本期內容是關于圖靈測試的觀點與評論，如果不了解圖靈測試可以移步第一期文章認知心理學 (中文) 1 - 什么是智能？誰能夠思考？。

復雜認知/認知心理學系列文集中文版文集！我們如何變得如此聰明？我們是如何感知周圍的世界、學習語言、制定決策、記得過去、并預測未來的？大量的神經組織如何思考？怎么樣產生想法？這些都是本系列專欄將嘗試回答或啟發(fā)思考的問題。

本系列主要分為三種內容：第一種是傳統(tǒng)的講座形式的知識內容，涵蓋不同的概念與例子；第二種是體驗不同的認知科學實驗；第三種是不定期地對丹尼爾·卡尼曼的《思考，快與慢》這本心理學著作的閱讀與討論。

本期思考：圖靈測試是回答“機器可以思考嗎 Can machines think?”這個問題的好方法嗎？為什么或者為什么不？

值得注意的是，認知科學家們在這些問題上存在很大的分歧。每個人的觀點都可能有各自的依據，這些問題并沒有“標準答案”。

閱讀下面的不同人的評論，思考上面的問題，并注意：

- 您在閱讀的過程中有什么讓你感到驚訝？出現了哪些主題？

- 您閱讀這些評論時有何反應？有什么有趣，令人驚訝，令人困惑的地方？

?

?

? 人們都持什么觀點？Where do we stand?

–“我對圖靈測試及其評估計算機是否被認為正在思考的能力感到非常矛盾。我認為該測試有很多優(yōu)點，并且可以與人類抗衡，但是在了解了可以通過圖靈測試的人工智能系統(tǒng)之后，我不確定我是否會考慮它在“思考”。（意思是這位朋友認為有的通過圖靈測試的人工智能系統(tǒng)很瓜批。）

– “I feel very conflicted about the Turing test and its ability to assess whether or not a computer is considered thinking. I think that it has a lot of merit and that it makes sense to put it up against humans, but after reading about the AI system that has arguably passed the Turing test, I am not sure that I would consider it thinking.”

?

–“我相信計算機的工作永遠無法與人類的思想相提并論，而機器也永遠無法去“思考”，這是一種種類似于人類的方式?！?/span>

– “I don’t believe that the work of computers can ever parallel that of human thought and machines will never be able to “think” is a way that can resemble a human.”

?

–“我認為圖靈測試是回答以下問題的好方法：‘機器可以證明它們與人類相似嗎？’”

– “I think the Turing Test is a good way to answer the question, ‘Can machines prove they are similar to humans?’”

?

–“圖靈測試在測試機器方面的表現非常出色，它能夠給出滿足人類受眾的答案?！?/span>

– “The Turing test does a really good job testing machines’ ability to give the answer that will satisfy human audiences.”

?

–“我覺得圖靈測試不是回答機器可以思考的好方法。圖靈測試模仿了智能的行為，但是機器實際上不是智能的?！?/span>

– “I feel as though that the Turing Test is not a good way to answer if machines can think. The Turing Test mimics the behaviour of intelligence but that machines are not actually intelligent.”

?

–“我將思維定義為使用理性或邏輯判斷來理解和推理某些事物。”

– “I define thinking as using rational or logical judgement in order to comprehend and reason with something.”

?

–“我個人認為，如果在特定情況下使用，圖靈測試可能是一個不錯的指標。 神經網絡的概念是我們未多講的一項人工智能技術。據我了解，他們從各種刺激中‘學習’，并可以根據以前的經驗創(chuàng)建處理網絡，以對新刺激做出反應?！?/span>

– “I personally think that the Turing test can be a decent metric if used in specific circumstances. One piece of AI tech that we’ve not spoken a lot about is the concepts of neural networks. From what I understand, they ‘learn’ from various stimuli and can create networks of processing to respond to new stimuli based on previous experiences.”

?

–“圖靈測試是一種確定計算機是否可以思考的好方法，因為該測試涵蓋了不同類型的詢問。諸如亞馬遜Alexa或蘋果產品的Siri之類的新技術中的人工智能正在演變?yōu)閷⑷祟惻c計算機區(qū)分開來的更像人類的功能?！?/span>

– “The Turing Test is an okay way of finding out if computers can think because the test covers different types of interrogation. The AI in newer technology, such as the Amazon Alexa or Apple’s Siri, is evolving into those more human-like features that distinguish humans from computers.”

?

–“盡管一臺機器通常缺乏自由意志并且無法創(chuàng)造自己的思想，但我認為使用代碼成為人類冒名頂替者的戰(zhàn)略行為（因此欺騙詢問者）確實在某種程度上暗示了思想?！?/span>

– “Although a machine often lacks free will and is unable to create its own thoughts, I would argue that the strategic act of using code to become a human imposter (thus fooling the interrogator) does suggest thought to some degree.”

?

–“我很喜歡圖靈的論文中，有一段內容解釋了人們在將人工智能與我們的智能進行比較時如何如此輕描淡寫，因為我們不希望任何東西威脅到我們自己對人類優(yōu)于所有其他人類的看法。當我繼續(xù)思考這個問題時，我將不得不進行自我檢查，并確保我不會讓這種偏見阻止我真正考慮所有可能性。”

– “I really enjoyed the part in the reading which explained how people are so dismissive when comparing artificial intelligence to our own intelligence because we don’t want anything to threaten our own perception of humankind being superior to all other beings. As I continue to ponder this question, I will have to check myself and make sure that I am not letting this bias prevent me from truly considering all of the possibilities.”

?

–“我認為圖靈測試可以幫助回答‘機器可以思考嗎？’這個問題，但這不應是唯一的測試?！?/span>

– “I think the Turing test can aid in answering the question, ‘Can machines think?’, but it should not be the only test.”

?

–“圖靈測試是回答機器是否可以思考的好方法。我們知道其他人的想法的唯一原因是，他們做著并說出使我們推斷出他們可以思考的事情。圖靈測試一旦通過，就會欺騙人們相信機器是人類，因為他們認為與之交流的事物像人類一樣在思考。”

– “The Turing Test is a good way to answer if machines can think. The only reason we know other humans think is that they do and say things that make us deduce that they can think. The Turing Test, if passed, tricks a human into believing a machine is a human because they think the thing they are communicating with is thinking like a human.”

?

?

? 人與計算機之間有相似之處There are similarities between humans and computers.

?

–“人類和計算機的想法相似，因為我們都從過去的經驗中學到東西。機器與人一樣，會從獎懲中學習，而經驗則可以指導他們的決策?！?/span>

– “Humans and computers think similarly because we both learn from past experiences. Same as human, a machine learns from rewards and punishment, and the experience informs their decision making.”

?

–“從最基本的意義上講，人與計算機的想法是相同的：它們從外界接收刺激和信息，然后根據一組規(guī)則進行處理和響應。”

– “In the most basic sense, humans and computers think the same: they receive stimuli and information from the outside world, then process and respond to it according to a set of rules.”

?

–“機器在某種程度上確實像人一樣思考。人們獲得知識，評估他們認為正確的答案或正確的事情并采取行動。我們傾向于將相似性或差異性的信息分組，將信息組織成層次或模式，得出不同主題和思想之間的結論和聯(lián)系。”

– “Machines, in a way, do think like humans. Humans gain knowledge, assess what they think is the right answer or thing to do and act on it. We tend to group information on similarities or differences, we organize information into levels or schemas, we draw conclusions and connections between different topics and ideas.”

?

?

? 人與計算機之間存在差異There are differences between humans and computers.

?

–“但是，只有人類有學習的欲望。發(fā)展與變化是我們固有的，是有機的?！?/span>

– “However, only humans have a desire to learn. Development and change is inherent to us, being organic and all.”

?

–“我同意馬庫斯的觀點，他認為機器與人類不同，因為機器主要依靠‘模式識別’，而不是真正的理解。人類的思維也會因情感而變得扭曲。例如，如果我們感到沮喪或饑餓，我們會失去精力，因此可能會屈服于我們的第一個想法（消耗最少腦力的想法）。電腦沒有情感，因此他們的思維不會受到干擾?！?/span>

– “I agree with Marcus, who says that machines differ from humans in that they rely mostly on ‘pattern recognition,’ rather than true understanding. Human thinking can also become twisted by emotion. For instance, if we are frustrated or hungry, we lose energy and therefore may succumb to our first thought (one that saps the least amount of brain power). Computers don’t have emotions, therefore their thinking goes undisrupted.”

?

–“計算機與人類思維之間的區(qū)別在于，人類在思考時具有自主權，這意味著人類可以主動地隨意思考。而計算機則不然，他們的編程決定（或限制）了他們的想法?！?/span>

– “The difference between computers and humans thinking is that humans have self-autonomy when thinking, meaning humans can be random on their initiative. However, computers are limited; their programming decides their thinking.”

?

–“我認為計算機不可能創(chuàng)造出像真實人類一樣的新事物。但是，如果我們可以對一個系統(tǒng)進行編程，讓計算機學習新事物并運用他們所學到的知識，則是很有可能的?！?/span>

– “I think it is impossible for computers to create new things like real human beings; however, it’s highly possible if we could program a system to let computers learn new things and apply the knowledge they learned.”

?

–“即使我們知道某種行為可能會導致懲罰，我們仍會考慮導致我們繼續(xù)進行的這種行為，因為盡管一再被警告，我們可能會堅持認為它是正確的行動方針。而機器是基于邏輯和被告知正確的思想來思考的。”

– “Even if we know a certain behavior may lead to a punishment, we still think about that behavior which leads to us going through with it due to our insistence of it being the right course of action or pure curiosity on the outcome despite being repeatedly told no. Machines think based on logic and based on what it is told is right.”

?

–“（人與機器）主要區(qū)別在于兒童可以成長為形成自己的意見，并選擇拒絕接受所學的內容的個體。計算機無法這樣做。我認為這是計算機工作方式與人類思維方式之間的主要區(qū)別，是人們能夠結合情感和過去的經驗來改變其行為的能力?！?/span>

– “The main difference is that children can grow to form their own opinions and choose to reject what they have been taught. A computer cannot do so. I think that is the main difference between the way computers work and human thinking, the ability to draw on emotions and past experiences together to change their behavior.”

?

–“人工智能可能能夠模擬對話響應或開個玩笑，但仍無法根據個人經驗來綜合信息。因此，盡管通過了圖靈測試，它仍無法像我們一樣‘思考’?！?/span>

– “An artificial intelligence may be able to emulate a conversational response, or tell a joke, but it is not yet able to synthesize information from personal experience. Therefore, although the Turing Test has been successfully beaten, it is not yet able to ‘think’ as we do.”

–“也許若重新評估了測試的參數，如文章所建議的那樣；也許若我們可以確定機器是否可以學習并分析其所知道的知識，那將是一個更好的測試?！?/span>

– “Maybe if the parameters of the test are reevaluated, like how the article suggests that maybe if we can determine whether a machine can learn and analyze what it knows then that would possibly be a better test.”

?

–“我相信元認知（元認知是指對‘認知’的認知。例如對自己的各種認知活動進行積極的監(jiān)控和調節(jié)；對自己的感知、記憶、思維等認知活動本身的再感知、再記憶、再思維就稱為元認知）對于能夠真正思考至關重要。如果沒有某種自我認知或意識，元認知是不可能的，因為如果不先意識到自己有大腦，就無法考慮大腦的運作方式。

– “I believe metacognition is vital to being able to truly think. This is impossible without some semblance of a consciousness/awareness of self, as you cannot contemplate how your brain is functioning without first being aware that you have a brain.”

?

–“對于人類，我們遵循大腦中養(yǎng)成的思維習慣。很難說我們的情報是否真的比計算機復雜得多?！?/span>

– “For humans, we follow the thinking habits developed in our brains. It’s hard to say whether our intelligence is genuinely far more sophisticated than computers’.”

?

?

閱讀完上面的不同評論，再次思考：

- 圖靈測試是回答“機器可以思考嗎Can machines think?”這個問題的好方法嗎？為什么或者為什么不？

- 您閱讀這些評論時有何反應？有什么有趣，令人驚訝，令人困惑的地方？

?

思考過后，您可以嘗試寫一篇短小的反思文章：

您認為圖靈測試是回答“機器可以思考嗎”這個問題的好方法嗎？為什么？人與計算機的“思維”在哪些方面相似？它們有什么不同？

值得注意的是，認知科學家們在這些問題上存在很大的分歧。每個人的觀點都可能有各自的依據，這些問題并沒有“標準答案”。

本期內容到結束，感謝閱讀！期待更多中文版專欄請多多支持UP主喔~