- Key challenges – Marvin Minsky [MIT] (symbolic school) & Frank Rosenblatt [Cornell] (connectionist school)
- Deep neural networks in the 1960’s with Rosenblatt — high on research agenda — (My thought: Why not until the 90’s or 2000’s if high on research agenda?)
- Output of each layer needs to be convex otherwise you will not reach a local minima.
- Breakthrough in number of layers in 2000’s
- 3-D chips on the horizon (My thought: How about N-D dimensional chips — holographic?)
- “Life begins with 1,000,000,000 examples”
- LSTM – recomputes the whole net for each sequence (My thought: how about caching with dynamic programming to speed up the computation?)
- HMM — sequential model (doesn’t do long-term associations)
- LSTM — does do long-term associations (My thought: what is the granularity of the sequence until limit reaches Plank?)
- Neocortex is a hierarchy of sequential models (RNN+Skip-List)
My journey in kernel code with Apple macOS Sierra 10.12.3
Today, I was interested in writing kernel extensions for the XNU kernel and I ran into the following problem with the example code in the Documentation from Apple. I was having multiple errors with the “so-called” working example verbatim from Apple technical documentation creators. I am going to wait on the resolution to this problem.
Fluid intelligence in “Organic” Intelligence
- Crystallized intelligence is overrated and static
- Fluid intelligence should be the sole focus and is dynamic
- Artificial intelligence is on the wrong path due to its heavy dependence on crystallized intelligence
- Using long-term memory to do tasks means nothing about the malleability of the working memory operators
- Fluid intelligence is about “pattern recognition, abstract reasoning, and problem-solving” [1]
- Can machines do pattern recognition, abstract reasoning, and problem-solving that matches the best pattern recognizers, abstract reasoners, and problems solvers in the universe? No.
- AI would thus do terrible on fluid intelligence quotient tests and thus, not pass the grade of AI.
[1] https://en.wikipedia.org/wiki/Fluid_and_crystallized_intelligence
Steve Jobs – Interview at CHM (2011) – Walter Isaacson (First 20 minutes)
Key points from the interview with Walter Isaacson:
- Technology Personification
- Traits of Steve Jobs – passion, perfectionism, impatient, storytelling
- He changed the way we listen to music
- Humanities + Technology
- Creativity + Engineering
- Brutally Honest
- Mind Distortion Reality Field (Game Theory)
- Don’t Preach, do the Fareed Zakaria or Christopher Nolan?
- Night shift
- Simplicity in bridge between computer and machine in terms of games (e.g. iPod, iPhone)
- Invention is collaboration (exception: Einstein)