For 80 years, we have tried to make our computers more brain-like.

80 年來，我們一直試圖讓我們的計算機更像大腦。

I would like to show you that there is value in making brain organoids, so human cell cultures, more computer-like.

我想告訴大家，讓大腦器官（即人類細胞培養物）變得更像計算機是有價值的。

We call this brain organoid plus AI, organoid intelligence, or OI.

我們稱這種大腦類器官加人工智能為類器官智能，或 OI。

It is essentially combining, connecting AI with a brain organoid system and have them talk to each other.

從本質上講，它是將人工智能與大腦器官系統連接起來，並讓它們相互對話。

This all is possible because of the dramatic increases in the capabilities of AI.

這一切之所以成為可能，是因為人工智能的能力大幅提升。

Artificial intelligence is the synergy of three exponential growth curves.

人工智能是三條指數增長曲線的協同作用。

The first one is data.

第一個是數據。

The data in the world are doubling every 18 months, or roughly 90% of all data in the world have been produced in the last three years.

世界上的數據每 18 個月翻一番，也就是說，世界上大約 90% 的數據是在過去三年中產生的。

Computers are doubling every second year in capacity at half the price.

計算機的容量每兩年翻一番，而價格卻只有原來的一半。

That's what we know as Moore's law for 60 years.

這就是我們所熟知的摩爾定律，已有 60 年的歷史。

But AI algorithms, since the introduction of deep learning in 2010, double in capacity every three months, which means this year's large language model, for example, is about eight times as powerful as the one from last year.

但是，自 2010 年引入深度學習以來，人工智能算法的能力每三個月就會翻一番，這意味著，以今年的大型語言模型為例，其功能大約是去年的八倍。

This altogether is giving rise to a tremendous use of AI in all parts of life.

這一切正促使人工智能在生活的各個領域得到廣泛應用。

However, we also know that these large language models are not reasoning.

不過，我們也知道，這些大型語言模型並不是推理模型。

They are learning tons of things, and they can use these informations, but they are not actually reasoning.

他們學習了大量的知識，可以使用這些資訊，但實際上他們並沒有進行推理。

They are faking it.

他們是裝出來的。

However, our computer, the brain, just 1.4 kilogram on our shoulders, is actually quite a supercomputer.

然而，我們的計算機--大腦，肩上只有 1.4 千克的重量，實際上卻是一臺相當大的超級計算機。

It was only two years ago when the first time a supercomputer exceeded the computational capacity of a single human brain of estimated one exaflop.

就在兩年前，超級計算機的計算能力首次超過了估計為一個 exaflop 的單個人腦的計算能力。

This supercomputer was a $600 million installation on 6,800 square feet.

這臺超級計算機耗資 6 億美元，佔地 6800 平方英尺。

At the same time, the human brain consumes only about one millionth of energy as this supercomputer to the same computational work.

與此同時，人腦在進行同樣的計算工作時，所消耗的能量只有這臺超級計算機的百萬分之一。

This shows you that there's a lot of things we can learn, and it is actual intelligence.

這告訴你，我們可以學到很多東西，而且是實實在在的智慧。

It is not faked intelligence.

這不是偽造的情報。

And still, our little computer is doing many things better than even the best supercomputer.

儘管如此，我們的小電腦在很多方面甚至比最好的超級計算機做得更好。

For example, a child is able to distinguish between, let's say, a horse and a zebra with about 10 pictures each.

例如，孩子能用 10 張左右的圖片區分馬和斑馬。

A computer needs hundreds, if not thousands, of these pictures.

一臺電腦需要成百上千張這樣的圖片。

Then you show the child one single unicorn, and it will distinguish new unicorns from now on, while the computer needs again hundreds of these pictures.

然後，你給孩子看一隻獨角獸，它就會從現在開始分辨出新的獨角獸，而電腦又需要數百張這樣的圖片。

Our large language models are updated perhaps once per year.

我們的大型語言模型可能每年更新一次。

We learn progressively.

我們循序漸進地學習。

Whatever you take from my presentation today, you have integrated into your knowledge from now on.

無論你從我今天的演講中收穫了什麼，從現在起，你都將把它融入到你的知識中去。

We call this progressive learning.

我們稱之為漸進式學習。

We also have the ability to decide very quickly, fast thinking, intuition.

我們還擁有快速決策、快速思考和直覺的能力。

If we hear going through the jungle a noise, we are jumping, because we assume there could be a tiger.

如果我們聽到穿過叢林的聲音，我們就會跳起來，因為我們認為可能有老虎。

The answer from our chat models will be there after the meal.

我們的哈拉模特將在餐後給出答案。

We also have to see that real creativity is probably not yet there.

我們還必須看到，真正的創造力可能還沒有出現。

It is a rearrangement of human creativity, which we are observing.

這是我們正在觀察的人類創造力的重新佈局。

So let's talk about making brain organics more computer-like.

是以，讓我們來談談如何讓大腦有機體更像計算機。

Our cell cultures are actually keeping our brain very much in solitary confinement.

我們的細胞培養實際上是在把我們的大腦單獨監禁。

We are boring our cells to death.

我們的細胞快被悶死了。

They have nothing to do.

他們無事可做。

There is no input and no output to these systems.

這些系統沒有輸入，也沒有輸出。

Brain organoids, these balls of cells, which are derived from stem cells and are available now for about 10 years, are actually having an infrastructure of all the different cell types.

腦器質性組織，即這些細胞球，來源於幹細胞，問世已有10年左右，實際上擁有所有不同類型細胞的基礎設施。

They're having the neurons, all the different types of them.

他們有神經元，所有不同類型的神經元。

They're having oligodendrocytes, which are wrapping themselves around these axons of the neurons to allow faster connection.

它們擁有少突膠質細胞，這些細胞包裹著神經元的軸突，從而實現更快的連接。

And we also have the astrocytes, a very important cell type for learning and memory, because they are responsible to take away all of the many connections of the brain, which we do not need.

我們還有星形膠質細胞，這是一種對學習和記憶非常重要的細胞類型，因為它們負責把大腦中的許多連接都拿走，而我們並不需要這些連接。

And this is how memory establishes.

記憶就是這樣建立起來的。

They are pruning synapses.

它們正在修剪突觸。

So all together, they're forming a functional unit.

是以，它們共同組成了一個功能單元。

And this is what we want to use in a standardized way for reproducible reactions, mass-produced, and also, as I will show you in a minute, from different genetic backgrounds of patients in order to develop disease models for these.

這就是我們要以標準化方式用於可重複反應、批量生產的東西，而且，正如我稍後要向你們展示的那樣，這些東西來自不同遺傳背景的病人，以便為這些東西開發疾病模型。

In 2016, we were the first to mass-produce these type of organoids.

2016 年，我們率先批量生產了這類有機體。

And since then, we have used them for many different disease models.

從那時起，我們就把它們用於許多不同的疾病模型。

This is an article from the front page of the Financial Times at the time.

這是當時《金融時報》頭版的一篇文章。

So the first thing is we need to talk to these brain organoids.

是以，首先我們需要與這些腦器官對話。

So we need the type of EEG, an electroencephalogram.

是以，我們需要腦電圖的類型。

This is how it looks for a designer, for AI in this case.

這就是設計者的樣子，這裡指的是人工智能。

Down here, you see the actual EEG, which at the time, nine electrodes, which we published two years ago.

在這裡，你可以看到實際的腦電圖，當時有九個電極，我們在兩年前發表了這篇論文。

But we are increasing the number of electrodes around this mesh electrode.

但我們正在增加這個網狀電極周圍的電極數量。

So at the moment, we are at 16.

是以，目前我們有 16 人。

But you have to imagine that this brain organoid is only half a millimeter in size.

但你必須想象一下，這個大腦器官只有半毫米大小。

It is just having 50,000 neurons and 50,000 helper cells.

它只是擁有 5 萬個神經元和 5 萬個輔助細胞。

However, it is already at the moment undergoing training.

不過，它目前已經在接受培訓。

And it is controlling through Bluetooth, for example, this little robot, where we're trying to explore its environment and find roadblocks in the map it is establishing.

例如，它可以通過藍牙控制這個小機器人，我們正試圖探索它所處的環境，並在它繪製的地圖上找到路障。

But we have an entire suite of little computer games you can see here, which we call our mind gym.

但我們有一整套小電腦遊戲，你可以在這裡看到，我們稱之為 "頭腦健身房"。

And in this mind gym, we are using these type of video games to find out how well it performs.

在這個智力健身房裡，我們使用這些類型的視頻遊戲來了解它的性能如何。

And does it perform better and better over training periods and over the next day?

它在訓練期間和第二天的表現是否越來越好？

This is ongoing work.

這是一項持續性工作。

But we have the first indications that there's really long-term memory establishing.

但我們有初步跡象表明，長期記憶的建立確實存在。

Very, very simple functionalities, to be clear here.

說白了，就是非常非常簡單的功能。

At the same time, we're making these brains bigger.

與此同時，我們正在讓這些大腦變得更大。

We are trying to achieve a brain organoid of one centimeter, which requires us to perfuse it not with blood, but with cell culture media to get oxygen and nutrients to the center that we can work with a larger one.

我們正試圖實現一釐米的大腦器官模型，這就要求我們不是用血液灌注，而是用細胞培養基灌注，以便將氧氣和營養物質輸送到中心，這樣我們就可以用更大的大腦器官模型來工作了。

One centimeter is already twice a mouse brain, just to give you some type of proportion here.

一釐米已經是小鼠大腦的兩倍了，這只是給你一個比例。

And this is part of a surpass program, which Johns Hopkins University, my university, is financing.

這也是我所在的約翰霍普金斯大學資助的超越計劃的一部分。

This is opening up for a lot of ethical questions.

這引發了許多倫理問題。

Nature titled once, can lab-grown brains become conscious?

大自然》曾一度提出，實驗室培育的大腦能否有意識？

We don't know.

我們不知道。

But we have to prepare for the possibility of something like this.

但我們必須為可能發生這樣的事情做好準備。

So we are working very closely with ethicists.

是以，我們正在與倫理學家密切合作。

They are meeting with us every week when we plan our experiments.

在我們計劃實驗時，他們每週都會與我們會面。

They come to the labs.

他們來到實驗室。

They see what we are doing.

他們看到我們在做什麼。

We call this embedded ethics.

我們稱之為嵌入式倫理。

But they're also asking people on the street, what do you think about this type of research?

但他們也會在街上詢問人們，你對這類研究有什麼看法？

When does it start to feel perhaps better not?

什麼時候開始感覺好一些？

And does this change if we tell you that this is to develop, for example, drugs for Alzheimer's?

如果我們告訴你，這是為了開發治療阿爾茨海默氏症的藥物，這是否會有所改變？

Because this is one of the short-term perspectives of such a system.

因為這是這種系統的短期前景之一。

It will not beat AI as it is.

它無法擊敗人工智能。

It will probably not even beat a handheld calculator for the foreseeable future.

在可預見的未來，它甚至可能比不上手持計算器。

But it is something where we can build disease models, where we can try, if you take cells from a child with autism, are they communicating differently?

但是，我們可以建立疾病模型，嘗試從患有自閉症的兒童身上提取細胞，它們的交流方式是否有所不同？

Are those with their new developmental disorders incapable of learning?

那些患有新發育障礙的人是否沒有學習能力？

Is this changed if their cells were taken from an Alzheimer patient?

如果他們的細胞取自老年痴呆症患者，情況是否會發生變化？

So we are already preparing with various expert groups in Hopkins to produce brain organics of this type, and we are characterizing them, and we are hoping to employ organic intelligence.

是以，我們已經與霍普金斯大學的各個專家小組合作，準備生產這種類型的大腦有機物，我們正在對它們進行特徵描述，我們希望能夠運用有機智能。

At the same time, we are trying to form a community.

與此同時，我們還在努力組建一個社區。

We launched this just one and a half years ago with an article as the first article ever in the flagship journal Frontiers in Science.

就在一年半前，我們在旗艦期刊《科學前沿》上發表了一篇文章，成為有史以來的第一篇文章。

And we are trying to form a community.

我們正在努力組建一個社區。

We are trying to bring people together because it needs a multidisciplinary team.

我們正試圖將人們聚集在一起，因為這需要一個多學科團隊。

We have carried out a workshop, and instantly the White House has made this one of the bold biotechnologies goals of the U.S. to create these type of systems.

我們舉辦了一次研討會，白宮立即將此作為美國大膽的生物技術目標之一，以創建此類系統。

We have created a journal, a sub-journal of my journal, Frontiers in Artificial Intelligence, which is called Frontiers in Organic Intelligence now, and we have convinced the National Science Foundation to open a program, which is now open for Engineering Organic Intelligence, And we also created a newsletter, we created a website, and the community is at the moment 450 people.

我們創辦了一份期刊，是我的期刊《人工智能前沿》的子期刊，現在叫做《有機智能前沿》，我們還說服了美國國家科學基金會開設了一個項目，現在這個項目正在為 "工程有機智能 "開放。

It is free to join, and I would hope to see some of you do this.

加入是免費的，我希望看到你們中的一些人加入。

It is a remarkable scientific exercise.

這是一項了不起的科學工作。

I consider it to lift science fiction, and to show that indeed this is the fact, a friend of mine after one of my first presentations on Organic Intelligence said to me, Thomas, do you recall Star Trek in 1966?

我認為這只是科幻小說，為了證明這確實是事實，我的一位朋友在我關於有機智能的第一次演講後對我說：托馬斯，你還記得 1966 年的《星際迷航》嗎？

I didn't, I was three years old.

我沒有，當時我才三歲。

But in this episode, Captain Kirk finds this computer which is running on brains.

但在這一集中，柯克船長髮現了這臺依靠大腦運行的電腦。

This is one of the incredible visions of this project, to really understand better how a brain could drive a computer, how we possibly could create neuromorphic algorithms and architectures which are embracing biology as a driving principle for Organic Intelligence.

這也是這個項目令人難以置信的願景之一：真正更好地理解大腦如何驅動計算機，我們如何才能創造出神經形態算法和架構，將生物學作為有機智能的驅動原則。

Thank you very much.

非常感謝。