for moving
materials between workstations [29] (see figure 5). In fact, AIVs are
collaborative Robots and can safely work with human workers.
AIVs are customizable. For
the visits last week, they were welcoming people to the factory saying “welcome
to Tesla”. The automaker also uses them at its Fremont Factory. Tesla modified one to make it
look like R2-D2 from Star Wars.
It can be seen that in some cases robot dependence is
OK but if it may lead to almost zero human touch scenario, particularly, in
first and third case then it could be bad but in case 2 and 5 its OK. In case
4, we show a robot which is ideal for delivery but need to be more sober
looking. One has to look at this aspect from ethical angle.
Further [17], Robots have qualities including mass production and self-replication, mind transfer from one
robot to other Robot simply read data migration or replication & ease to
upgrade via remote connectivity, zero fatigue, no evolved psychological
predisposition, moral superiority, immunity to damaging biological functions and possibility
of adopting to many desired shapes based on situations, i.e., dynamic
morphologies which make Robots more useful than humans in many scenarios or
situations. The good news is that robots have no sense to know
that they are superior and hence do not compete with human and human is free to
think that they are far superior in many ways. This situation is today but what
happens tomorrow is a question Mark?
Figure 5: Robots which move material from one work
stations to other work stations [29]
Way forward
ML [10] is in race to become the most preferred solution for
the next enterprise intelligence business offerings.
Microsoft
Text analysis API can extract key phrases; detect the topic and language of the
text. Sentiment analysis is only provided in English, French, Spanish, and
Portuguese, but more languages will also come soon.
Amazon and Google target developers whose data is already on
their clouds. IBM and Microsoft are aiming for far broader territory, and
while IBM is offering a lot, it also has the most to lose as Microsoft cloud
services also include lucrative business from areas such as gaming adding a
separate financial channel for revenue.
ML based algorithm
[28] can mimic other peoples' voices or create new ones from scratch. A Canadian start-up has developed a voice
imitation program capable of mimicking a person's voice after just a minute of
listening to them speak or from a minute of sample in form
of audio clip only. This technology could be used in the form of mimicry to
very risky situations to impersonate someone based on imagination and role of
its possessors.
The top robotics manufacturers
[27] are located in Japan, Germany and Switzerland as of today. China employs
the highest number of these manufacturing robots as compared to any other
country, but that number is still very low as compared to number of human employees.
South Korea, however, has the most robots per human worker. It has, therefore,
has potential to become a more industrious, efficient, and profitable place to make
things than Chinese factories.
As China and South Korea
not only need to increase investments in robotics in their industries to remain
competitive in the world market but also need to fine-tune rising labor costs
with more intelligent and smart automation. This will bring a boom in domestic
robotics market in China and Korea and will result in mushrooming of the number
of industrial and AI based robots in various engineering roles. This scenario, however, paints a decimal
picture of USA [27] which may be left behind in the race if it does not do
something different.
AI and robotics has come
up as a really powerful combo [30]. Fanuc and Nvidia have teamed up to produce AI-powered robots.
Fanuc and Kuka robots are already used by Tesla in their Gigafactory where
machine builds the machine. Swiss
company ABB and Watson have entered in partnership [35], where ABB will use IBM's
AI technology in its Industrial robots and other connected devices. Such intelligent and smart robots and will
fill the need of industry in future although it may also replace human labor
with machine.
Within 30 years robots [32]
will outnumber humans on earth and they will have IQ of minimum 10,000 and will
be 1000 times smarter than humans. Think of job scenario then?
References
[1] Progress and
Perils of Artificial Intelligence (AI)
[2] Invited Chapter 6 - Evolutionary Algorithms and Neural Networks, Pages 111-136, R.G.S. Asthana
in book, Soft Computing and Intelligent Systems (Theory and Applications),
Academic Press Series in Engineering, Edited by:Naresh K. Sinha, Madan M. Gupta
and Lotfi A. Zadeh ISBN: 978-0-12-646490-0
http://www.sciencedirect.com/science/book/9780126464900
[3] Future 2030
by Dr. RGS Asthana, Senior Member IEEE
[4] Machine
Learning (ML) and Artificial Intelligence (AI) – Part 1, by Dr. RGS Asthana, Senior Member IEEE
[5] Machine Learning
(ML) and Artificial Intelligence (AI) – Part Two, by Dr. RGS Asthana, Senior Member
IEEE
[6] Google releases TensorFlow 1.0 with new
machine learning tools
[7] Supervised learning
[8] What is the difference between
supervised and unsupervised learning algorithms?
https://www.quora.com/What-is-the-difference-between-supervised-and-unsupervised-learning-algorithms
[9] Internet of Things (IoT)
[10] WIRED: Machine Learning and
Cognitive Systems: The Next Evolution of Enterprise Intelligence (Part I)
[11] How Steve Wozniak Got Over His
Fear of Robots Turning People into Pets
[12] Cognitive System
[13] Elon Musk: just outlined how
He’ll merge the human brain and AI
[14] 5 Things Humans Do Every Day
That Robots Will Soon Take Over
[15] Rotimatic
[16] How to convince your baby
boomer parents to Let Elder-Care Robots nurse them?
[17] 12 reasons robots will always
have an advantage over human
[18] Cognitive services
[19] Watson Services: Take your
first step into the cognitive era with our variety of smart services.
[20] Apple Emotient acquisition
[21] IBM, Apple, bring Watson into
the iOS enterprise
[22] Google Cloud Platform: Powerful
Image analysis
[23] How IBM, Google, Microsoft, and
Amazon do machine learning in the cloud
[24] Microsoft Cognitive Services
Leap Forward
[25] Boston
Dynamics has been using its robot ‘dog’ to deliver packages in Boston
[26] Tesla receives massive
shipment of robots for Model 3 production line – first pictures
[27]
America may miss out on the next industrial revolution
[28]
Speech-imitating algorithm can steal your voice in 60 seconds
[29] Tesla Gigafactory: a look at the
robots and ‘machine building the machine’ at the battery factory
[30] Fanuc
Robots to be powered by Nvidia; Already powering Tesla
[31] Machine Learning in Robotics – 5 Modern Applications
[32] Get set for smart RO bots with
an IQ of 10,000
[33] Will
AI and Robotics Create a New Form of Slavery? (Part 1 - Humans as Masters)
[34] How
Robots Work
[35] Watson
could be the key to smarter manufacturing robots
[36]
Google equips its Translate services with Neural Machine learning
Hello Molly Mclaughlin
ReplyDeleteI just read your article The Best Credit Card Processing Services of 2018 and it was really helpful for me because I was not aware of such type of credit card processing services.
But I have done further research on it and I found that there is another service which you have not mentioned in your article i.e Fake Credit Card Generator service which is available on credit card name and number generator
. Now the question is this service is related to credit card processing services as you mention in your article?
Hope you will answer this to clear my doubt and also tell me that is there any new update regarding credit card processing services in 2019
Thanks!
It is very good post. See my post on my site visit my site
ReplyDelete