What "Overfitting" implies for our own learning practices.

I'm not a specialist in Machine Learning at all but interested in it and have read books about it.

人工知能はどのようにして 「名人」を超えたのか?―――最強の将棋AIポナンザの開発者が教える機械学習・深層学習・強化学習の本質

The Master Algorithm: How the Quest for the Ultimate Learning Machine Will Remake Our World

What interests me is the concept of "Overfitting."

Overfitting:https://en.wikipedia.org/wiki/Overfitting

When machine learning learns too much, it fails in predicting the future.

Machine learning is too good at memorizing the detail of everything that it fails in generalizing. Therefore, deep learning engineers intentionally cut some connections between layers that connect networks.

I think this has a great implication for our (human's) own learning practices.

Sometimes, we experience that hard working does not necessarily bring about good results.

Studying hard sometimes does not come along with good test results.

Working hard sometimes lowers productivity.

As I wrote before, machine learning is not only about machine but also about learning.

We don't normally question what learning is but in this age of machine learning, it's time to think about it.

Also, it's time to think about what it is that only humans can learn.

To think deeply; How, Why and What

I feel like many people are good at asking HOW questions.

How to make money?

How to learn a new language?

How to get married?

Then, we ask WHY?

Why do I need money?

Why do I have to learn a new language?

Why marry?

But to think deeply, I think we need more WHAT questions.

What is money?

What is language?

What is marriage?

Machine Learning is not about Machine but about Learning

It seems like many people think that machine learning is about machine.

But this book changed my view; machine learning is about learning.

The Master Algorithm: How the Quest for the Ultimate Learning Machine Will Remake Our World

To make machines learn new things, first we need to unpack the concept of learning.

What is learning? How do we learn?

We do not think about these questions often.

But it seems like these are the key questions to machine learning.

"Less is more" in workplaces

Now I work for a small startup.

A small startup always has resource constraint and needs to think about where to allocate scarce resource.

On the other hand, when I was working for a big transnational corporation, there was enough budget.

Having being working for a small startup for a while, I realized that resource constraint makes me think more than when the resource was enough.

When you have enough resource, you just use it to satisfy your demand.

But when you don't, you need to think a lot about how to achieve your goal.

I experienced this kinda situation when I was studying english abroad.

I did't have enough money to go to language school. So I had to think about how to learn English effectively without paying much money.

So I went to bars, bought a glass of beer and talked to random people. It was a good strategy. Language schools are full of non-english speakers but bars are full of native English speakers. I got to speak English with native English speakers every night.

Having less makes you think.

This is probably the biggest advantage of having less.

Side effect of stable society

The other day, when I was talking to my friend, she said

"I believe in money but not in people. Why? Because people might betray you but money does not."

I think this is reasonable.

But only when the value of money is relatively stable.

Since I live in Japan now, the value of money does not fluctuate to a large extent.

Most probably, the value of 10,000 yen remains the same after a year.

In this "stable" society, no wonder people believe in money rather than people.

Yes, people are generally susceptible to changes. Even you believe in a person today, he might betray you tomorrow.

But traveling around the world might remind you of another aspect of money; the value of money fluctuates.

When I was working in Nicaragua, Costa Rica and the Philippines, I encountered people who are not interested in believing in money at all.

I was wondering why.

I just thought that they are prone to present consumption rather than saving for the future.

But it might not be that simple.

They might have experienced drastic inflation in the past.

They might have experienced bank account blocking in the past.

Having been experienced these events, they don't believe in money. They believe in people.

Do you believe in money or in people?

No judgment.

That's your choice.

Epistemology in Workplaces

Donald Rumsfeld, when he was asked if there was an evidence of Iraq owning weapons of mass destruction, said,

"there are things that we know,

there are things that we know that we don't know

and there are things that we don't know that we don't know"

This is an interesting statement from the viewpoint of epistemology.

It seems like every knowledge falls into one of those categories.

We know the speed of light.

We don't know what exists outside the universe.

But we cannot tell what it is that we don't know that we don't know because we just don't know it.

I know that this kind of theoretical and philosophical argument is generally disliked because it is not useful. It does not generate money.

Right?

But working as a manager in a company, I find this framework of thinking quite useful.

By expanding the reach of known knowns, we can be experts.

But this is not good enough.

To be a good manager, we need to be aware of known unknowns too.

This is to say that we need to know what it is that we don't know.

Since the required knowledge for managers is broad, ranging from IT to law to marketing to finance, it is almost impossible to know all of them.

What's important is to know what you don't know and who knows what you don't know.

And the third category?

I think this has to do with our world view.

Though it is impossible to estimate the portion of this category by definition, if you think this area is not too big, you might have an organized world view.

Your world is filled up with known knowns and known unknowns.

With this world view, one might think that you can grasp the overall image of how the world works.

However, if you think that unknown unknowns share the majority of portion in the world, you might have a chaotic world view.

Nothing is predictable and everything can happen.

Which world view do I have?

Needless to say.

話の聞き方 その②

ずーーーっと前に、「話の聞き方」というタイトルで記事を書きましたが、なぜかふとその第２弾を思いついたので、久しぶりにこのテーマについて書きます。

さて、僕は社会人１年目の時に、上司に謝罪を求めた事があります。

それは飲み会で無理矢理飲まされそうになったからです。

僕はお酒を飲むのが好きですが、人からの強制を何より嫌う僕は、あえて「すみません飲めないんです」と言って断りました。（嫌なやつ）

まあ、何より、僕はその上司が嫌いだったので、単純に喧嘩を売りたかったというのもあった。（まじで嫌なやつ）

そしたら、酔っぱらってたその上司は「お前、客からの酒でも飲めないって言うつもりなのか？」と言って飲まそうとして来ました。まあ予想通りの展開です。笑

で、次の日、新人研修でアルコールハラスメントについて習った事を引き合いに出して「人にお酒を強要するのはいけないことだと思いますので、僕に謝罪してください」とその上司にメールしたところ（記録を残すため。笑　　最悪なやつ。笑）当時の部の人事担当者が僕を個室に呼び出して、

「君、上司に謝罪を要求するって一体どういう事かね？それじゃあ会社生活やっていけないよ。」

と怒ってました。

ーーーーーー

さて、その、僕を怒った人事担当者ですが、凄く真面目な人で、自分の上司の言う事は凄く真面目に聞いて、まあ、良い子ちゃんの典型みたいなサラリーマンでした。

でも、その人は、それから数ヶ月後、鬱病っぽくなってしまったらしいです。（原因が何かとか詳しくは知らないので、色々言う立場にはありませんが。）

僕は心の中で思いました。「会社生活やっていけないのはあなたでしたね。」

ーーーーーーーーーーーーーー

さて、この当時の人事担当者をサンプルにして、彼の頭の中を少し分析してみましょう。（こういう、ナメた「分析」がとても嫌われるのは百も承知ですが。笑）

この人事担当者はおそらく「自分は会社生活３０年、様々な事をやってきた。会社でどういう人が上手くやっていけるかよくわかっている。だから、この新人に自分の学んで来た事を教えてあげて、彼が上手くやっていけるように育ててやろう」と思っている。

とても良い人です。嫌みに聞こえるかもしれませんが、これは嫌みではなく、当時の会社は本当に良い人達ばかりで、新人を教育して立派な社会人にしてやろう、と思っている上司が多かったと思います。今でも繋がっている人がたくさんいます。

自分の理想とする社会人像Aがあり、そのAという像を体現すれば、皆社会で上手くやって行けると思っている。だからそれを新人に体現させようとする。

繰り返しますが、とても良い人です。

ただ、おめでたすぎる。というか、考えが浅過ぎるというしかないでしょう。地獄への道は善意で敷き詰められている、という言葉は知らなかったのでしょうか。

事実、彼はその後会社で上手くやっていけませんでした。鬱になっちゃったんだから。とても気の毒です。

さて、彼が彼の信念（Aという理想像を追い求めるという信念）に従って仕事をして失敗したので、それはそれで仕方ない（自己責任、という言葉を使うのは嫌ですが、まあそういうことでしょう）ですが、問題は、それを僕に押し付けようとしていたことです。

彼がそれ（Aという社会人の理想像）を僕に押し付けて、もし僕がそれを素直に聞いていて、Aを体現し、彼の様に鬱病になって、もし自殺でもしてしまっていたら、彼は責任をとってくれるのでしょうか？

とってくれるわけありません。というかとれません。死んでしまった人は少なくとも今の医療じゃ生き返らせられないので。

そういう意味で、彼は極めて無責任と言うことができるでしょう。

結局、自分の理想を人に押し付けるだけのマスタベーション人事担当者だった、と言われても仕方ないでしょう。

（＊注１　ただし、全ての教育は価値観の押しつけである、と考える事もできるので、かれは仕事として仕方なくやっていた、と考える事もできます。例えば、算数を教える、という学校教育は価値観の押しつけと考える事もできます。算数を使わない生き方もあるので、それを無理矢理教えてその点数によって人を評価するなんてひどい、という考え方もできなくはないということです。）

（＊注２　彼が鬱病っぽくなってしまった原因を僕は何もしりません。上の議論は、彼の社会人としての理想像が彼の鬱病の発症に寄与した、という仮定のもと書いています。もしこの仮定が完全に間違いである場合（彼の社会人としての理想像と鬱病の発生の間には何も因果関係がない場合）、上の議論は全く成り立ちません。）

ーーーーーーーーーーーーーーーー

人の話を良く聞く事は良い事だとされています。

ただし、上の例からも分かる様に、「人の話を聞いて理解すること」と、「人の話をすべて鵜呑みにすること」はまったく違います。
それは前の記事でも書いた通りです。

人の話を良く聞き、完全に理解し、それのどこが間違っているのか、どこが自分とは相容れないのかをいつも考えるという思考のクセが大事だと思います。