The rise of human-computer cooperation
-
0:01 - 0:03I'd like to tell you about two games of chess.
-
0:03 - 0:07The first happened in 1997, in which Garry Kasparov,
-
0:07 - 0:11a human, lost to Deep Blue, a machine.
-
0:11 - 0:13To many, this was the dawn of a new era,
-
0:13 - 0:16one where man would be dominated by machine.
-
0:16 - 0:19But here we are, 20 years on, and the greatest change
-
0:19 - 0:22in how we relate to computers is the iPad,
-
0:22 - 0:24not HAL.
-
0:24 - 0:26The second game was a freestyle chess tournament
-
0:26 - 0:29in 2005, in which man and machine could enter together
-
0:29 - 0:34as partners, rather than adversaries, if they so chose.
-
0:34 - 0:36At first, the results were predictable.
-
0:36 - 0:38Even a supercomputer was beaten by a grandmaster
-
0:38 - 0:41with a relatively weak laptop.
-
0:41 - 0:44The surprise came at the end. Who won?
-
0:44 - 0:46Not a grandmaster with a supercomputer,
-
0:46 - 0:48but actually two American amateurs
-
0:48 - 0:52using three relatively weak laptops.
-
0:52 - 0:54Their ability to coach and manipulate their computers
-
0:54 - 0:57to deeply explore specific positions
-
0:57 - 0:59effectively counteracted the superior chess knowledge
-
0:59 - 1:02of the grandmasters and the superior computational power
-
1:02 - 1:04of other adversaries.
-
1:04 - 1:07This is an astonishing result: average men,
-
1:07 - 1:11average machines beating the best man, the best machine.
-
1:11 - 1:14And anyways, isn't it supposed to be man versus machine?
-
1:14 - 1:18Instead, it's about cooperation, and the right type of cooperation.
-
1:18 - 1:21We've been paying a lot of attention to Marvin Minsky's
-
1:21 - 1:24vision for artificial intelligence over the last 50 years.
-
1:24 - 1:26It's a sexy vision, for sure. Many have embraced it.
-
1:26 - 1:29It's become the dominant school of thought in computer science.
-
1:29 - 1:32But as we enter the era of big data, of network systems,
-
1:32 - 1:35of open platforms, and embedded technology,
-
1:35 - 1:38I'd like to suggest it's time to reevaluate an alternative vision
-
1:38 - 1:41that was actually developed around the same time.
-
1:41 - 1:45I'm talking about J.C.R. Licklider's human-computer symbiosis,
-
1:45 - 1:49perhaps better termed "intelligence augmentation," I.A.
-
1:49 - 1:51Licklider was a computer science titan who had a profound
-
1:51 - 1:54effect on the development of technology and the Internet.
-
1:54 - 1:57His vision was to enable man and machine to cooperate
-
1:57 - 2:01in making decisions, controlling complex situations
-
2:01 - 2:02without the inflexible dependence
-
2:02 - 2:05on predetermined programs.
-
2:05 - 2:07Note that word "cooperate."
-
2:07 - 2:10Licklider encourages us not to take a toaster
-
2:10 - 2:12and make it Data from "Star Trek,"
-
2:12 - 2:16but to take a human and make her more capable.
-
2:16 - 2:18Humans are so amazing -- how we think,
-
2:18 - 2:21our non-linear approaches, our creativity,
-
2:21 - 2:23iterative hypotheses, all very difficult if possible at all
-
2:23 - 2:24for computers to do.
-
2:24 - 2:26Licklider intuitively realized this, contemplating humans
-
2:26 - 2:29setting the goals, formulating the hypotheses,
-
2:29 - 2:32determining the criteria, and performing the evaluation.
-
2:32 - 2:34Of course, in other ways, humans are so limited.
-
2:34 - 2:37We're terrible at scale, computation and volume.
-
2:37 - 2:39We require high-end talent management
-
2:39 - 2:41to keep the rock band together and playing.
-
2:41 - 2:43Licklider foresaw computers doing all the routinizable work
-
2:43 - 2:47that was required to prepare the way for insights and decision making.
-
2:47 - 2:49Silently, without much fanfare,
-
2:49 - 2:52this approach has been compiling victories beyond chess.
-
2:52 - 2:55Protein folding, a topic that shares the incredible expansiveness of chess —
-
2:55 - 2:59there are more ways of folding a protein than there are atoms in the universe.
-
2:59 - 3:01This is a world-changing problem with huge implications
-
3:01 - 3:03for our ability to understand and treat disease.
-
3:03 - 3:07And for this task, supercomputer field brute force simply isn't enough.
-
3:07 - 3:10Foldit, a game created by computer scientists,
-
3:10 - 3:12illustrates the value of the approach.
-
3:12 - 3:15Non-technical, non-biologist amateurs play a video game
-
3:15 - 3:18in which they visually rearrange the structure of the protein,
-
3:18 - 3:20allowing the computer to manage the atomic forces
-
3:20 - 3:23and interactions and identify structural issues.
-
3:23 - 3:26This approach beat supercomputers 50 percent of the time
-
3:26 - 3:28and tied 30 percent of the time.
-
3:28 - 3:32Foldit recently made a notable and major scientific discovery
-
3:32 - 3:35by deciphering the structure of the Mason-Pfizer monkey virus.
-
3:35 - 3:38A protease that had eluded determination for over 10 years
-
3:38 - 3:40was solved was by three players in a matter of days,
-
3:40 - 3:42perhaps the first major scientific advance
-
3:42 - 3:45to come from playing a video game.
-
3:45 - 3:47Last year, on the site of the Twin Towers,
-
3:47 - 3:48the 9/11 memorial opened.
-
3:48 - 3:51It displays the names of the thousands of victims
-
3:51 - 3:54using a beautiful concept called "meaningful adjacency."
-
3:54 - 3:56It places the names next to each other based on their
-
3:56 - 3:59relationships to one another: friends, families, coworkers.
-
3:59 - 4:02When you put it all together, it's quite a computational
-
4:02 - 4:06challenge: 3,500 victims, 1,800 adjacency requests,
-
4:06 - 4:09the importance of the overall physical specifications
-
4:09 - 4:11and the final aesthetics.
-
4:11 - 4:14When first reported by the media, full credit for such a feat
-
4:14 - 4:16was given to an algorithm from the New York City
-
4:16 - 4:20design firm Local Projects. The truth is a bit more nuanced.
-
4:20 - 4:22While an algorithm was used to develop the underlying framework,
-
4:22 - 4:25humans used that framework to design the final result.
-
4:25 - 4:28So in this case, a computer had evaluated millions
-
4:28 - 4:31of possible layouts, managed a complex relational system,
-
4:31 - 4:33and kept track of a very large set of measurements
-
4:33 - 4:36and variables, allowing the humans to focus
-
4:36 - 4:39on design and compositional choices.
-
4:39 - 4:40So the more you look around you,
-
4:40 - 4:42the more you see Licklider's vision everywhere.
-
4:42 - 4:45Whether it's augmented reality in your iPhone or GPS in your car,
-
4:45 - 4:48human-computer symbiosis is making us more capable.
-
4:48 - 4:50So if you want to improve human-computer symbiosis,
-
4:50 - 4:51what can you do?
-
4:51 - 4:53You can start by designing the human into the process.
-
4:53 - 4:56Instead of thinking about what a computer will do to solve the problem,
-
4:56 - 5:00design the solution around what the human will do as well.
-
5:00 - 5:01When you do this, you'll quickly realize that you spent
-
5:01 - 5:04all of your time on the interface between man and machine,
-
5:04 - 5:07specifically on designing away the friction in the interaction.
-
5:07 - 5:10In fact, this friction is more important than the power
-
5:10 - 5:12of the man or the power of the machine
-
5:12 - 5:14in determining overall capability.
-
5:14 - 5:16That's why two amateurs with a few laptops
-
5:16 - 5:19handily beat a supercomputer and a grandmaster.
-
5:19 - 5:22What Kasparov calls process is a byproduct of friction.
-
5:22 - 5:24The better the process, the less the friction.
-
5:24 - 5:28And minimizing friction turns out to be the decisive variable.
-
5:28 - 5:31Or take another example: big data.
-
5:31 - 5:32Every interaction we have in the world is recorded
-
5:32 - 5:35by an ever growing array of sensors: your phone,
-
5:35 - 5:38your credit card, your computer. The result is big data,
-
5:38 - 5:40and it actually presents us with an opportunity
-
5:40 - 5:42to more deeply understand the human condition.
-
5:42 - 5:45The major emphasis of most approaches to big data
-
5:45 - 5:47focus on, "How do I store this data? How do I search
-
5:47 - 5:49this data? How do I process this data?"
-
5:49 - 5:51These are necessary but insufficient questions.
-
5:51 - 5:54The imperative is not to figure out how to compute,
-
5:54 - 5:56but what to compute. How do you impose human intuition
-
5:56 - 5:58on data at this scale?
-
5:58 - 6:01Again, we start by designing the human into the process.
-
6:01 - 6:04When PayPal was first starting as a business, their biggest
-
6:04 - 6:07challenge was not, "How do I send money back and forth online?"
-
6:07 - 6:11It was, "How do I do that without being defrauded by organized crime?"
-
6:11 - 6:13Why so challenging? Because while computers can learn
-
6:13 - 6:16to detect and identify fraud based on patterns,
-
6:16 - 6:17they can't learn to do that based on patterns
-
6:17 - 6:20they've never seen before, and organized crime
-
6:20 - 6:22has a lot in common with this audience: brilliant people,
-
6:22 - 6:26relentlessly resourceful, entrepreneurial spirit — (Laughter) —
-
6:26 - 6:29and one huge and important difference: purpose.
-
6:29 - 6:31And so while computers alone can catch all but the cleverest
-
6:31 - 6:34fraudsters, catching the cleverest is the difference
-
6:34 - 6:36between success and failure.
-
6:36 - 6:38There's a whole class of problems like this, ones with
-
6:38 - 6:41adaptive adversaries. They rarely if ever present with a
-
6:41 - 6:44repeatable pattern that's discernable to computers.
-
6:44 - 6:48Instead, there's some inherent component of innovation or disruption,
-
6:48 - 6:50and increasingly these problems are buried in big data.
-
6:50 - 6:53For example, terrorism. Terrorists are always adapting
-
6:53 - 6:55in minor and major ways to new circumstances, and despite
-
6:55 - 6:58what you might see on TV, these adaptations,
-
6:58 - 7:00and the detection of them, are fundamentally human.
-
7:00 - 7:04Computers don't detect novel patterns and new behaviors,
-
7:04 - 7:07but humans do. Humans, using technology, testing hypotheses,
-
7:07 - 7:11searching for insight by asking machines to do things for them.
-
7:11 - 7:14Osama bin Laden was not caught by artificial intelligence.
-
7:14 - 7:16He was caught by dedicated, resourceful, brilliant people
-
7:16 - 7:21in partnerships with various technologies.
-
7:21 - 7:23As appealing as it might sound, you cannot algorithmically
-
7:23 - 7:25data mine your way to the answer.
-
7:25 - 7:28There is no "Find Terrorist" button, and the more data
-
7:28 - 7:30we integrate from a vast variety of sources
-
7:30 - 7:32across a wide variety of data formats from very
-
7:32 - 7:36disparate systems, the less effective data mining can be.
-
7:36 - 7:38Instead, people will have to look at data
-
7:38 - 7:41and search for insight, and as Licklider foresaw long ago,
-
7:41 - 7:44the key to great results here is the right type of cooperation,
-
7:44 - 7:45and as Kasparov realized,
-
7:45 - 7:48that means minimizing friction at the interface.
-
7:48 - 7:51Now this approach makes possible things like combing
-
7:51 - 7:54through all available data from very different sources,
-
7:54 - 7:57identifying key relationships and putting them in one place,
-
7:57 - 8:00something that's been nearly impossible to do before.
-
8:00 - 8:02To some, this has terrifying privacy and civil liberties
-
8:02 - 8:05implications. To others it foretells of an era of greater
-
8:05 - 8:07privacy and civil liberties protections,
-
8:07 - 8:10but privacy and civil liberties are of fundamental importance.
-
8:10 - 8:13That must be acknowledged, and they can't be swept aside,
-
8:13 - 8:15even with the best of intents.
-
8:15 - 8:18So let's explore, through a couple of examples, the impact
-
8:18 - 8:20that technologies built to drive human-computer symbiosis
-
8:20 - 8:23have had in recent time.
-
8:23 - 8:26In October, 2007, U.S. and coalition forces raided
-
8:26 - 8:29an al Qaeda safe house in the city of Sinjar
-
8:29 - 8:31on the Syrian border of Iraq.
-
8:31 - 8:33They found a treasure trove of documents:
-
8:33 - 8:35700 biographical sketches of foreign fighters.
-
8:35 - 8:38These foreign fighters had left their families in the Gulf,
-
8:38 - 8:41the Levant and North Africa to join al Qaeda in Iraq.
-
8:41 - 8:43These records were human resource forms.
-
8:43 - 8:46The foreign fighters filled them out as they joined the organization.
-
8:46 - 8:47It turns out that al Qaeda, too,
-
8:47 - 8:49is not without its bureaucracy. (Laughter)
-
8:49 - 8:51They answered questions like, "Who recruited you?"
-
8:51 - 8:54"What's your hometown?" "What occupation do you seek?"
-
8:54 - 8:58In that last question, a surprising insight was revealed.
-
8:58 - 9:00The vast majority of foreign fighters
-
9:00 - 9:02were seeking to become suicide bombers for martyrdom --
-
9:02 - 9:07hugely important, since between 2003 and 2007, Iraq
-
9:07 - 9:11had 1,382 suicide bombings, a major source of instability.
-
9:11 - 9:13Analyzing this data was hard. The originals were sheets
-
9:13 - 9:16of paper in Arabic that had to be scanned and translated.
-
9:16 - 9:18The friction in the process did not allow for meaningful
-
9:18 - 9:21results in an operational time frame using humans, PDFs
-
9:21 - 9:23and tenacity alone.
-
9:23 - 9:25The researchers had to lever up their human minds
-
9:25 - 9:28with technology to dive deeper, to explore non-obvious
-
9:28 - 9:31hypotheses, and in fact, insights emerged.
-
9:31 - 9:34Twenty percent of the foreign fighters were from Libya,
-
9:34 - 9:3750 percent of those from a single town in Libya,
-
9:37 - 9:39hugely important since prior statistics put that figure at
-
9:39 - 9:41three percent. It also helped to hone in on a figure
-
9:41 - 9:44of rising importance in al Qaeda, Abu Yahya al-Libi,
-
9:44 - 9:47a senior cleric in the Libyan Islamic fighting group.
-
9:47 - 9:50In March of 2007, he gave a speech, after which there was
-
9:50 - 9:53a surge in participation amongst Libyan foreign fighters.
-
9:53 - 9:56Perhaps most clever of all, though, and least obvious,
-
9:56 - 9:58by flipping the data on its head, the researchers were
-
9:58 - 10:01able to deeply explore the coordination networks in Syria
-
10:01 - 10:04that were ultimately responsible for receiving and
-
10:04 - 10:06transporting the foreign fighters to the border.
-
10:06 - 10:09These were networks of mercenaries, not ideologues,
-
10:09 - 10:11who were in the coordination business for profit.
-
10:11 - 10:13For example, they charged Saudi foreign fighters
-
10:13 - 10:15substantially more than Libyans, money that would have
-
10:15 - 10:18otherwise gone to al Qaeda.
-
10:18 - 10:20Perhaps the adversary would disrupt their own network
-
10:20 - 10:23if they knew they cheating would-be jihadists.
-
10:23 - 10:26In January, 2010, a devastating 7.0 earthquake struck Haiti,
-
10:26 - 10:29third deadliest earthquake of all time, left one million people,
-
10:29 - 10:3210 percent of the population, homeless.
-
10:32 - 10:35One seemingly small aspect of the overall relief effort
-
10:35 - 10:37became increasingly important as the delivery of food
-
10:37 - 10:39and water started rolling.
-
10:39 - 10:41January and February are the dry months in Haiti,
-
10:41 - 10:44yet many of the camps had developed standing water.
-
10:44 - 10:46The only institution with detailed knowledge of Haiti's
-
10:46 - 10:47floodplains had been leveled
-
10:47 - 10:50in the earthquake, leadership inside.
-
10:50 - 10:53So the question is, which camps are at risk,
-
10:53 - 10:55how many people are in these camps, what's the
-
10:55 - 10:57timeline for flooding, and given very limited resources
-
10:57 - 11:00and infrastructure, how do we prioritize the relocation?
-
11:00 - 11:03The data was incredibly disparate. The U.S. Army had
-
11:03 - 11:06detailed knowledge for only a small section of the country.
-
11:06 - 11:08There was data online from a 2006 environmental risk
-
11:08 - 11:11conference, other geospatial data, none of it integrated.
-
11:11 - 11:14The human goal here was to identify camps for relocation
-
11:14 - 11:16based on priority need.
-
11:16 - 11:19The computer had to integrate a vast amount of geospacial
-
11:19 - 11:21information, social media data and relief organization
-
11:21 - 11:25information to answer this question.
-
11:25 - 11:27By implementing a superior process, what was otherwise
-
11:27 - 11:30a task for 40 people over three months became
-
11:30 - 11:33a simple job for three people in 40 hours,
-
11:33 - 11:36all victories for human-computer symbiosis.
-
11:36 - 11:38We're more than 50 years into Licklider's vision
-
11:38 - 11:40for the future, and the data suggests that we should be
-
11:40 - 11:43quite excited about tackling this century's hardest problems,
-
11:43 - 11:46man and machine in cooperation together.
-
11:46 - 11:48Thank you. (Applause)
-
11:48 - 11:51(Applause)
- Title:
- The rise of human-computer cooperation
- Speaker:
- Shyam Sankar
- Description:
-
Brute computing force alone can’t solve the world’s problems. Data mining innovator Shyam Sankar explains why solving big problems (like catching terrorists or identifying huge hidden trends) is not a question of finding the right algorithm, but rather the right symbiotic relationship between computation and human creativity.
- Video Language:
- English
- Team:
- closed TED
- Project:
- TEDTalks
- Duration:
- 12:12
Morton Bast edited English subtitles for The rise of human-computer cooperation | ||
João Gil edited English subtitles for The rise of human-computer cooperation | ||
Morton Bast edited English subtitles for The rise of human-computer cooperation | ||
Morton Bast edited English subtitles for The rise of human-computer cooperation | ||
Morton Bast edited English subtitles for The rise of human-computer cooperation | ||
Thu-Huong Ha approved English subtitles for The rise of human-computer cooperation | ||
Thu-Huong Ha edited English subtitles for The rise of human-computer cooperation | ||
Morton Bast accepted English subtitles for The rise of human-computer cooperation |