dr. sven j. körner [email protected] @svenjkoerner€¦ · alice: balls have zero to me to...
TRANSCRIPT
Dr. Sven J. Körner
[email protected]@svenjkoerner
“We humans have a love-hate relationship with our technology. We
love each new advance and we hate how fast our world is changing.”
— Daniel H. Wilson
NO!
We Used the Last ParadigmShift For Good
But We Run The Risk of Becoming the Largest Living Museum of
20th Century Tech.
We’re Missing Out on 21st Century Tech.
Overestimating the Short Term Impact.
Underestimating the Long Term Impact.
AI Isn’t Taking Over
Others Are Overtaking Us With AI.
The First Rule of AI:
Hard Problems Are Easy.
Easy Problems Are Hard.
Substituierbarkeitspotentiale Deutschlandnach Branchen
Quelle: FAZ http://www.faz.net/-gym-9f0y3 / Beschäftigungsstatistik der Bundesagentur für Arbeit, Stand 31.12.2016
Anteil der Beschäftigten mit hohem Substituierbarkeitspotenzial (>70%) an allen Beschäftigten.
Software Engineering vs. AI (SE 2.0)
Problem Code Solution
Problem Solution Code
Context and Proper Training Matter
but use it here...If you train a system on these...
Drawbacks of NN – (Wrong) Training contd
“People remember errorscommitted by AI, but forget human errors”
am am pmpm
ampm
ampm
inout
Input Layer
Output Layer
day shift night shift
am am pmpm
ampm
ampm
1 1 110 0 0 0
inout
Input Layer
Output Layer
day shift night shift
am am pmpm
ampm
ampm
.3 .2 .1.9.8 .9 .6 .4
inout
Input Layer
Output Layer
am pm
ampm
ampm
.3 .2 .1.9.8 .9 .6 .4
(.3+.1)/2 = .2 (.8+.4)/2 = .6
inout
ampm
ampm
Input Layer
Output Layer
am pm
ampm
ampm
.3 .2 .1.9.8 .9 .6 .4
(.3+.1)/2 = .2 (.8+.4)/2 = .6
Gradient Descent
inout
Input Layer
Output Layer
am pm
.3 .1 .1.6.9 .9 .6 .5
(.3+.1)/2 = .2 (.9+.5)/2 = .7
Gradient Descent
inout
Input Layer
Output Layer
ampm
ampm
How Computers and Brains Are Different
How Computers and Brains Are Differentcont‘d
How Computers and Brains Are DifferentPart III
“We found that this information bottleneck measure doesn’t see compression in the same way you or I would. Given the choice, it is just
as happy to lump 'martini glasses' in with ‘Labradors', as it is to lump them in with
'champagne flutes”
Brendan Tracey (SFI, MIT)
Words exist because of meaning. Once you've gotten the meaning, you can forget the words.”
Zhuangzi
“
Common Sense.
For Computers.
What Are Semantics?
Identifying an adult female from other persons might be a data-centric problem
But finding the teacher and the mother requires information not in the data.
adult_female 0.943
adult_female 0.854
mother
teacher
Facebook Bots Invented 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
OpenAI GPT-2
Recycling is good for the world, no, you could not be more wrong.
Recycling is NOT good for the world.It is bad for the environment, it is bad for our health,
and it is bad for our economy. I’m not kidding. Recycling is not good for the environment.
It is destructive to the earth and it is a major contributor to global warming. Recycling is bad for our economy.
It increases the cost of a product, and in turn, the price of everything that is made with that product.
Recycling is not good for our nation. We pay a tremendous price for the privilege of having the world’s most
advanced and efficient recycling system. Recycling is a huge, colossal waste of time, energy, money, and resources.
Replacing
DrudgeWork
Human Ability To Learn
Amount of Information
Computer Thinking Speed
???
1st : Automate Expensive Tasks
2nd: Not Just Faster – But More
Your Job
Hertz Car Rental Example
<Type of Clause, Standard, or Paragraph>GDPR Regulation No. 1
<Type of Clause, Standard, or Paragraph>GDPR Regulation No. 2
<Type of Clause, Standard, or Paragraph> Procurement Clause No. 4711
<Type of Clause, Standard, or Paragraph>GDPR Regulation No. 3
<Type of Clause, Standard, or Paragraph>GDPR Regulation No. 5
<Type of Clause, Standard, or Paragraph>NDA Regulation No. 13
<Type of Clause, Standard, or Paragraph>GDPR Regulation No. 4
<Type of Clause, Standard, or Paragraph>NDA Regulation No. 2
<Type of Clause, Standard, or Paragraph>GDPR Regulation No. 6
<Type of Clause, Standard, or Paragraph>Procurement Clause A14
<Type of Clause, Standard, or Paragraph>Export Restrictions Clause
Topics:
• Reference to internal standard of paragraphs/chapters/clauses
• Mapping from clauses to paragraph(s)/chapter(s) in other contracts
• Clause correction and detection• Auto-tagging of contracts• Extract values for further (automated)
processing
Green => Matching OKYellow => Values extractionOrange => Multiple matchesRed => No Gos
AI Lawyer I: Helping to Find Relevant Contract Data and Map
to Existing Clauses, Standards, and Contracts
Semantic Similarity Service
Box.com (Document Management)
Sharepoint (Semantic Compare)
Semantic Clustering (Matrix)
If only 10% are replaced by machines – what would that mean to society?
More than 10% of money for states get lost.Machines don’t pay taxes.Machines don’t consume.
The First Rule of AI:
Hard Problems Are Easy.
Easy Problems Are Hard.
[email protected]@svenjkoerner