专栏首页超级架构师3 Lessons IBM's Watson Can Teach Us About Our Brains' Biases

3 Lessons IBM's Watson Can Teach Us About Our Brains' Biases

COGNITIVE COMPUTING IS TRANSFORMING THE WAY WE WORK. IT ALSO OFFERS A WINDOW TO THE LIMITATIONS OF THE MIND TO HELP US OVERCOME THEM.

BY JANE PORTER

What can artificial intelligence teach us about human intelligence? Quite a bit, it turns out. IBM's Watson had its crowning moment back in 2011 when the cognitive computing system famously destroyed its Jeopardy champion contenders, but lately Watson has been wading into far more murky cognitive territory—creative thinking.

"The crowning achievement of human intelligence is our ability to be creative," saysSteve Abrams, director of IBM Watson. Abrams is in the business of recreating cognitive processes in computers. "I have to know how people work," he says. And one of the biggest pieces of understanding how people work is uncovering what it means to think creatively.

From Chef Watson, which teamed up with the Institute of Culinary Education to develop unexpected food combinations to Watson Tone Analyzer, which performs automatic linguistic analysis to interpret not just the informationm but the tone of people's writing, Watson is reaching into all sorts of creative applications.

And the cognitive computing system, which has been snapped up and applied in various ways by some 5,000 different application developers in the past year, can also teach us quite a bit about the way our minds work. Years of studying cognitive computing and developing cognitive systems for Watson point to a number of ways in our brains function during creative thinking and problem solving.

"What is creativity? It's having a message to convey or a problem to solve in the face of some sort of constraints, and figuring out how to achieve that goal in the face of those constraints," says Abrams.

But what are some of the cognitive roadblocks that might be standing in the way of our ability to think and problem solve most creatively? Fast Company spoke with Abrams about what he's learned about the cognitive biases we bring to the table that a computing system like Watson is designed to overcome.

THE BRAIN'S AMBIGUITY EFFECT

In the absence of information, we tend to avoid choosing options we don't know enough about or don't feel fully confident in. Behavioral scientists call this the "ambiguity effect"—a cognitive bias in which your decision-making is impaired by a lack of information, says Abrams.

Take, for example, the choice by some investors to put their money in a less risky fixed yield instrument rather than opting for a more volatile investment like stocks and funds that have less predictable results, but offer higher returns over time. "There are people for whom that unknown factors into the equation more than it should," says Abrams. Recognizing those biases and seeking out information in the face of doubt can help counter that ambiguity effect from preventing you to make the best and most informed decision.

OUR TENDENCY TOWARD CONFIRMATION BIAS

Another bias we're often susceptible to is confirmation bias, or the tendency to look for and favor evidence that confirms what you already believe to be true. This affects the way we approach problem solving—from the key words we choose to look up on a search engine to the details we selectively notice in favor of our own opinions.

ON WATSON'S RECOMMENDATION, THE CHEF WENT AHEAD, RESISTING HIS OWN CONFIRMATION BIAS AND FOUND THE PAIRING WAS SURPRISINGLY COMPLEMENTARY.

Take for example, our preferences in food taste. When Chef Watson called up a flavor pairing between apples and olive oil, the immediate reaction from the human chef involved was skepticism. Apples and olive oil are not a typical combination. But going on Watson's recommendation, the chef went ahead, resisting his own confirmation bias and found the pairing was surprisingly complementary.

ARE YOU SUFFERING FROM NOT INVENTED HERE SYNDROME?

To be clear, this isn't an actual syndrome, but the phenomenon (sometimes referred to as NIH) is very much out there. It's the tendency to be more critical of someone else's ideas rather than your own. We tend to apply higher standards to other people's solutions rather than our own, rejecting external solutions as less superior.

In the tech world, this kind of NIH is often what leads people to reject software or developments made outside their company in favor of their own in-house version, however inferior it may be. But this happens across industries and organizations.

"Those biases that we aren't even conscious of having impact the way we interpret results," says Abrams. "We are trying to build a system that will help us come to a more reasonable set of interesting options, despite the fact that we as human beings have these biases."

本文分享自微信公众号 - 首席架构师智库(jiagoushipro)

原文出处及转载信息见文内详细说明,如有侵权,请联系 yunjia_community@tencent.com 删除。

原始发表时间:2015-10-04

本文参与腾讯云自媒体分享计划,欢迎正在阅读的你也加入,一起分享。

我来说两句

0 条评论
登录 后参与评论

相关文章

  • 呃哦:区块链可能没有我们想象的那么安全

    在对一百万份智能合约的分析中,一项新的分析工具发现34,200个安全漏洞。 在我们转向基于区块链的数字经济之前,我们需要解决这个系统中的缺陷。 ? 区块链有可能...

    首席架构师智库
  • Instant Messaging at LinkedIn: Scaling to 10000 of Connections

    We recently introduced Instant Messaging on LinkedIn, complete with typing indic...

    首席架构师智库
  • 区块链是未来吗? 或者它只是一种时尚?

    区块链行业和美国大淘金者的共同点 今天用加密货币发生的事情让人联想到90年代的网络泡沫时代。投资科技公司的人们以及更加连接未来的雄心勃勃的想法几乎在一夜之间赚取...

    首席架构师智库
  • 【论文推荐】最新6篇生成式对抗网络(GAN)相关论文—半监督对抗学习、行人再识别、代表性特征、高分辨率深度卷积、自监督、超分辨

    【导读】专知内容组整理了最近六篇生成式对抗网络(GAN)相关文章,为大家进行介绍,欢迎查看! 1. Classification of sparsely lab...

    WZEARW
  • 用信息瓶颈的迁移学习和探索

    Transfer and Exploration via the Information Bottleneck

    用户1908973
  • Graph application with Python, Neo4j, Gephi & Linkurious.js

    I love Python, and to celebrate Packt Python week, I’ve spent some time developi...

    fishexpert
  • 面向边缘计算的深度学习:直接从压缩数据中提取神经网络 (CS)

    由于手机的普及和计算能力的增长,以及人工智能的进步,许多智能应用被开发出来,有意义地丰富了人们的生活。为此,边缘智能领域越来越受到人们的关注,它的目的是将数据的...

    管欣8078776
  • GCN 论文英语表达总结

    -------------------------------------------------------一条开始认真脸的分界线--------------...

    张凝可
  • 使用infogan学习可解释的隐变量特征学习-及代码示例(代码和官方有差异)

    In this week’s post I want to explore a simple addition to Generative Adversaria...

    用户1908973
  • 【论文推荐】最新八篇生成对抗网络相关论文—BRE、图像合成、多模态图像生成、非配对多域图、注意力、对抗特征增强、深度对抗性训练

    WZEARW

扫码关注云+社区

领取腾讯云代金券