Featured

First of all, who am I?

This is the blog of a mad scientist who does subjective science.

You may ask first, what is subjective science?

Before giving you my answer, let me talk a little bit about myself.

I have always firmly believed that science represents rationality. Scientists collect objective data, anaylze and draw conlucions on propositions and theorems that could possibly explain this world we are living in right now. I used to believe that science was exclusive as one needs to possess a tremendous amount of specialty knowledge to step into an academic science field and participate in science researches, to really trace and work upon the progress made by all the preceding scientists……

Until I watched a youtube video about a physics experiment done in public. When the man in the video said that he was going to introduce the wave property of light to some random tourists at the beach, I assumed that his purpose was to make fun of other people. It took me a whole week of class in high school to understand what is light and how it could behave as a wave. It was as astonishing as the time I was told that there was no Santa Claus when I was six. There was no chance that man could effectively teach people in five minutes. It was mission impossible.

He made it. And I loved seeing how people showed a great “Ah ha!” face when they were enlightened by their new discovery about light. It was then I realized that science education for public is not as hard as I imagined. The man in the video simply used a combination of analogies and metaphors to help visualize the concept of “wave”. And showed people how light had the corresponding properties by letting them observe the famous double-slit experiment. The crucial part of science education, I concluded, is to explain science in an “everyday language”, and to show science with perceivable (visible, audible, etc) evidence.

What do I mean by “everyday language”?

Let’s see how wikipedia describes the double-slit experiment:

“The experiment belongs to a general class of “double path” experiments, in which a wave is split into two separate waves that later combine into a single wave. Changes in the path lengths of both waves result in a phase shift, creating an interference pattern. “

Do you understand what is double-slit experiment now? I suppose most people would say no. (Physics major or scientists can skip this part) What is a phase shift? What is an interference pattern? This would be endless if we click on every physics jargon that leads to another web page.

Suppose you are inside a closed dark room and there are two narrow short vertical slits on the wall. There is light projected on the other side of the room through the slits. If the light is a beam and only travels in a straight line, we should expect to see two slit-like lines of light projected. Yet what we actually see is a pattern of multiple bright and dark bands. This is the observation that opposes to our assumption “light is a beam”. Boom, now you know the general observation and procedure of the double-slit experiment.

Science is not exclusive. Different methods of communication and various tools of demonstration make science inclusive, comprehensible, or even fun. And I want to contribute to building the bridge between you and science in my own way.

Then what is this blog about?

What if science is not about humans discovering things? What if it’s the things that let themselves to be discovered?

Imagine a new technology which could make things speak. Yes, things, all sorts of things.

I can ask the cloud at what time it would rain before I go out for work. I can ask the tree holes about people’s secrets. I can even ask the electricity how fast can it travel through the wires.

It’s simply that sometimes they react to me, sometimes they don’t. Depend on their moods.

Imagine what could happen if this technology is used for science? Science could speak for itself. Reports will be written in an active-voice instead of passive-voice. We will be investigating subjective science instead of objective science.

Citation:

https://en.wikipedia.org/wiki/Double-slit_experiment

AI is on its way……

Have you ever imagined AI (Artificial Intelligence) writing a fiction novel? Is it possible for them to organize words and languages just as we do? Do they even possess the ability to think, or even imagine?

In other words, does artificial intelligence possess subjectivity? Now follow me to look at some of the interesting texts that AI has generated.

This talktotransformer.com is the website that I looked at. It is a random text generation website that allows you to type the first sentence of an article and then the AI will finish the rest for you. I experimented with the text generator to see how “intelligent” the AI was. In fact, it was not as clever as I thought it to be.

In the first experiment, I typed in “I love cats” as the first sentence and I expected the AI to agree with me and compliment how cutes those kitties could be. When the AI started to type the first few words on my screen, I was astonished and amazed.

“I love cats. A lot. They’re adorable and playful and absolutely……”

The words in blue color were the texts completed by the AI, and it successfully made me feel like it was a true cat-lover. Yet the amazement did not last long. As the typing went on, the whole paragraph became a little weird:

“……gorgeous,’ Ehlrich says. ‘But they are all-too-often the exception rather than the rule. That is just what happens in politics.’”

Politics? What do cats have to do with politics? And cats are not generous at all! I could almost declare that the first experiment had failed as I generated more versions of texts.

“I love cats. But man, that Tuna Shit Tuna shit on the dollar bag was one of the most disgusting things I’ve ever ordered. It was a huge tuna roll in a can, and it was terrible.”

As you can see, the AI started to talking some nonsense. (I mean, how can a bad-taste tuna can relate to loving cats?)

My second experiment was to test the potential of this text generator. I was curious if it could possibly help me do my sociology or physics homework in the future. I typed in “Dark matter is……” and wondered if it would give any definition of dark matter from a physicist’s perspective. The result turned out quite well, to my surprise.

“Dark matter is a substance made of only one thing: energy. Whether or not you believe in a God or gods is not your business. But if you don’t believe that we’re all part of the same universe, then that’s where your religion has gone wrong. In fact, all belief systems are the same in terms of their distribution of the Universe’s energy.”

Quite impressive, right? In spite of a little religious content, the AI did a pretty excellent job defining dark matter for me. Compared to the first experiment, I suppose the AI behind the system was more sophisticated and knowledgeable in dealing with academic objective questions than subjective emotional questions.

The next step was to fine-tune the AI by changing its data base (where the AI extracts the words and combine them into its own article). I decided to investigate two main questions:

  1. If I add an intense amount of physics research paper (30) to the existing data base, will the AI be able to answer a more complicated physics question?
  2. If I add an intense number of cat-related interviews and speeches (30), will the AI be able to become a true “cat-lover”?

For the first question, I pulled random physics research paper from google scholar, including researches from various fields like particle physics, nuclear physics, etc. I combined all the texts into one txt file and uploaded it to the data base. I then tried generating texts with the updated data base. Since the first try was more like building a prototype, I did not expect it to present perfect results. As I anticipated, the first article the AI generated made no sense at all. Since I copied every single details from each research paper including the weird data, formula, notation and citation, the result the AI generated was not even an article! It was a bunch of meaningless symbols and numbers that I obviously could not understand. Maybe I had to choose pure texts myself so that the paragraph generated could at least be “readable”, or it’s simply that the AI was messing around with me.

Thus, I extracted only the words from all the research papers, getting rid of the data, the graphs, and all the cited names of great physicists. I then ran the machine again to see whether the AI would surprise me this time. I set the first two letters to be “The atom” and the result was amazing.

“An atom is an object that is arranged in a certain way.

For example, an egg is an egg, and if the egg is flat, then the floor of the egg is flat.

However, if the moon is in the constellation of Pisces, the egg is able to rise up to an extremely high temperature and pressure and then collapse in a certain number of pockets called compartments.”

Compared to its previous explanation on dark matter, at least within this one, the contents seem relevant. Though the AI was not directly explaining the concept of atom to me, its analogy and metaphor used surprised me and made me think of its future potentials. Suppose there is an extremely large amount of data within AI’s database, if we type in random physics definitions, will the AI be able to explain it in some unprecedent but rather effective metaphors? Or, if we type in two seemingly unrelated physical quantities, will it be able to find any correlations or simply inspire physicists with some random ideas? These predictions of the future seem totally feasible to me. Though at first, I was expecting the AI to have enough capacity to help me write my physics lab report if possible.

Again, I went through the same fine-tune procedure to test if the AI could speak like a cat lover. I collected interviews, speeches and captions that were related to human feelings about cats. As I typed in “I love cats”, this is what I received:

“I love cats, and there is nothing more to be said about my dear favorite. Cats are as romantic as love to me; I just always knew it. But really, it comes from what I think is a real place in people’s minds. I think there’s something so primal about a cat. I think that we all take ourselves a little too seriously sometimes. And it feels nice to be able to be a bit giddy on the inside for once.”

The result as you see, is quite intriguing. The AI not only expresses how adorable cats are, but even reflects on inner reasons why humans tend to love cats. It is a little bit horrifying for me at the same time because what if the AI could precisely predict human thoughts and emotions some time in the future? That does not sound real at all, but what if?

However, if we look at the positive aspect, the AI could contribute to so many fields. It could first of all, be used for education as for example, it could easily choose the best way to explain some academic jargons with metaphors and analogies. It could also be used for service industries, with the AI recognizing and analyzing the emotion of web-users and reacting correspondingly. Yet I still fear AI, knowing that one day, if it possesses self-studying ability, humans might be extinguished by its own creatures. But who knows? Maybe humans will be pushed to their limit in order to compete with AI and our intelligence will reach a whole new level.

I suppose we’ll just have to wait and see.