Oregon State University develops talk show robots

Posted May 28, 20206 min read

Scholars at Oregon State University have been studying how to make robots grasp a sense of humor and learn to perform better in talk shows. To this end, the team took the talk show robot Jon for dozens of performances and summarized some success tips.

Most of the robots in our traditional impression are cold metal textures. Even voice assistants such as Siri, although there is no physical image, the cold voice also conveys a robot temperament.

There is a saying, "When human-computer interaction is as natural as human communication, the real era of intelligence comes."

In the direction of human-computer interaction, scientists and engineers have been working hard, and also hope that robots and AI can get a sense of humor to better get along with humans. After all, humor is one of humanity's most precious qualities.

For example, Siri and Microsoft Xiaobing are all trying to learn to tell a joke.

Let robots tell jokes:a good way to improve human-computer interaction

It is not easy to have a robot or voice assistant AI with a sense of humor, and it is even considered to be the ultimate problem in the field.

(We have discussed the issue of AI having a sense of humor before. For details, see ["Let Artificial Intelligence Learn Humor, Human-Machine Dialogue Is Not Awkward"]( http://mp.weixin.qq.com/s?__biz = MzU3NTQ2NDIyOQ == & mid = 2247485023 & idx = 1 & sn = 2186b90720beff6ea2497509b9168bd3 & chksm = fd23f015ca547903b13e4db5226138b5e91194c31f87e3b7359d68d4ec1307b70076fc2627d # 2121 ##

In May 2019, a robot named "Titanium" appeared on the stage of the talk show "Swordsman". It claimed to be funny, and the paragraphs were frequent, making the audience laugh out loud.

However, the over-representation of "Titan" in action and language interaction has caused the audience to question it. Later, it was confirmed that it was really just a robot shell with a real person inside to control its movements and performances.

Since the development of social robots, its interactivity and understanding of human emotions have always been very challenging. Like "Titan" may have certain components that meet the needs of the program. This is a freely and super capable robot. , In reality it is still difficult to do.

Oregon State University has a research team that has been working on human-computer interaction and robotic sense of humor since 2011. In order to solve this problem and let social robots use humor to better attract people and look more like humans, Naomi Fitter, an assistant professor of robotics, plans to start by building talk show robots.

The team first selected the NAO robot developed by Softbank Robot as the carrier and performer to run the talk show performance program they developed.

So far, Naomi Fitter has performed 32 West Coast tours with her talk show robot Jon. Discover how the robot gradually understands and learns human sense of humor.

Naomi Fitter believes that entertainment robot technology can provide a good opportunity for scientific communication

32 performances, constantly improving performance skills

We know that talk shows are generally composed of "opening speech-shaking baggage." In addition to the content of the joke, the speed of the joke and the pause time between the opening speech and the baggage shaking are also crucial to the talk show performance. **

Because the research direction at this stage is to make the humorous interaction between robots and humans more natural, rather than automatically generating jokes. Therefore, the researchers followed the general routine of talk shows and pre-written performance materials suitable for robots for Jon.

These materials are from the perspective of robots, and the topics include interpersonal relationships, daily life, politics, dreams and frustration.

For example:I was always embarrassed when my nephew asked me "Where does the robot come from?" I can only tell it that the robot comes from its mother, the express truck.

Jon judges the audience's response by radio during the performance to decide whether the next stalk needs to be paused or whether he needs to strengthen his tone.

The researchers wrote about 8 minutes of material, including 26 jokes and 22 emotional tags. Then use the NAO robot as a performer to run its comedy performance program.

During the performance process, the robot's microphone is used to collect performance records and evaluate the audience's reaction

Robot Jon performed 32 performances at the two Open Mais, which were performed at two open Mai venues in the Los Angeles area and Oregon.

First study:Pause skills add points to performance

The first study was conducted in the Los Angeles area and included a total of 22 performances with 10-20 adult audiences on the scene. They had not seen a robot talk show before.

The robot performed two modes of performance.

Schematic diagram of the two modes

The first mode is bad-timing mode:that is, after telling each joke, wait for 5 seconds fixedly before starting the next one. In this mode, the robot has completed 12 jokes;

The team uses two methods based on volume and sound count to determine whether the audience is still laughing and applauding.

Then set a reference waiting time of 170ms for the robot. If the audience stays quiet during this period, it will start. If it has not been quiet, it will wait for the 5 second timeout to start the next joke(experts suggest that 5 seconds is usually enough Too.)

The second mode is the appropriate timing mode:**The robot decides to pause or continue speaking according to the reaction of the audience on the spot. If the audience is laughing, pause; if the audience returns to silence, continue.** In this mode, Jon completed 10 jokes.

The research shows that if a robot comedian has a good grasp of the performance timing, that is, giving the audience appropriate reaction time, etc., it will be much more interesting than a robot that does not have a good timing to perform on its own.

Comparison of success rates in the two modes, the success rate is higher in the appropriate timing mode

Because, when a joke succeeds, the audience often laughs or applauds, but because the robot will start the next joke after a fixed delay time, the audience will miss part of the next joke, resulting in failure.

The picture shows that the audience is always laughing when telling joke 2, because they are still immersed in joke 1 **** Second study:"Speak and observe" while speaking **

The second study is based on 10 performances in Oregon. The team takes into account the audience's different responses to the performance and calls it "Adaptive Performance".

The research team added a joke with a label to the joke material. The label is divided into fierce, lukewarm and silent.

** Adaptive performances will choose different jokes based on the audience's different reactions; non-adaptive performances will simply "remember"

Whenever a robot tells a joke, it chooses which "label" joke is based on the level of audience laughter. In the second study, all performances were selected with appropriate timing.

However, the results of this study show that whether or not is labeled has little effect on the success rate of the performance.

Conclusion:Shaking the burden at the right time, the probability of success is high

Based on the above research, Fitter concluded:"In bad-timing mode, the robot always waits a full five seconds after each joke, regardless of the audience's reaction.

In adaptive-timing mode, the timing strategy used by the robot pauses when the laughter comes to mind, and continues after the laughter subsides, just like a good human comedian.

So overall, when the joke is spoken at the right time, the audience s response rate is higher. In this technique, Jon is already familiar with light cars.

Research significance:Improve the robot's understanding of human sense of humor

Fitter said that the 32 performances provided enough data for the research team to determine the significant differences between the different modes of the robot talk show. Moreover, these data can help "autonomous social robots improve their humor". **

At the same time, this research also helps to provide some answers to the key questions of comedy social interaction. It will help researchers in the field of robotics and AI to understand the response of human groups to entertaining social robots in real environments, and will also help comedians evaluate the success of a joke or routine.

Will robots bring change in the comedy world dominated by human actors

For example, before the next performance, Li birthday can let the robot rehearse themselves, if the effect is not good, and then improve in time. In this way, the talk show actors may not need to perform on stage in the future, just focus on writing jokes behind the scenes?

Just like a joke made by Jon:If you guys like me, then book me! I will not only tell jokes, but also replace your work ?

0(13) .gif