0:0029:53

Whatever your question is, math already has a map to the answer

Bogdan Bocșe 29:53 2020

Math can already answer almost anything: the hard part is asking the question

A virtual library holds every possible page of 3,200 characters, and any sentence you type can be found at an exact shelf and page. For anyone building AI, teaching math or fighting misinformation, the bottleneck is no longer the answer. Bogdan Bocșe, a solution architect and managing partner of AI startups Knosis.ai and Envisage.ai, asks what kind of shared language would let humans and machines frame the right questions.

Learn how math can teach kids isomorphisms, expose hidden privilege and help fight fake news

Watch on YouTube

In short

Bogdan Bocșe argues that mathematics already maps the space of possible answers, so the real challenge for humans and machines is learning to ask the right questions. He shows how simple math ideas can teach young children abstract concepts, expose hidden privilege and inspire a more inclusive shared language, then answers audience questions on starting with AI, fighting fake news and building digital twins.

“Because math can provide the answer, but the challenge is, can we provide the question?”
“It's not the machine's job to decide what's wrong and what's right for us.”

What you will learn

  • Reframe problem solving around asking precise questions, since mathematics already maps much of the space of answers.
  • Explain abstract concepts such as functions, bijections and isomorphisms through simple visual stories that young children can follow.
  • Use prime factor decomposition, as in Eugenia Cheng's thought experiment, to map hidden layers of bias and privilege.
  • Start experimenting with computer vision in the browser using TensorFlow.js, then define a concrete objective before choosing tools.
  • Evaluate tamper-proof publishing records as one approach to making fake news harder to deny or erase.

Details

Level
All levels
Format
Talk
Technical depth
Introductory technical
For
All career stages
Length
29:53
Published
July 2020

Questions this talk answers

Where should a beginner start with AI and machine learning?

Bocșe recommends starting with TensorFlow.js in the browser, using TensorFlow for server-side work, and then defining a concrete objective to zoom in on.

Can technology alone solve fake news?

He believes AI and blockchain can help, but a lasting solution needs a change in human mentality and discipline, for example a permanent record that stops publishers pretending they never published something.

Can math be used to create a person's digital twin?

Almost: he expects technology to create a shallow copy indistinguishable to others within five to ten years, but not to reach what is inside your head, such as a secret you never tell.

Is lambda calculus the inclusive language he is describing?

It is one tile of that larger language, but it does not handle cases where an association is not a complete function, such as a superposition or a probability distribution.

Why is the growth of programming languages a problem?

Human minds cannot sustain that growth in diversity: nobody can learn all of them, and they become harder to manage over time.

About this talk

Mathematics seems to be discovered, nonetheless, the same way the perfect form for a cutting tool was discovered by primitive humans. Just as the tool, mathematics still represents a concoction of essential secrets of the Universe and circumstantial limitation of the human body, mind and perception.

Topics

Related talks

Browse the library