researchwithviolet

[researchwithviolet]

‘Learning How to Learn’ by Babara Oakley and Terrence Sejnowski, with Alistair MsConville.

I’ve selected this podcast topic because when everyone was discussing about the topics they listened to for the first two articles, I found this topic practically interesting. I have always seen myself as the ‘Black Sheep’ of the family as both my younger sisters are straight A’s student with full scholarships. And then here I stand, who can’t even secure a fifty percent scholarship. I have also always find myself zoning out and day dreaming while studying. I can never sit still and study, I would get distracted by the sound of the birds, or the smell of my mother’s cooking, or playing with the stationaries on my desk… My mom never did really think highly of me too. When I scored over 3.0 GPA during my first semester in high school, she told her friends I only scored 3 point something and that I was never smart enough to score more. When I won fourth place in a music competition, she said from the audience seat that she did not think my skills could win any prize, to the judge who was presenting the trophy to me on stage. After listening to a few discussions about ‘Learning How to Learn’ in the auditorium, I knew I got to know about the techniques and how to learn things effectively and successfully.

Listen to the podcast here:

Over the weekend, I have listened to the podcast trice, zoning out a few times in the middle of it, and rewinding again to relisten. Until at the third time I started implying the ‘pomodoro technique’ I kept hearing Barbara discussed about, where I listened for about 25 minutes, and take a coffee break, make some notes, and continued doing the same thing until I finished listening to the podcast.

The book ‘Learning How to Learn’ is a modified version of Barbara’s best seller book ‘A Mind for Numbers’ to help people, especially kids and teens, to understand how our brain works and show us how to ‘hack’ complex subjects with simple tricks.

Barbara Oakley was a successful linguist working for the U.S. Army when she first went out into the world [1]. She used to be bad at math and science and hated it throughout her childhood, thinking those subjects were just not meant for her. But as an adult, technology is constantly evolving and Barbara had been put into different positions where the skills to operate technological equipment were required. It was challenging for the linguist as technology comes from science and math, and so Barbara decided she had to learn these subjects in order to level up the game in her career progression. By doing so, she discovered the secrets to learning and earned herself a Ph.D. in Systems Engineering[2].

In Barbara’s research, we have two types of memory. The working memory, and the long-term memory, which are related to the two types of thinking, the focused mode (also known as task positive network) and the diffuse mode (also known as task negative network).

Working Memory

The working memory, as defined by Professor Nelson Cowan, a psychologist and a professor of Psychological Sciences at the University of Missouri[3], are small amount of information that is being held in our minds when we are doing cognitive tasks like solving something and learning something new[3].

Long-Term Memory

The long-term memory, as defined by Professor Nelson Cowan, is the vast amount of information we gathered from our lives and stored in our brains[4]. It is in the back of our mind that will come to play subconsciously when we are executing a task we know how to do.

As per my understanding, it is like how RAM and ROM works in computer. We have the RAM, which the computer used when executing a program, and the ROM, which stores all our saved data.

Focused Mode (Task Positive Network)

In the podcast interview, Barbara explained the focused mode as an activation mode. She described it as like activating math skills when presented with mathematical questions, activating speech skills when talking, and activating writing skills when writing. This mode is activated when we need to activate something to solve something.

When we are focused, we put specific parts of the brain to work. When we are using our focused mode, it means we are paying attention[2].

Diffuse Mode (Task Negative Network)

According to Barbara, diffuse mode is when our mind is not thinking about anything in particular. When we are in diffuse mode, we are using parts of the brain that are different from the parts we use when we are focusing[2]. This mode happens when sets of neurons that are in resting state connects. It is usually when ideas and creativity come to us.

Often, when people get stuck working on something, and they decided to take a break from it, they come back with solutions after the break. What happened was that the focused mode part of our brain had the time to take a break from focusing and the diffuse mode part kicked in. Our brain managed to take a step back and process the information from a different perspective while we are not focusing on it. This happens subconsciously. During the interview and in her book, Barbara states that our brain go back and forth between these two modes to help us learn, but we can never be in both modes at one time.

The Pomodoro Technique

The Pomodoro Technique was developed by Francesco Cririllo in the late 1980s. This method implies using a timer to break work into intervals. The technique suggests to focus working on a task for 25 minutes, and then take a 5-10 minutes break, and repeat the process till the task is completed[5].

Barbara strongly recommends this technique to help achieve effective learning. But do not think of the task to be completed as it might demotivate your will to learn. Instead, focus on the process. For example, instead of thinking ‘I need to finish this research paper’, think of ‘I’m going to work on something for 25 minutes, and I can have my coffee’. This way you can trick your brain to not think on the stressful task on hand and build better focus. During this technological era, it is also important to switch off all distractions during the 25 minutes. Put your phones away, turn off the notifications, just for 25 minutes to keep you from straying away, and you will be able to work efficiently and effectively.

Another effective way of learning complex topic is to use metaphors. From the topics we have covered above, there are many scenarios where we are able to relate, where we are able to understand immediately because of the metaphors used. While Barbara was writing her book ‘A Mind for Numbers’, which is also an adult version book for ‘Learning How to Learn’, Barbara talked to many professors who were rated as top professors, to her surprised, many of the professors said the best way to convey their ideas to students so that the students remember is to use metaphors.

Barbara also talked about other tips towards the end of the interview. Persistency is one of the keys to effective learning, repeated practices can help create links to our long-term memory and enable us to solve the problem when it arises. One of the other effective ways is to have discussions, talk about the topics you have learned. When we discussed something with other people, this helps imbedding the data into our brains, and we may have mistakenly understood some information that when we discussed it out, others can help correct us. Barbara said that leaning is also a social activity as much as an individual activity. And lastly, Barbara also suggested us to push ourselves in the process of learning. Do not fall into ‘lazy learning’, go explore different ways of learning, and challenging ourselves to learn new things.

While looking for more information about Barbara’s book, I came across a free pdf version of the book and I have read through 20 pages of the book. It is over 350 pages in the pdf file, and I intent to continue reading a little bit every day. I have also come across her online course and enrolled into the class. The course starts on 10th October 2022, and it is free to join until December 2022. I hope that towards the end of Barbara’s course and her book, I will be able to control my daydreaming tendencies, understand the benefits of being a slower learner than my sisters, and be more content and confident with myself.

External Links:

Purchase Barbara’s book ‘Learning How to Learn: How to Succeed in School Without Spending All Your Time Studying; A Guide for Kids and Teens’: https://www.goodreads.com/book/show/36647421-learning-how-to-learn

Purchase the book ‘A Mind for Numbers: How to Excel at Math and Science (Even If You Flunked Algebra)’:https://www.goodreads.com/book/show/18693655-a-mind-for-numbers

The PDF book for ‘Learning How to Learn’ can be found here: (Kindly purchase from the above link to support if you are financially stable) https://www.pdfdrive.com/learning-how-to-learn-how-to-succeed-in-school-without-spending-all-your-time-studying-a-guide-for-kids-and-teens-e195220593.html

The online course can be found here: (Free to join till December 2022. This course contains assignments that need to be submitted, so make sure to commit when you enroll for the course) https://www.coursera.org/learn/learning-how-to-learn

References:

[1] https://barbaraoakley.com/about-me/

[2] Oakley, B. (2018). Learning How to Learn: How to Succeed in School without Spending All Your Time Studying: a Guide for Kids and Teens. J.P. Tarcher, U.S./Perigee Bks., U.S.

[3] https://en.wikipedia.org/wiki/Nelson_Cowan

[4] https://www.ncbi.nlm.nih.gov/pmc/articles/PMC4207727/#:~:text=Working%20memory%20is%20the%20small,widely%2Dused%20terms%20in%20psychology.

[5] https://en.wikipedia.org/wiki/Pomodoro_Technique

[researchwithviolet]

A Commentary on “Machines like Us: TOWARD AI WITH COMMON SENSE”

I've selected my second topic - “Machines like Us: TOWARD AI WITH COMMON SENSE” with Professor Ronald Brachman from the link https://soundcloud.com/bridging-the-gaps/machines-like-us-toward-ai-with-common-sense-with-professor-ronald-brachman?si=1948f264aeba47ba93c34459ed64a9cf&utm_source=clipboard&utm_medium=text&utm_campaign=social_sharing. From my last commentary, we learned about artificial intelligence having common sense being far-fetched. In this article, we will learn more about artificial intelligence and common sense with Professor Ronald J. Brachman, the co-author of the book - Machines Like Us: Toward AI with Common Sense.

Professor Ronald J. Brachman is currently the director of Jacobs Technion-Cornell Institute, and a professor in Cornell University's Department of Computer Science. He had held before any leadership positions in the computer science and artificial inteligence field throughout his career years, including the President of Association for the Advancement of Artificial Intelligence(AAAI). He was the Chief Scientist of Yahoo and the Head of Yahoo Labs before joining Jacobs' Institute

The discussion started off with Professor Ronald's view on the progress of narrowly focused artilicial intelligence system.

I've selected my second topic - “Machines like Us: TOWARD AI WITH COMMON SENSE” with Professor Ronald Brachman from the link https://soundcloud.com/bridging-the-gaps/machines-like-us-toward-ai-with-common-sense-with-professor-ronald-brachman?si=1948f264aeba47ba93c34459ed64a9cf&utm_source=clipboard&utm_medium=text&utm_campaign=social_sharing. From my last commentary, we learned about artificial intelligence having common sense being far-fetched. In this topic, Dr. Waseem and tackle more about artificial intelligence and common sense with Professor Ronald J. Brachman, the co-author of the book - Machines Like Us: Toward AI with Common Sense.

Professor Ronald J. Brachman is currently the director of Jacobs Technion-Cornell Institute, and a professor in Cornell University's Department of Computer Science. He had held before any leadership positions in the computer science and artificial intelligence field throughout his career years, including the President of Association for the Advancement of Artificial Intelligence (AAAI). He was the Chief Scientist of Yahoo and the Head of Yahoo Labs before joining Jacobs' Institute

The discussion started off with Professor Ronald's view on the progress of narrowly focused artificial intelligence system. The professor said with enthusiasm that the progress was spectacular, especially in the recent years. Artificial intelligence had transcended from machine learning system. Through algorithmic development, computational power, availability of data, and training through machine learning, artificial intelligence system are able to complete narrowed tasks that are challenging to human. Dr. Waseem followed with the question of since we are already achieving so much with narrowed task artificial intelligence system, why are we still pursuing to achieve generalised artificial intelligence. Professor Ronald stated that generalised artificial intelligence has always been the vision in this field. It is a fact that the narrowed focus system are significant achievements in the history of artificial intelligence, but the goal was always to make artificial intelligence work in an open setting. They want artificial intelligence to perform well in unanticipated situations, meaning also to achieve human level intelligence.

The professor followed with the mars space rover example. The space rover’s task is to search for other lifeform on mars, find dusts, rocks, collect their samples and package it to be sent back for study. But what if the machine encountered other things outside from what it was tasked to do? What if the machine found different minerals or metals? If it was a human, he or she will collect the samples and bring them back for further studies. But the space rover would not do so because it is not programmed to do so. What they want to achieve is for the machine to be versatility and autonomous. They want to develop artificial intelligence to be able to perform under unanticipated input and take a certain amount of responsibilities to make decisions.

From the professor’s point of view, artificial intelligence cannot yet be responsible for their own actions. For example, if something was destroyed by the machine while it was operating, the responsibilities will fall onto the programmers of the machine. No one will blame the machine for the damages. And if accidents happened where human lives are involved, we cannot take artificial intelligence to the court to be judged. Artificial Intelligence cannot justify their actions nor provide any reasoning. There are a few examples of failed artificial intelligence system that made the news. For example, there was a trend on Tik Tok where people do ‘The Penny Challenge’. It is a challenge challenging people to plug in a charger halfway onto an outlet and hold a penny against the exposed prongs. In December 2021, a 10 year old girl was doing some challenges with her mom and when she asked Alexa to give them a challenge, Alexa challenged the girl to do ‘The Penny Challenge’. This is something a normal human being will not do. We have common sense kicking in telling us what in dangerous, what should we do and what not to do. This is also why Professor Ronald’s partner – Hector Levesque said that artificial intelligence will continue to have fatal flaw. The mistakes made by artificial intelligence are unpredictable and often dangerous.

To put a product out into the world, the professor thinks that the product needs to have a deeper sense of things. It needs to be able to predict what will happen and fall back to think through its’ current situations. They want to create a product that can perform even on problems that were not trained on before. Artificial intelligence lacks knowledge and reasoning, it needs to be able to act like human in terms of thinking like human, have autonomy and sense of responsibility which we will call common sense.

[researchwithviolet]

A Commentary on 'Artifical Intelligence: A Guide for Thinking Humans'

For my first article, I’ve selected the topic ‘"Artificial Intelligence: A Guide for Thinking Humans" with Professor Melanie Mitchell’ from the link https://soundcloud.com/bridging-the-gaps/artificial-intelligence-a-guide-for-thinking-humans-prof-melanie-mitchell. When I read through the introductory paragraph Dr. Waseem Akhtar wrote, the first sentence caught my interest immediately. Over the past decade, I found myself to be fascinated by what artificial intelligence can do but at the same time terrified that it will dominates mankind and take over the world. In this topic, Dr. Waseem and Professor Melanie discussed about the possibilities of artificial intelligence progression in the future years.

Professor Melanie is the Davis Professor of Complexity at the Santa Fe Institute, USA. She is a computer scientist, majoring in areas of analogical reasoning, complex system, genetic algorithms, and cellular automata. She is currently focusing her research on conceptual abstraction, analogy-making, and visual recognition in artificial intelligence system.

In the beginning of the interview, Perceptron was introduced. It was one of the first programs that attempts to create automated intelligence during the 1950s by Frank Rosenblatt. His idea was to stimulate neurons and recognize perception input, which means to allow machines to learn from various training examples so that they can recognize new similar examples. Professor Melanie also stated that in around 1960s, scientists and mathematicians gathered to discuss about the future of artificial intelligence in the Dartmouth Workshop. Some of the pioneers of artificial intelligence including Marvin Minsky, Herbert Simon, and Claude Shannon believed that human level artificial intelligence could be achieved in the next 15 – 20 years. They all thought that artificial intelligence will be able to do all the things a human can do, such as solving mathematical equations, driving cars, etc... And it would not take a long period of time to develop such system.

Dr. Waseem followed with questioning Professor Melanie’s opinion on the definition of artificial intelligence and the possibility of achieving human level artificial intelligence in the coming few years, as there are still speculations proclaiming that we will be able to see such technology in the next decade. Professor Melanie thinks that this sort of speculations will always be there as it has been since the 1950s. In her opinion, artificial intelligence is a big umbrella encompassing computational methods for getting machines to do things that can imitates human intelligence. However, intelligence is hard to measure, nor it is definable as human intelligence is constantly evolving and has changed over the years. Therefore, it is a challenge for machines to do things that requires that level of said human intelligence. For example, how can we get machines to behave like human? How can a system act like human? Professor Melanie says that for machines to achieve human level artificial intelligence, machines must be able to deal with all kinds of situations, to be able to generalize, to make analogies, and to have common sense. In some ways, artificial intelligence can perform better than human, but they can only perform narrow defined tasks. She used playing chess as an example. In 1996, there was a chess competition whereby the computer was able to beat the then world’s best chess player - Garry Kasparov. However, the computer can’t perform any other thing except playing chess.

Dr. Waseem then asked Professor Melanie to talk about the emergence of machine learning. Professor Melanie mentioned about the 'Expert System' being introduced after Perceptron. Perceptron works by learning and recognizing information. Whereas the expert system worked by using a set of rules to perform certain tasks. Scientists interview experts and professionals to extract rules to perform specific tasks and program it into the computer. However, it did not work well as many tasks were performed subconsciously by human and couldn’t be extracted as rules to be programmed into the computer. Machine learning is then continued to be used and evolved throughout the history of artificial intelligence.

Dr. Waseem and Professor Melanie then discussed about machine learning capabilities and about future possibilities. Professor Melanie further explained about the limitations in machines. Our system is very good at performing a certain data they were trained on, but they are unable to perform well when situations are changed, that is because the lack of generalization. This is also what they called ‘Brutalness’ in the world of artificial intelligence. When discussed about machine’s learning and understanding system. Professor Melanie stated that the term ‘understanding’ is also indefinable. Artificial Intelligence can learn through inputs, but they cannot understand the ‘why’. The professor once again used chess game as an example. When human play chess, we understand that there is an opponent, we understand that winning is good, we can explain every move that we made. However, a machine will only move the chess pieces based on what they are input with. Human can make analogies that enables us to predict something, such as what the person will do next, what move will our opponent move next. We can explain why we do something, have common sense about things, but machines cannot. Professor Melanie then concluded that artificial intelligence is not as flexible. That is because they do not have generalized intelligence. This is one of the main challenges artificial intelligence is facing to evolve to human level artificial intelligence.

Dr. Waseem further asked about Professor Melanie’s view on the possibility to achieve human level artificial intelligence over the near future. Even though various experiments and researches are still on going, and that artificial intelligence has evolved rapidly over the past decade, Professor Melanie still thinks that it is a fallacy because artificial intelligence can still not have human like generalized intelligence. Yes, they are incredibly good at performing narrowed tasks, but they cannot develop what human would called ‘common sense’. And to make artificial intelligence trustworthy and reliable, machines must first be able to know what common sense is, which seems like something impossible to achieve in the near future. That is why scientists say that ‘Common Sense is Dark Matter of Artificial Intelligence’.

Professor Melanie lastly concluded with the possibility breakthroughs in artificial intelligence such as meta-learning, which means learning certain algorithms by learning from other machine learning algorithms. Dr. Waseem pointed out that there are currently many people who are terrified with the rapidly evolving artificial intelligence system, some even called it 'Summoning the demon', Professor Melanie laughed about it and explained that we are nowhere near the point where artificial intelligence is able to dominate mankind, we have more issues pertaining to be solved about artificial intelligence for the moment, such as fine tuning facial and voice recognition. If artificial intelligence is taking over the world, it will also be a far future issue to worry about.

External Links: