1 00:00:01,100 --> 00:00:04,519 covering the week's top textbooks like 2 00:00:04,519 --> 00:00:04,529 covering the week's top textbooks like 3 00:00:04,529 --> 00:00:07,460 covering the week's top textbooks like Linux bias google has released the code 4 00:00:07,460 --> 00:00:07,470 Linux bias google has released the code 5 00:00:07,470 --> 00:00:08,900 Linux bias google has released the code for their internally developed 6 00:00:08,900 --> 00:00:08,910 for their internally developed 7 00:00:08,910 --> 00:00:11,690 for their internally developed artificial intelligence tapas they can 8 00:00:11,690 --> 00:00:11,700 artificial intelligence tapas they can 9 00:00:11,700 --> 00:00:13,280 artificial intelligence tapas they can take a natural language question such as 10 00:00:13,280 --> 00:00:13,290 take a natural language question such as 11 00:00:13,290 --> 00:00:15,140 take a natural language question such as what's the name of the latest iPhone and 12 00:00:15,140 --> 00:00:15,150 what's the name of the latest iPhone and 13 00:00:15,150 --> 00:00:17,210 what's the name of the latest iPhone and fetch the answer from a relational 14 00:00:17,210 --> 00:00:17,220 fetch the answer from a relational 15 00:00:17,220 --> 00:00:19,460 fetch the answer from a relational database or spreadsheet and it's now 16 00:00:19,460 --> 00:00:19,470 database or spreadsheet and it's now 17 00:00:19,470 --> 00:00:21,830 database or spreadsheet and it's now open source the search giant's 18 00:00:21,830 --> 00:00:21,840 open source the search giant's 19 00:00:21,840 --> 00:00:24,070 open source the search giant's researchers detailed the AI on Thursday 20 00:00:24,070 --> 00:00:24,080 researchers detailed the AI on Thursday 21 00:00:24,080 --> 00:00:26,839 researchers detailed the AI on Thursday tapas is based on Bert a natural 22 00:00:26,839 --> 00:00:26,849 tapas is based on Bert a natural 23 00:00:26,849 --> 00:00:28,820 tapas is based on Bert a natural language processing technique that 24 00:00:28,820 --> 00:00:28,830 language processing technique that 25 00:00:28,830 --> 00:00:31,030 language processing technique that Google uses in its search engine a 26 00:00:31,030 --> 00:00:31,040 Google uses in its search engine a 27 00:00:31,040 --> 00:00:32,990 Google uses in its search engine a sizable portion of the world's 28 00:00:32,990 --> 00:00:33,000 sizable portion of the world's 29 00:00:33,000 --> 00:00:35,120 sizable portion of the world's information is relational that is to say 30 00:00:35,120 --> 00:00:35,130 information is relational that is to say 31 00:00:35,130 --> 00:00:37,540 information is relational that is to say organized into rows and columns 32 00:00:37,540 --> 00:00:37,550 organized into rows and columns 33 00:00:37,550 --> 00:00:39,709 organized into rows and columns navigating from these rows and columns 34 00:00:39,709 --> 00:00:39,719 navigating from these rows and columns 35 00:00:39,719 --> 00:00:41,709 navigating from these rows and columns historically required either manually 36 00:00:41,709 --> 00:00:41,719 historically required either manually 37 00:00:41,719 --> 00:00:44,840 historically required either manually shift lifting through a spreadsheet or 38 00:00:44,840 --> 00:00:44,850 shift lifting through a spreadsheet or 39 00:00:44,850 --> 00:00:47,540 shift lifting through a spreadsheet or writing SQL queries natural language 40 00:00:47,540 --> 00:00:47,550 writing SQL queries natural language 41 00:00:47,550 --> 00:00:49,430 writing SQL queries natural language processing makes the task considerably 42 00:00:49,430 --> 00:00:49,440 processing makes the task considerably 43 00:00:49,440 --> 00:00:51,380 processing makes the task considerably easier for users which is why the 44 00:00:51,380 --> 00:00:51,390 easier for users which is why the 45 00:00:51,390 --> 00:00:53,660 easier for users which is why the technology has been extensively adopted 46 00:00:53,660 --> 00:00:53,670 technology has been extensively adopted 47 00:00:53,670 --> 00:00:55,369 technology has been extensively adopted by Google and other players in the 48 00:00:55,369 --> 00:00:55,379 by Google and other players in the 49 00:00:55,379 --> 00:00:59,330 by Google and other players in the analytics market the search giant says 50 00:00:59,330 --> 00:00:59,340 analytics market the search giant says 51 00:00:59,340 --> 00:01:01,400 analytics market the search giant says that the toughest beats or matches the 52 00:01:01,400 --> 00:01:01,410 that the toughest beats or matches the 53 00:01:01,410 --> 00:01:03,500 that the toughest beats or matches the three top open-source algorithms for 54 00:01:03,500 --> 00:01:03,510 three top open-source algorithms for 55 00:01:03,510 --> 00:01:06,170 three top open-source algorithms for parsing relational data they train the 56 00:01:06,170 --> 00:01:06,180 parsing relational data they train the 57 00:01:06,180 --> 00:01:08,539 parsing relational data they train the AI on 6.2 million tables from the 58 00:01:08,539 --> 00:01:08,549 AI on 6.2 million tables from the 59 00:01:08,549 --> 00:01:10,700 AI on 6.2 million tables from the English version of Wikipedia and then 60 00:01:10,700 --> 00:01:10,710 English version of Wikipedia and then 61 00:01:10,710 --> 00:01:12,859 English version of Wikipedia and then set it to work on a trio of academic 62 00:01:12,859 --> 00:01:12,869 set it to work on a trio of academic 63 00:01:12,869 --> 00:01:15,469 set it to work on a trio of academic datasets benchmark tests that showed 64 00:01:15,469 --> 00:01:15,479 datasets benchmark tests that showed 65 00:01:15,479 --> 00:01:16,820 datasets benchmark tests that showed that the neural network provides 66 00:01:16,820 --> 00:01:16,830 that the neural network provides 67 00:01:16,830 --> 00:01:18,890 that the neural network provides accurate comparable answers as the rival 68 00:01:18,890 --> 00:01:18,900 accurate comparable answers as the rival 69 00:01:18,900 --> 00:01:22,070 accurate comparable answers as the rival algorithms across all three data sets 70 00:01:22,070 --> 00:01:22,080 algorithms across all three data sets 71 00:01:22,080 --> 00:01:24,649 algorithms across all three data sets the type of language processing google 72 00:01:24,649 --> 00:01:24,659 the type of language processing google 73 00:01:24,659 --> 00:01:26,929 the type of language processing google has implemented into tapas allows the AI 74 00:01:26,929 --> 00:01:26,939 has implemented into tapas allows the AI 75 00:01:26,939 --> 00:01:29,090 has implemented into tapas allows the AI to consider not only the question posed 76 00:01:29,090 --> 00:01:29,100 to consider not only the question posed 77 00:01:29,100 --> 00:01:31,190 to consider not only the question posed by users in the data they wish to query 78 00:01:31,190 --> 00:01:31,200 by users in the data they wish to query 79 00:01:31,200 --> 00:01:33,980 by users in the data they wish to query but also the structure of the relational 80 00:01:33,980 --> 00:01:33,990 but also the structure of the relational 81 00:01:33,990 --> 00:01:37,219 but also the structure of the relational tables in which the data is stored tapas 82 00:01:37,219 --> 00:01:37,229 tables in which the data is stored tapas 83 00:01:37,229 --> 00:01:39,050 tables in which the data is stored tapas can go beyond just fetching data and 84 00:01:39,050 --> 00:01:39,060 can go beyond just fetching data and 85 00:01:39,060 --> 00:01:42,050 can go beyond just fetching data and also perform basic calculations for 86 00:01:42,050 --> 00:01:42,060 also perform basic calculations for 87 00:01:42,060 --> 00:01:44,240 also perform basic calculations for example if a business user evaluating 88 00:01:44,240 --> 00:01:44,250 example if a business user evaluating 89 00:01:44,250 --> 00:01:46,280 example if a business user evaluating sales data asked for the average revenue 90 00:01:46,280 --> 00:01:46,290 sales data asked for the average revenue 91 00:01:46,290 --> 00:01:47,749 sales data asked for the average revenue across their company's three most 92 00:01:47,749 --> 00:01:47,759 across their company's three most 93 00:01:47,759 --> 00:01:50,240 across their company's three most popular products the AI would reply with 94 00:01:50,240 --> 00:01:50,250 popular products the AI would reply with 95 00:01:50,250 --> 00:01:53,060 popular products the AI would reply with the calculated answer not just the data 96 00:01:53,060 --> 00:01:53,070 the calculated answer not just the data 97 00:01:53,070 --> 00:01:56,060 the calculated answer not just the data set Pappas is available now on the 98 00:01:56,060 --> 00:01:56,070 set Pappas is available now on the 99 00:01:56,070 --> 00:02:05,590 set Pappas is available now on the Google research github repository 100 00:02:05,590 --> 00:02:05,600 101 00:02:05,600 --> 00:02:08,640 [Music]