Hinton still hasn't figured out how brain works, but he got the Nobel Prize for Physics for AI. That proves ppl still have no clue what is intelligence and the mechanics of it.
Mind and brain, the former is software and the latter hardware, I suppose so. How they form will to drive mind going through all kinds of sufferings to reach the goal of the mind set for is what we call computation. I still believe that when a being striken by accidents to be poor vegetable and lying in bed can still see god to the realms that is Stephen Hawking says in his the Theory of Everything 😊
Boltzmann machine are pattern which's GPU of AL, backpropagation are induction in number theory which common language is number, for example n*(n+1)/2=55 add from 1 to 10, have such pattern AI can predict any sum from 1 to n by backpropagation[induction], for prime counting function p(2^2)=4-(4-0)*(1/2)+1-1=2 from traditional sieve of prime 2, can transform into p(2^2)=4*(1/2)+0/2=2 by Boltzmann machine because have same pattern, 0/2=mod(4,2)/2, deduce p(5)=5*(1/2)+1/2=3, p(6)=6*(1/2)+0/6=3, p(7)=7*(1/2)+1/2=4, p(3^2-1)=8*(1/2)+0/2=4, start at 3^2 must add sieve of 3 by backpropagation, p(3^2)=9-(9-1)*(1/2)-(9-0)*(1/3)+(9-3)/6+2-1=4=9*(2-1)*(3-1)/(2*3) + (1/2-3/6+0/3)+1...etc to p(5^2-1)=24*(1/3)+(0/2-0/6+0/3)+1=9, start 5^2 add sieve of 5, p(31)=31*(4/15)+(1-4/15)+2=11, 1-4/15=11/15=4/15+4/15+3/15 is sum of zero of zeta function Z(1)=llp/(p-1), 4/15=(2-1)*(3-1)*(5-1)/(2*3*5), (4/15)/(2-1)=4/15=1/2-1/6-1/10+1/30 : 1st zero which correspond 14.134725 at higher number, (3-1)*(5-1)/(3*5)/(3-1)=4/15 =1/3-1/15: 2nd zero[21.02...]. (5-1)/5/(5-1)=3/15 : 3rd zero[25.01..], by Boltzmann machine AI can deduce zero all start at pn^2[2^2, 3^2, 5^2...] that prove RH by induction[backpropagation] from x^(1/2)=e^((1/2)*logx), by Boltzmann machine can deduce all zero on line 1/2 of x axis on complex plane which's different language but have same pattern[Boltzmann machine].
The Royal Swedish Academy of Sciences said it awarded the prize to the two men because they used "tools from physics to develop methods that are the foundation of today's powerful machine learning" that is "revolutionising science, engineering and daily life." 瑞典皇家科學院表示,之所以將這一獎項授予這兩個人,是因為他們使用“物理學工具開發了為當今強大的機器學習奠定基礎的方法”,“徹底改變了科學、工程和日常生活”。 src: - Nobel physics prize 2024 won by AI pioneers John Hopfield and Geoffrey Hinton
謝謝您的字幕與翻譯 這是一段啟發性很高的影片.
7:42 翻譯錯誤: 「還有技術能力 but also the skill」 應為 「還有規模 but also the scale」,後面那句也是。
Thats the most romantic 45 minutes that i ever had… 👏👏👏👏 thank you for this beautiful clip
太感謝中文翻譯啦❤
看到你的评论才发现有中文字幕。。。
相當珍貴的對談紀錄~人類AI革命正在開始
感謝fox分享,從頭到尾都好讚
感謝分享
信息量很大,谢谢博主 好人一生平安啊
感謝精彩分享,辛苦了!
能看到听到这种视频真好,真好。
感謝提供
這完全符合我對卡內基的印象,美國高等學府都是一群瘋起來不要命的人。卡內基更是當中學風最誇張的學校之一
头脑如此活跃,身体如此硬朗,看来有机会再拿诺奖。讲的真好啊❤
他是站著在回答問題的。
謝謝你
腦神經是個體的,想的都是局部的事情,還有和經驗,情緒聯繫在一起,推論有限的將來。而意識是宇宙性的,是廣大無垠的。
Embedding 是指把 data 轉換成向量並嵌入到一個向量空間裡的意思。
great!
1973年辛頓就發現機器也有情緒反應
太刺激了🤩
we need a method which can learn without label
采访Geoffrey Hinton,让我想起了研究原子弹的奥本海默。
但是AI比原子弹要恐怖,未来机器人即将代替人类,有可能会像电影里那样。可能会出现机器人怎么和平相处问题。
还有一本书,叫《最后一人类》,美国的科学家一直在研究,怎么讲人类的大脑拷贝下来,当人死了之后,讲拷贝下来的大脑,移植到计算机上,这样就能达到永生。
不会有比原子弹这种瞬间毁灭一切更恐怖的东西,当一切毁灭了,那就都没有意义了
Triple 應該是指 bi-gram model?
下一步 是 算力功耗了
普通硬體計算架構 無法支撐指數成長的模型
要突破, 就需要再 類似量子電腦的新運算技術, 有重大突破才可進入下個技術爆發點
还是没有达到人脑的高效,人脑也没有消耗这么多的能量和神经元,在这个角度突破才是真正的进化
很好奇Geoffrey提到的Donald Hebb 跟 John von Neumann的書名
后面的图书 是摆设还是摆着好看
人類自詡靈魂,是 生物體,執行強大“運算” 的錯覺嗎?
或者同上問,若能達到原子等級100%的大腦複製,靈魂也一併複製了嗎??
電子級的複製
好吧,隨著 AI 進步,學習和“思考” 都會越來越強大,並且超越人類。
那,我們得到的AI機器,會是更像欣梅爾,還是芙莉蓮 ?
But how to let people have a good command of Ai should be concerned
20:25 他相信的language到底如何實作?
45 minutes standing…
That is how long a professor stand giving lecture.
15:43
AI目前是藉由 語言 仿生學習的過程,並展現智力或智慧 嗎?
不同語言的人類語言,會生出不同“性格”的AI 嗎? AI 在未來,會因為現有語言的字詞缺乏,而創造出新字詞。或AI會創造全然不同的人類語言的AI語言嗎??
只有我覺得他們站太久了嗎?
久坐不好(X)
Hinton 有慢性背痛,坐不了多久
@@ivantsai5209原來如此,感謝解惑!
你是会开小差的
也才不到一小時耶
大師就是不一樣
Hinton still hasn't figured out how brain works, but he got the Nobel Prize for Physics for AI. That proves ppl still have no clue what is intelligence and the mechanics of it.
好奇Netflix推薦給Geoffrey Hinton的片單都是些什麼影片造成選擇障礙
太多了,很多人都有同样的困境
雖然有翻譯,但還是完全聽不懂.
你要单独去理解科学家说的专业术语。说的专业术语就像数据一样,大家都听不懂。
你明白,AI通过不断的学习,可能会比人类还要聪明。
目前人類還在改進馮紐曼架構,等到用矽光子解決記憶體瓶頸後,電腦還會加快一百倍,訓練能力會更強,創造力肯定是能超越人類
Mind and brain, the former is software and the latter hardware, I suppose so. How they form will to drive mind going through all kinds of sufferings to reach the goal of the mind set for is what we call computation. I still believe that when a being striken by accidents to be poor vegetable and lying in bed can still see god to the realms that is Stephen Hawking says in his the Theory of Everything 😊
cucumber!
Boltzmann machine are pattern which's GPU of AL, backpropagation are induction in number theory which common language is number, for example n*(n+1)/2=55 add from 1 to 10, have such pattern AI can predict any sum from 1 to n by backpropagation[induction], for prime counting function p(2^2)=4-(4-0)*(1/2)+1-1=2 from traditional sieve of prime 2, can transform into p(2^2)=4*(1/2)+0/2=2 by Boltzmann machine because have same pattern, 0/2=mod(4,2)/2, deduce p(5)=5*(1/2)+1/2=3, p(6)=6*(1/2)+0/6=3, p(7)=7*(1/2)+1/2=4, p(3^2-1)=8*(1/2)+0/2=4, start at 3^2 must add sieve of 3 by backpropagation, p(3^2)=9-(9-1)*(1/2)-(9-0)*(1/3)+(9-3)/6+2-1=4=9*(2-1)*(3-1)/(2*3) + (1/2-3/6+0/3)+1...etc to p(5^2-1)=24*(1/3)+(0/2-0/6+0/3)+1=9, start 5^2 add sieve of 5, p(31)=31*(4/15)+(1-4/15)+2=11, 1-4/15=11/15=4/15+4/15+3/15 is sum of zero of zeta function Z(1)=llp/(p-1), 4/15=(2-1)*(3-1)*(5-1)/(2*3*5), (4/15)/(2-1)=4/15=1/2-1/6-1/10+1/30 : 1st zero which correspond 14.134725 at higher number, (3-1)*(5-1)/(3*5)/(3-1)=4/15 =1/3-1/15: 2nd zero[21.02...]. (5-1)/5/(5-1)=3/15 : 3rd zero[25.01..], by Boltzmann machine AI can deduce zero all start at pn^2[2^2, 3^2, 5^2...] that prove RH by induction[backpropagation] from x^(1/2)=e^((1/2)*logx), by Boltzmann machine can deduce all zero on line 1/2 of x axis on complex plane which's different language but have same pattern[Boltzmann machine].
Sit down, please
跟数学、生物学、计算机科学有关,跟物理学有什么关系?
囧. 可能是物理界 沒有什麼大突破, 加上沒有諾貝爾數學 電腦科學獎
我來回答你的問題吧,數學初階學會知道0、1、2、3、4、5、6、7、8、9這是初階知識,還記得座標四大象限嗎?,下一階加法跟減法這是使用方法(可以看成是其中一個象限),下一階微分積分(這裡開始就是藝術抽象學,如果你真的懂的話,微積分看起來就是一個跟生活有關的數學模式,只是公式裡面帶入生活這個未知數變化是無窮盡,所以我只會講到這),接下來這裡我只會講到科學研究跟天文學恆星運作,這裡就是涉及更大尺度的運算,所以也就到這裡停了。
接著到生物學,生物學裡面從細胞、微生物、器官跟系統、系統組合運動學(這裡可以指任何運動包含物理學、男女交配做愛運動學、物與物的運動關係等等,太多能講解的,我就是不多說),生物學核心就是理解生命與系統運動的關係。
所以基本上大部分的知識學習都可以跟人工智能或人工智慧扯上關係,因為人工智慧就是把所有的學科跟經驗合併在一起,並且淘汰掉不適合的,讓最大效益化不斷持續並且持續產生,以達到所謂的永久不間斷的生命條件,但是這是物理方面的,精神方面是無法達到,因為我本人靈魂出竅過,靈魂離開肉體是感覺很輕的就好像沒有感覺似的,然後對肉體這個阻礙反而很討厭回到這裡,我先講到,如果你有什麼不理解,我在看看要怎麼講解你聽的懂的方式。
我也不是一開始就什麼都懂,也是慢慢花時間理解知識,然後花時間理解科學、哲學、神學,並且最後所有的盡頭的學科就是神學,本人沒有宗教、信仰,從來不拜神、佛、道、寺廟、鬼神等等,見識過另外空間的釋迦牟尼佛,確實真實存在,包含造物主也是見識過並且存在著,地球現在就是被造物主掌控著,它的神力不是你們普通人能理解的,還有科學最終是會到盡頭的...,信不信就是你們人類的事了
大概是各個領域終究是殊途同歸,今年正好是那個時刻的到來。
The Royal Swedish Academy of Sciences said it awarded the prize to the two men because they used "tools from physics to develop methods that are the foundation of today's powerful machine learning" that is "revolutionising science, engineering and daily life."
瑞典皇家科學院表示,之所以將這一獎項授予這兩個人,是因為他們使用“物理學工具開發了為當今強大的機器學習奠定基礎的方法”,“徹底改變了科學、工程和日常生活”。
src:
- Nobel physics prize 2024 won by AI pioneers John Hopfield and Geoffrey Hinton
14:00
11:31
31:38