Loading presentation...

Present Remotely

Send the link below via email or IM


Present to your audience

Start remote presentation

  • Invited audience members will follow you as you navigate and present
  • People invited to a presentation do not need a Prezi account
  • This link expires 10 minutes after you close the presentation
  • A maximum of 30 users can follow your presentation
  • Learn more about this feature in our knowledge base article

Do you really want to delete this prezi?

Neither you, nor the coeditors you shared it with will be able to recover it again.


Make your likes visible on Facebook?

Connect your Facebook account to Prezi and let your likes appear on your timeline.
You can change this under Settings & Account at any time.

No, thanks

TOK Presentation: Can a Machine Know?

No description

Pearce Ropion

on 16 April 2015

Comments (0)

Please log in to add your comment.

Report abuse

Transcript of TOK Presentation: Can a Machine Know?

Subjective part of knowledge

If you are not
that you 'know' something, then you do not know at all.

AIs don't have consciousness, they are only conscious that they hold information.
What is Consciousness?
Physical Facts
Knowledge Question
To what extent can machines know?
Can a Machine Know?
A TOK Presentation by Salomé Cap and Pearce Ropion
Who is Watson?
Watson Wins at Jeopardy
What are the
criteria of knowledge
for humans?


What Defines Truth?
distinguishes knowledge
from belief
Truth Dilemma
As humans, is what we think we know true?
Conclusion for Truth as a Criteria
The same dilemma for truth is applicable, so truth 'beyond reasonable doubt' is applicable for AIs too.
Conclusion of Belief as a Criteria
AIs do not and cannot believe in anything: subjectivity is not possible.
They are not able to come with a judgement.
Truths and beliefs must be justified, and justified correctly.
Sense Perception
AI language is code: how they learn.

They search for denotative definitions of words but cannot interpret the "intended meaning" - not understanding of connotations, allusions, or tone.
AIs can perceive senses through detectors.

They cannot interpret that information beyond the analysis of components and results form research.
What are the limitations they face?

To what extent can a machine know according to criteria for human knowledge?
Knowledge Defined
Plato's tripartite theory of knowledge: a justified true belief
Humans generate the 'knowledge' that are programed into computers, machines and AIs.
>>The same dilemma applies: how can we be sure that what AIs could 'know' be knowledge?
This criteria is fulfilled by AIs, Watson fulfilled this criteria.

Conclusion of Justification as a Criteria
The criteria is partially fulfilled.

A superficial understanding of the reliable ways of knowing is present.

The less reliable ways of knowing cannot be used by AIs.

Machines cannot know completely.
Watson in Social Sciences
Watson in Natural Sciences
Watson in Ethics
Even with AIs that aren't able to know, they could become part of our daily lives because they are so useful and polyvalent.
Beating The Turing Test
Thank You
Chalmers, David. "How Do You Explain Consciousness?" TED. TED, Mar. 2014. Web. 24
Mar. 2015. <https://www.ted.com/talks/david_chalmers_how_do_you_explain_consciousness?language=en>.
"Consciousness." Theory of Knowledge. Theory of Knowledge, 12 Oct. 2009. Web. 24
Mar. 2015. <http://ibtokspot.blogspot.nl/search/label/Consciousness>.
"Eugene the Turing Test-beating 'human Computer'" The Guardian. Guardian, 9 June
2014. Web. 24 Mar. 2015. <http://www.theguardian.com/technology/2014/jun/09/eugene-person-human-computer-robot-chat-turing-test>.
Fry, Steven. "Stephen Fry Does 'The Knowledge'" YouTube. Google, 25 Nov. 2011. Web.
24 Mar. 2015. <
Gennaro, Rocco J. "Internet Encyclopedia of Philosophy." Internet Encyclopedia of
Philosophy. Internet Encyclopedia of Philosophy, n.d. Web. 24 Mar. 2015.
How to Create a Mind. Dir. Ray Kurzweil. Perf. Ray Kurzweil. TEDTalks, 2013. TEDx. TED, 6
Mar. 2013. Web. <http://tedxtalks.ted.com/video/How-To-Create-A-Mind-Ray-Kurzwe>.
Howard, Jeremy. "The Wonderful and Terrifying Implications of Computers That Can
Learn." TEDx. TED, Dec. 2014. Web. 24 Mar. 2015. <https://www.ted.com/talks/jeremy_howard_the_wonderful_and_terrifying_implications_of_computers_that_can_learn>.
"IBM's Watson Supercomputer Destroys Humans in Jeopardy." YouTube. Google, 13 Jan.
2011. Web. 23 Mar. 2015. <
Masnick, Mike. "No, A 'Supercomputer' Did NOT Pass The Turing Test." Techdirt. Floor64,
9 June 2014. Web. 24 Mar. 2015. <https://www.techdirt.com/articles/20140609/07284327524/no-supercomputer-did-not-pass-turing-test-first-time-everyone-should-know-better.shtml>.
Parkin, Simon. "Rewriting the Rules of Turing’s Imitation Game." MIT Technology Review.
MIT, 17 Mar. 2015. Web. 24 Mar. 2015. <http://www.technologyreview.com/news/535391/rewriting-the-rules-of-turings-imitation-game/>.
"Theory of Knowledge." Theory of Knowledge. Theory of Knowledge, 2015. Web. 24 Mar.
2015. <http://www.theoryofknowledge.info/>.
Cellan-Jones, Rory. "Hawking: AI Could End Human Race." BBC News. N.p., 2 Dec. 2014.
Web. 18 Mar. 2015. <http://www.bbc.com/news/technology-30290540>.
Can a Machine Know?
In February, 2011 Watson was broadcasted playing jeopardy against the top two players in the world.
there are shades of grey in differentiating knowledge from belief
knowledge is more than isolated facts, they are methodically related
Opinionated Beliefs
Types of Thought:
The validity of the statement is determined by reliability of the justification.

To find a reliable source of information, we must appeal to the Ways of Knowing.
Solution: accept it; in humans, knowledge requires less than certainty
true = 'beyond reasonable doubt'
This criteria is not fulfilled by AIs, Watson does not fulfill this criteria.
"Knowledge is primarily based on reason."
A computer's use of reason is superficial at best.
Limited to inductive reasoning, computers cannot reason in the same way humans can.
Thus, according to rationalism, AIs cannot know because this criteria is not complete.
Inductive reasoning uses a specific example to make a general rule.
Computers can pick up on a small piece of data and make a conclusion based on those facts
Humans can apply general rules to specific examples.
Tabula Rasa
They can be provided with facts which can be computed for a feedback.
"Computers are a blank slate."
They cannot actually understand the information they have.
Machines like Watson contain facts, can find the correlation between these facts and the question, can give you a probability of the answer being right, but doesn't understand any of the information.
AIs are able to store most of human derived knowledge.
The Remaining Ways of Knowing
Computers have the ability to store and access any file from any time with simplicity and ease, but cannot create links without human intervention.
"Knowledge is primarily based in experience."
AIs use memory in a very limited way.
A limited version of the primary aspect of knowledge (according to this theory)
>> a limited version of knowledge.
These are ways of knowing that are specific to humanity
AIs cannot have the ability to think creatively - to
AIs cannot have
in something that is not based in pure fact but is ascertained through belief.
AIs cannot have
, a "gut feeling" that something is right or wrong, without any supporting facts.
They cannot feel pain, sadness, anger, happiness, fear and hope. Only humans can experience an

The Approach
What is truth?
What is the problem with truth?
What is belief?
What does it mean to be conscious?
What does it mean to be justified?
What ways of knowing are computers able to use? Which ones can they not?
Can a machine know?
The RLS, Use of AIs in the World, Exploring Further

The Debate
Separate Perspectives
Can a Machine Know?
1: No, a machine cannot know:
AIs do not satisfy all the criteria for knowledge.
2: Yes, a machine can know:
AIs satisfy certain criteria for knowledge, and it is enough.
Types of Reasoning:
IBM's Watson is an artificially intelligent computer designed to
answers questions in human language.
"Jeopardy!" was the test to affirm the success of Watson as a question answering machine.
Watson beats them with a large margin even though it
did not
answer everything correctly.
Watson's databases stored the
of millions of web-pages and it accessed this
to answer the questions.
Did Watson actually know the answers? Although Watson has access to
, what does it mean to be able to
Are Watson's answers just computation of symbols, in this case words?
Because it is a machine, it has no ability to
the meaning of those symbols.
Any question that is not purely based in fact is unanswerable because the computer cannot
in the same way that human can.
Consciousness is being aware of one's existence.
Computers lack belief and the full depth of justification.
Knowledge is the understanding information.
AIs are unable to understand anything, they only have information.
Doesn't know the subtlety of the words - superficial grasp of language.
They cannot form experience from the stimulus.
Fields that need to take into account a lot of information (law)
Time sensitive cases (financial markets)
Complete knowledge isn't the goal in AIs, it is for the to be useful.

prosthetic limbs
aiding doctors
Other RLSs
June 2014 - a computer was able to convince a human that the computer was also human
The issue is that "[AIs] would take off on its own, and re-design itself at an ever increasing rate,"
"Humans, who are limited by slow biological evolution, couldn't compete, and would be superseded."
- Stephen Hawking
The computer, a chat-box named Eugene imitated a 13 year old boy to use the foolishness of a child against the judge
Does imitating humans count as knowing like humans do?
The Turing Test is a thought experiment designed to see if AIs have developed to the point where humans can't tell the difference
Humans are able to recall these memories, learn from them and build upon them, but AIs can’t.
Use them a tools for human development.
This is the most important WoK for AIs.
Full transcript