homework doing robot

• Skip to main content
• Keyboard shortcuts for audio player

A new AI chatbot might do your homework for you. But it's still not an A+ student

Emma Bowman

Enter a prompt into ChatGPT, and it becomes your very own virtual assistant. OpenAI/Screenshot by NPR hide caption

Enter a prompt into ChatGPT, and it becomes your very own virtual assistant.

Why do your homework when a chatbot can do it for you? A new artificial intelligence tool called ChatGPT has thrilled the Internet with its superhuman abilities to solve math problems, churn out college essays and write research papers.

After the developer OpenAI released the text-based system to the public last month, some educators have been sounding the alarm about the potential that such AI systems have to transform academia, for better and worse.

"AI has basically ruined homework," said Ethan Mollick, a professor at the University of Pennsylvania's Wharton School of Business, on Twitter.

The tool has been an instant hit among many of his students, he told NPR in an interview on Morning Edition , with its most immediately obvious use being a way to cheat by plagiarizing the AI-written work, he said.

Academic fraud aside, Mollick also sees its benefits as a learning companion.

Opinion: Machine-made poetry is here

He's used it as his own teacher's assistant, for help with crafting a syllabus, lecture, an assignment and a grading rubric for MBA students.

"You can paste in entire academic papers and ask it to summarize it. You can ask it to find an error in your code and correct it and tell you why you got it wrong," he said. "It's this multiplier of ability, that I think we are not quite getting our heads around, that is absolutely stunning," he said.

A convincing — yet untrustworthy — bot

But the superhuman virtual assistant — like any emerging AI tech — has its limitations. ChatGPT was created by humans, after all. OpenAI has trained the tool using a large dataset of real human conversations.

"The best way to think about this is you are chatting with an omniscient, eager-to-please intern who sometimes lies to you," Mollick said.

It lies with confidence, too. Despite its authoritative tone, there have been instances in which ChatGPT won't tell you when it doesn't have the answer.

That's what Teresa Kubacka, a data scientist based in Zurich, Switzerland, found when she experimented with the language model. Kubacka, who studied physics for her Ph.D., tested the tool by asking it about a made-up physical phenomenon.

"I deliberately asked it about something that I thought that I know doesn't exist so that they can judge whether it actually also has the notion of what exists and what doesn't exist," she said.

ChatGPT produced an answer so specific and plausible sounding, backed with citations, she said, that she had to investigate whether the fake phenomenon, "a cycloidal inverted electromagnon," was actually real.

When she looked closer, the alleged source material was also bogus, she said. There were names of well-known physics experts listed – the titles of the publications they supposedly authored, however, were non-existent, she said.

"This is where it becomes kind of dangerous," Kubacka said. "The moment that you cannot trust the references, it also kind of erodes the trust in citing science whatsoever," she said.

Scientists call these fake generations "hallucinations."

"There are still many cases where you ask it a question and it'll give you a very impressive-sounding answer that's just dead wrong," said Oren Etzioni, the founding CEO of the Allen Institute for AI , who ran the research nonprofit until recently. "And, of course, that's a problem if you don't carefully verify or corroborate its facts."

Users experimenting with the chatbot are warned before testing the tool that ChatGPT "may occasionally generate incorrect or misleading information." OpenAI/Screenshot by NPR hide caption

An opportunity to scrutinize AI language tools

Users experimenting with the free preview of the chatbot are warned before testing the tool that ChatGPT "may occasionally generate incorrect or misleading information," harmful instructions or biased content.

Sam Altman, OpenAI's CEO, said earlier this month it would be a mistake to rely on the tool for anything "important" in its current iteration. "It's a preview of progress," he tweeted .

The failings of another AI language model unveiled by Meta last month led to its shutdown. The company withdrew its demo for Galactica, a tool designed to help scientists, just three days after it encouraged the public to test it out, following criticism that it spewed biased and nonsensical text.

Untangling Disinformation

Ai-generated fake faces have become a hallmark of online influence operations.

Similarly, Etzioni says ChatGPT doesn't produce good science. For all its flaws, though, he sees ChatGPT's public debut as a positive. He sees this as a moment for peer review.

"ChatGPT is just a few days old, I like to say," said Etzioni, who remains at the AI institute as a board member and advisor. It's "giving us a chance to understand what he can and cannot do and to begin in earnest the conversation of 'What are we going to do about it?' "

The alternative, which he describes as "security by obscurity," won't help improve fallible AI, he said. "What if we hide the problems? Will that be a recipe for solving them? Typically — not in the world of software — that has not worked out."

Your browser is not supported

Sorry but it looks as if your browser is out of date. To get the best experience using our site we recommend that you upgrade or switch browsers.

Find a solution

• Skip to main content
• Skip to navigation

• Back to parent navigation item
• Collections
• Sustainability in chemistry
• Simple rules
• Teacher well-being hub
• Women in chemistry
• Global science
• Escape room activities
• Decolonising chemistry teaching
• Teaching science skills
• Post-lockdown teaching support
• Get the print issue
• RSC Education

• More from navigation items

Do your homework

Are your students using AI bots to do theirs? After all, AI is readily available and capable as we found out

We gave ChatGPT a prompt to ‘write an editorial (300 words) for a readership of teachers on the danger of students using AI bots to do their homework for them’. This what it produced:

Source: © Shutterstock

Could you spot a chemistry homework written by an AI bot?

As teachers, we have all seen the impact that technology can have on the classroom. From smartphones and tablets to educational software and online resources, there are countless tools that can help our students learn and grow. However, there is also a dark side to technology in education, and it is something that we need to be aware of and address.

One of the biggest concerns is the use of AI bots to do students’ homework for them. With the rise of artificial intelligence and the increasing availability of AI-powered tools, it is becoming easier and easier for students to use these bots to complete their assignments without actually learning the material.

We need to make it clear that cheating is unacceptable

On the surface, this may seem like a harmless way for students to save time and effort, but it is actually a serious problem for several reasons. First and foremost, using AI bots to do homework undermines the entire purpose of education. The goal of homework is to reinforce the concepts that students are learning in class, not to simply churn out answers without understanding them. By using AI bots to do their homework, students are cheating themselves out of a valuable learning opportunity.

Furthermore, using AI bots to do homework sets a dangerous precedent. It teaches students that they can take shortcuts and cheat their way through school, rather than putting in the hard work and effort that are necessary for success. This attitude can carry over into other areas of their lives, leading to a lack of motivation and a lack of real-world skills.

As teachers, it is our responsibility to educate our students about the dangers of using AI bots to do their homework. We need to make it clear that cheating is unacceptable and that the only way to truly learn and grow is through hard work and dedication. By taking a strong stance on this issue, we can help our students develop the skills and values that will serve them well throughout their lives. 

1 Reader's comment

Only registered users can comment on this article., more from opinion.

Why we should ditch working scientifically

2024-04-12T06:00:00Z By Ben Rogers

Explore a new approach to this this national curriculum strand, grounded in contemporary cognitive science

How do you feel about working in science education?

2024-02-29T05:00:00Z By Lisa Clatworthy

Tell us what it’s like by taking part in the Science teaching survey, and we can improve our support for you

Making science teaching more inclusive

2024-01-04T06:52:00Z By Nasima Hassan

Let’s introduce more representative examples in our chemistry classrooms

• Contributors
• Print issue
• Email alerts

Site powered by Webvision Cloud

The Tech Edvocate

• Advertisement
• Home Page Five (No Sidebar)
• Home Page Four
• Home Page Three
• Home Page Two
• Icons [No Sidebar]
• Left Sidbear Page
• Lynch Educational Consulting
• My Speaking Page
• Newsletter Sign Up Confirmation
• Newsletter Unsubscription
• Page Example
• Privacy Policy
• Protected Content
• Request a Product Review
• Shortcodes Examples
• Terms and Conditions
• The Edvocate
• The Tech Edvocate Product Guide
• Write For Us
• Dr. Lynch’s Personal Website
• The Edvocate Podcast
• Assistive Technology
• Child Development Tech
• Early Childhood & K-12 EdTech
• EdTech Futures
• EdTech News
• EdTech Policy & Reform
• EdTech Startups & Businesses
• Higher Education EdTech
• Online Learning & eLearning
• Parent & Family Tech
• Personalized Learning
• Product Reviews
• Tech Edvocate Awards
• School Ratings

8 Ways to Service an Air Conditioner

3 ways to stop a baby from vomiting, 3 ways to save instagram highlights, skills needed for reading comprehension: everything you need to know, how to change the language in android, 3 ways to permanently delete facebook messages, how to tie the sides of a shirt: 8 simple steps, phonics rules: everything you need to know, how to fill a flask: 8 steps, 3 ways to treat skin disease in hamsters, why do homework, when you can make a robot do it.

Robots are infiltrating our daily lives. They are cleaning our floors, driving us around on campus or the golf course, operating on people, flipping hamburgers, and they’re teaching children in Asia. There are even robots like mBot that teaches kids about robotics.

Robots are meant to automate repetitive tasks, leaving humans free for more creative and rewarding endeavors. And what is more repetitive than having to copy text from a book several times?

The writing robot that can copy texts for students

Enter a very enterprising young Chinese teen. She used the money she received over Chinese New Year to buy a so-called copying robot to do her homework for her. In all fairness, the homework, in this case, was to copy passages from a textbook. Chinese teachers give this kind of homework, even during the Lunar Year holidays, so students can practice the thousands of Chinese characters that make up Mandarin Chinese.

The device, a metal frame, and pen comes with the marketing message that it could “imitate all sorts of handwriting”. In her case, the device performed perfectly, finishing all her homework in record time and in her handwriting. The incident was reported , including the destruction of the device by the girl’s mother. Far from being criticized for her dishonesty, the girl was praised by netizens for her initiative and the company that manufactures the writing robot has been inundated with inquiries ever since.

In an investigation, the South China Morning Post found several entries for a so-called “copying robot” on the e-commerce platform Taobao. It seems like a good idea to let a robot do something for you that is boring and repetitive, but you have to ask yourself if that’s a sensible thing to do. Surely you miss out on a learning opportunity if you let a robot do your work for you.

That doesn’t mean that robots cannot be useful when it comes to school homework. Rather than doing the work for you, your robot can help you to do the work and master the concepts in the curriculum. 

The robot that can do math homework

For kids who might feel burdened by too much homework, the future holds much promise. Christopher Yan and Derek He have come up with a concept of a homework-solving robot. Their prototype can solve math homework. What’s more, when it’s completed, the robot will be able to write the answers according to the user’s handwriting, so a teacher won’t suspect that the work was not done by the student. Their project known as Homework Help , picked up the top honors at this year’s PennApps, one of the largest student hackathons in the world, held at University of Pennsylvania.

It seems there is a need for robots that can help students with their homework. Scientists have noticed that children form close emotional bonds with social robots. It is not yet clear why this is so, but educators can exploit this to help students with their studies. Robots can, and do, add a fun aspect to learning. They can also add a fun aspect to homework.

Harnassing Edtech’s Disruptive Power

Technology collaboration results in improved success.

Matthew Lynch

Related articles more from author.

What is SQL Server Business Intelligence (SQL Server BI)?

AI Chatbots Are Here to Stay: Learn How They Can Work for You

AI Is Hitting Your Computer Hardware: AMD Ryzen AI

What is autocad.

Get a GEEKOM Mini IT11 11th Gen Intel Core i7 for Under $450

Here Come The AI Teaching Bots

New AI chatbot can do students' homework for them

Quebec teachers brace for chatgpt helping students cheat.

Social Sharing

On the first day of class this semester, Laurent Nicolas didn't just go over his course outline.

The English teacher showed his students how ChatGPT — artificial intelligence chat software that has consumed huge quantities of data to generate original, human-like responses — could complete a written assignment in a matter of seconds.

Nicolas, who teaches at Collège André-Grasset, a private college in Montreal, asked ChatGPT to write a gothic poem about rock climbing. The response came moments later:

With every step, my heart beats fast, As I reach for the next hold tight, The echoes of the past, Whisper secrets in the night.

But one of Nicolas's students was already familiar with the bot. He raised his hand and shared that he'd used it to write an essay the previous semester.

Faced with the explosive attention ChatGPT has garnered since it was made publicly available last fall, Quebec teachers are grappling with how to manage — or make the most of — the AI program in their classrooms.

A new frontier for plagiarism

Jeremy Klughaupt, an English teacher at Collège de Maisonneuve, typed some of his essay questions into the program, and was horrified by the result.

In less than a minute, the AI spat out responses comparable with the work some of his students might produce.

Klughaupt said many of his colleagues are "pulling their hair out," unsure how to proceed with their course plans and evaluations. Spotting and dealing with assignments completed by Google Translate was already a challenge. Now, with ChatGPT at his students' fingertips, he worries that the potential for fraud and plagiarism grows as this technology continues to evolve.

"As teachers, a lot of us feel torn between these two roles that we have: to educate and, increasingly, to police students," Klughaupt said.

Teachers in his department are considering abandoning take-home writing assignments. But that is a decision he is not yet prepared to make. Students hone their writing skills when engaged in long-term, immersive projects, which can't all be done in the classroom, he said.

Andrew Piper, professor in the department of languages, literatures and cultures at McGill University, agrees.

"We don't want to just get rid of essay writing," he said on CBC Montreal's Daybreak . "Writing is a fabulous way to figure out what you're trying to say and develop your thoughts and has been for a long stretch of human history."

However, Piper believes the risks go beyond academic integrity. The program's ability to easily generate massive amounts of text, which can be personalized, targeted and convincing, is a cause for concern.

"The problem is, [ChatGPT] doesn't really understand how the world works," said Piper. "It doesn't actually know if something's true. And so in certain cases, that can lead to misinformation…. There's nothing in the system that sort of flags itself."

The Quebec Ministry of Education is assessing the impacts of the technology on student learning and teacher work, especially with regard to ministerial exams. Meanwhile, in New York, ChatGPT is banned in all public schools due to concerns about its impact on learning and with its accuracy.

Harnessing the power of AI

But not all Quebec educators are worried. In fact, some are optimistic about ChatGPT and the possible emergence of AI in the province's schools.

Educators like Walcir Cardoso, professor of linguistics at Concordia University, are hoping teachers harness its power to enhance the quality of education.

Cardoso thinks ChatGPT can be effective in improving students' vocabulary, grammar and pronunciation. Teachers should also consider alternatives to essays like debates and pre-recorded presentations, he said.

As for precautions, Cardoso proposes that teachers test the AI for responses to essay questions. And he suggests students should be free to cite the chatbot for certain assignments, as long as they are transparent.

"If students want to plagiarize, they will, regardless of what is available," said Cordoso.

• Experts and educators say viral AI tool ChatGPT brings challenges
• Curveball or game changer? ChatGPT, AI tools under watch on Canadian campuses

For its part, the Lester B. Pearson School Board is still assessing the opportunities and pitfalls that come with ChatGPT.

"It's very early on in the arrival of this technology to really be sure of what the power of it is and how it can be used effectively in the class," said Mat Canavan, director of education services at Lester B. Pearson School Board.

But he knows that if ignored, "it's going to come back and bite us" due to how quickly ChatGPT can complete an assignment.

Canavan says educators need to adapt and make the most of the tools available to them and their students. This means shifting learning to focus on creativity, voice and process — things the bot can't imitate.

"How do we incorporate artificial intelligence into the work that we're doing to make it so a student can learn deeper?" he said.

"In every instance where there's new technology that comes along, it causes a knee-jerk fear reaction. But it also creates the potential for more evolved and or in depth learning."

ABOUT THE AUTHOR

Joe Bongiorno is an author, former high school teacher and a journalist at the CBC. He has also reported for Canadian Geographic, Maisonneuve, Canada’s National Observer and others. You can reach him at [email protected].

With files from CBC Montreal's Daybreak

Related Stories

• From AI dating to flirt coaches: How AI is changing dating, for better or worse

StudyMonkey

Your personal ai tutor.

Learn Smarter, Not Harder with AI

Introducing StudyMonkey, your AI-powered tutor .

StudyMonkey AI can tutor complex homework questions, enhance your essay writing and assess your work—all in seconds.

No more long all-nighters

24/7 solutions to questions you're stumped on and essays you procrastinated on.

No more stress and anxiety

Get all your assignments done with helpful answers in 10 seconds or less.

No more asking friends for help

StudyMonkey is your new smart bestie that will never ghost you.

No more staying after school

AI tutoring is available 24/7, on-demand when you need it most.

AI Tutor for any subject

American college testing (act), anthropology, advanced placement exams (ap exams), arabic language, archaeology, biochemistry, chartered financial analyst (cfa) exam, communications, computer science, certified public accountant (cpa) exam, cultural studies, cyber security, dental admission test (dat), discrete mathematics, earth science, elementary school, entrepreneurship, environmental science, farsi (persian) language, fundamentals of engineering (fe) exam, gender studies, graduate management admission test (gmat), graduate record examination (gre), greek language, hebrew language, high school entrance exam, high school, human geography, human resources, international english language testing system (ielts), information technology, international relations, independent school entrance exam (isee), linear algebra, linguistics, law school admission test (lsat), machine learning, master's degree, medical college admission test (mcat), meteorology, microbiology, middle school, national council licensure examination (nclex), national merit scholarship qualifying test (nmsqt), number theory, organic chemistry, project management professional (pmp), political science, portuguese language, probability, project management, preliminary sat (psat), public policy, public relations, russian language, scholastic assessment test (sat), social sciences, secondary school admission test (ssat), sustainability, swahili language, test of english as a foreign language (toefl), trigonometry, turkish language, united states medical licensing examination (usmle), web development, step-by-step guidance 24/7.

Receive step-by-step guidance & homework help for any homework problem & any subject 24/7

Ask any question

StudyMonkey supports every subject and every level of education from 1st grade to masters level.

Get an answer

StudyMonkey will give you an answer in seconds—multiple choice questions, short answers, and even an essays are supported!

Review your history

See your past questions and answers so you can review for tests and improve your grades.

It's not cheating...

You're just learning smarter than everyone else

How Can StudyMonkey Help You?

Hear from our happy students.

"The AI tutor is available 24/7, making it a convenient and accessible resource for students who need help with their homework at any time."

"Overall, StudyMonkey is an excellent tool for students looking to improve their understanding of homework topics and boost their academic success."

Upgrade to StudyMonkey Premium!

Why not upgrade to StudyMonkey Premium and get access to all features?

GPB Originals

Browse by genre, featured programs, featured programs & series, more gpb news, for kids & teachers, ghsa sports, high school football, browse by type, browse by category, for parents & caregivers, support gpb, gpb newsletter cta, banner image, section branding, header content, tagged as: , a new ai chatbot might do your homework for you. but it's still not an a+ student.

December 19, 2022 5:00 AM

• Emma Bowman

Share this page

Primary content.

Why do your homework when a chatbot can do it for you? A new artificial intelligence tool called ChatGPT has thrilled the Internet with its superhuman abilities to solve math problems, churn out college essays and write research papers.

After the developer OpenAI released the text-based system to the public last month, some educators have been sounding the alarm about the potential that such AI systems have to transform academia, for better and worse.

"AI has basically ruined homework," said Ethan Mollick, a professor at the University of Pennsylvania's Wharton School of Business, on Twitter.

The tool has been an instant hit among many of his students, he told NPR in an interview on Morning Edition , with its most immediately obvious use being a way to cheat by plagiarizing the AI-written work, he said.

Academic fraud aside, Mollick also sees its benefits as a learning companion.

He's used it as his own teacher's assistant, for help with crafting a syllabus, lecture, an assignment and a grading rubric for MBA students.

"You can paste in entire academic papers and ask it to summarize it. You can ask it to find an error in your code and correct it and tell you why you got it wrong," he said. "It's this multiplier of ability, that I think we are not quite getting our heads around, that is absolutely stunning," he said.

A convincing — yet untrustworthy — bot

But the superhuman virtual assistant — like any emerging AI tech — has its limitations. ChatGPT was created by humans, after all. OpenAI has trained the tool using a large dataset of real human conversations.

"The best way to think about this is you are chatting with an omniscient, eager-to-please intern who sometimes lies to you," Mollick said.

It lies with confidence, too. Despite its authoritative tone, there have been instances in which ChatGPT won't tell you when it doesn't have the answer.

That's what Teresa Kubacka, a data scientist based in Zurich, Switzerland, found when she experimented with the language model. Kubacka, who studied physics for her Ph.D., tested the tool by asking it about a made-up physical phenomenon.

"I deliberately asked it about something that I thought that I know doesn't exist so that they can judge whether it actually also has the notion of what exists and what doesn't exist," she said.

ChatGPT produced an answer so specific and plausible sounding, backed with citations, she said, that she had to investigate whether the fake phenomenon, "a cycloidal inverted electromagnon," was actually real.

When she looked closer, the alleged source material was also bogus, she said. There were names of well-known physics experts listed – the titles of the publications they supposedly authored, however, were non-existent, she said.

"This is where it becomes kind of dangerous," Kubacka said. "The moment that you cannot trust the references, it also kind of erodes the trust in citing science whatsoever," she said.

Scientists call these fake generations "hallucinations."

"There are still many cases where you ask it a question and it'll give you a very impressive-sounding answer that's just dead wrong," said Oren Etzioni, the founding CEO of the Allen Institute for AI , who ran the research nonprofit until recently. "And, of course, that's a problem if you don't carefully verify or corroborate its facts."

An opportunity to scrutinize AI language tools

Users experimenting with the free preview of the chatbot are warned before testing the tool that ChatGPT "may occasionally generate incorrect or misleading information," harmful instructions or biased content.

Sam Altman, OpenAI's CEO, said earlier this month it would be a mistake to rely on the tool for anything "important" in its current iteration. "It's a preview of progress," he tweeted .

The failings of another AI language model unveiled by Meta last month led to its shutdown. The company withdrew its demo for Galactica, a tool designed to help scientists, just three days after it encouraged the public to test it out, following criticism that it spewed biased and nonsensical text.

Similarly, Etzioni says ChatGPT doesn't produce good science. For all its flaws, though, he sees ChatGPT's public debut as a positive. He sees this as a moment for peer review.

"ChatGPT is just a few days old, I like to say," said Etzioni, who remains at the AI institute as a board member and advisor. It's "giving us a chance to understand what he can and cannot do and to begin in earnest the conversation of 'What are we going to do about it?' "

The alternative, which he describes as "security by obscurity," won't help improve fallible AI, he said. "What if we hide the problems? Will that be a recipe for solving them? Typically — not in the world of software — that has not worked out."

Copyright 2022 NPR. To see more, visit https://www.npr.org .

Bottom Content

Related news.

Indiana lawmakers ban cellphones in class. Now it's up to schools to figure out how

Many schools — but not all — in the state and around the U.S. already ban phones in class. This requires it now in Indiana.

• Kirsten Adair

AI images and conspiracy theories are driving a push for media literacy education

One of the nation's best-known media literacy events for high school students is expanding as demand grows for skills to identify deepfake images and online conspiracy theories.

• Kim Malcolm

Hackers are targeting a surprising group of people: young public school students

Districts store all kinds of sensitive student data, which means the consequences of a school cyberattack can follow pupils well into adulthood. And it's not just their credit that's at risk.

• Kavitha Cardoza

Support Quality Journalism

GPB is committed to bringing you comprehensive news coverage from Georgia, across the country and around the world.  Your support makes this possible.  Please consider making a gift today to support this vital public service.

Suggestions or feedback?

MIT News | Massachusetts Institute of Technology

• Machine learning
• Social justice
• Black holes
• Classes and programs

Departments

• Aeronautics and Astronautics
• Brain and Cognitive Sciences
• Architecture
• Political Science
• Mechanical Engineering

Centers, Labs, & Programs

• Abdul Latif Jameel Poverty Action Lab (J-PAL)
• Picower Institute for Learning and Memory
• Lincoln Laboratory
• School of Architecture + Planning
• School of Engineering
• School of Humanities, Arts, and Social Sciences
• Sloan School of Management
• School of Science
• MIT Schwarzman College of Computing

Showing robots how to do your chores

Press contact :, media download.

*Terms of Use:

Images for download on the MIT News office website are made available to non-commercial entities, press and the general public under a Creative Commons Attribution Non-Commercial No Derivatives license . You may not alter the images provided, other than to crop them to size. A credit line must be used when reproducing images; if one is not provided below, credit the images to "MIT."

Previous image Next image

Training interactive robots may one day be an easy job for everyone, even those without programming expertise. Roboticists are developing automated robots that can learn new tasks solely by observing humans. At home, you might someday show a domestic robot how to do routine chores. In the workplace, you could train robots like new employees, showing them how to perform many duties.

Making progress on that vision, MIT researchers have designed a system that lets these types of robots learn complicated tasks that would otherwise stymie them with too many confusing rules. One such task is setting a dinner table under certain conditions.  

At its core, the researchers’ “Planning with Uncertain Specifications” (PUnS) system gives robots the humanlike planning ability to simultaneously weigh many ambiguous — and potentially contradictory — requirements to reach an end goal. In doing so, the system always chooses the most likely action to take, based on a “belief” about some probable specifications for the task it is supposed to perform.

In their work, the researchers compiled a dataset with information about how eight objects — a mug, glass, spoon, fork, knife, dinner plate, small plate, and bowl — could be placed on a table in various configurations. A robotic arm first observed randomly selected human demonstrations of setting the table with the objects. Then, the researchers tasked the arm with automatically setting a table in a specific configuration, in real-world experiments and in simulation, based on what it had seen.

To succeed, the robot had to weigh many possible placement orderings, even when items were purposely removed, stacked, or hidden. Normally, all of that would confuse robots too much. But the researchers’ robot made no mistakes over several real-world experiments, and only a handful of mistakes over tens of thousands of simulated test runs.  

“The vision is to put programming in the hands of domain experts, who can program robots through intuitive ways, rather than describing orders to an engineer to add to their code,” says first author Ankit Shah, a graduate student in the Department of Aeronautics and Astronautics (AeroAstro) and the Interactive Robotics Group, who emphasizes that their work is just one step in fulfilling that vision. “That way, robots won’t have to perform preprogrammed tasks anymore. Factory workers can teach a robot to do multiple complex assembly tasks. Domestic robots can learn how to stack cabinets, load the dishwasher, or set the table from people at home.”

Joining Shah on the paper are AeroAstro and Interactive Robotics Group graduate student Shen Li and Interactive Robotics Group leader Julie Shah, an associate professor in AeroAstro and the Computer Science and Artificial Intelligence Laboratory.

Bots hedging bets

Robots are fine planners in tasks with clear “specifications,” which help describe the task the robot needs to fulfill, considering its actions, environment, and end goal. Learning to set a table by observing demonstrations, is full of uncertain specifications. Items must be placed in certain spots, depending on the menu and where guests are seated, and in certain orders, depending on an item’s immediate availability or social conventions. Present approaches to planning are not capable of dealing with such uncertain specifications.

A popular approach to planning is “reinforcement learning,” a trial-and-error machine-learning technique that rewards and penalizes them for actions as they work to complete a task. But for tasks with uncertain specifications, it’s difficult to define clear rewards and penalties. In short, robots never fully learn right from wrong.

The researchers’ system, called PUnS (for Planning with Uncertain Specifications), enables a robot to hold a “belief” over a range of possible specifications. The belief itself can then be used to dish out rewards and penalties. “The robot is essentially hedging its bets in terms of what’s intended in a task, and takes actions that satisfy its belief, instead of us giving it a clear specification,” Ankit Shah says.

The system is built on “linear temporal logic” (LTL), an expressive language that enables robotic reasoning about current and future outcomes. The researchers defined templates in LTL that model various time-based conditions, such as what must happen now, must eventually happen, and must happen until something else occurs. The robot’s observations of 30 human demonstrations for setting the table yielded a probability distribution over 25 different LTL formulas. Each formula encoded a slightly different preference — or specification — for setting the table. That probability distribution becomes its belief.

“Each formula encodes something different, but when the robot considers various combinations of all the templates, and tries to satisfy everything together, it ends up doing the right thing eventually,” Ankit Shah says.

Following criteria

The researchers also developed several criteria that guide the robot toward satisfying the entire belief over those candidate formulas. One, for instance, satisfies the most likely formula, which discards everything else apart from the template with the highest probability. Others satisfy the largest number of unique formulas, without considering their overall probability, or they satisfy several formulas that represent highest total probability. Another simply minimizes error, so the system ignores formulas with high probability of failure.

Designers can choose any one of the four criteria to preset before training and testing. Each has its own tradeoff between flexibility and risk aversion. The choice of criteria depends entirely on the task. In safety critical situations, for instance, a designer may choose to limit possibility of failure. But where consequences of failure are not as severe, designers can choose to give robots greater flexibility to try different approaches.

With the criteria in place, the researchers developed an algorithm to convert the robot’s belief — the probability distribution pointing to the desired formula — into an equivalent reinforcement learning problem. This model will ping the robot with a reward or penalty for an action it takes, based on the specification it’s decided to follow.

In simulations asking the robot to set the table in different configurations, it only made six mistakes out of 20,000 tries. In real-world demonstrations, it showed behavior similar to how a human would perform the task. If an item wasn’t initially visible, for instance, the robot would finish setting the rest of the table without the item. Then, when the fork was revealed, it would set the fork in the proper place. “That’s where flexibility is very important,” Ankit Shah says. “Otherwise it would get stuck when it expects to place a fork and not finish the rest of table setup.”

Next, the researchers hope to modify the system to help robots change their behavior based on verbal instructions, corrections, or a user’s assessment of the robot’s performance. “Say a person demonstrates to a robot how to set a table at only one spot. The person may say, ‘do the same thing for all other spots,’ or, ‘place the knife before the fork here instead,’” Ankit Shah says. “We want to develop methods for the system to naturally adapt to handle those verbal commands, without needing additional demonstrations.”  

Share this news article on:

Related links.

• Paper: “Planning With Uncertain Specifications (PUnS)”
• Paper: “Bayesian Interference of Temporal Task Specifications from Demonstrations”
• Ankit J. Shah
• Interactive Robots Group
• Computer Science and Artificial Intelligence Laboratory
• Department of Aeronautics and Astronautics

Related Topics

• Computer science and technology
• Artificial intelligence
• Assistive technology
• Aeronautical and astronautical engineering
• Computer Science and Artificial Intelligence Laboratory (CSAIL)

Related Articles

Algorithm tells robots where nearby humans are headed

Teaching robots to teach other robots

Robot helps nurses schedule tasks on labor floor

Previous item Next item

More MIT News

A biomedical engineer pivots from human movement to women’s health

Read full story →

MIT tops among single-campus universities in US patents granted

A new way to detect radiation involving cheap ceramics

A crossroads for computing at MIT

Growing our donated organ supply

New AI method captures uncertainty in medical images

• More news on MIT News homepage →

Massachusetts Institute of Technology 77 Massachusetts Avenue, Cambridge, MA, USA

• Map (opens in new window)
• Events (opens in new window)
• People (opens in new window)
• Careers (opens in new window)
• Accessibility
• Social Media Hub
• MIT on Facebook
• MIT on YouTube
• MIT on Instagram

Subscribe or renew today

Every print subscription comes with full digital access

Science News

Robots are becoming classroom tutors. but will they make the grade.

Mechanical mentors try to find their place as teacher’s helper

CAST OF CHARACTERS  Researchers are testing a motley crew of robots to serve as tutors and learning companions for children in classrooms or at home.  

Share this:

By Maria Temming

February 12, 2019 at 7:00 am

Pondering a tablet screen displaying a town scene, a pre-K student tilts her head to the side and taps her lip thoughtfully.

“What are we trying to find?” asks the plush, red and blue robot called Tega that’s perched on the desk beside the girl. The bot resembles a teddy bear–sized Furby.

“We are trying to find lavender-colored stuff,” the girl explains. Lavender is a new vocabulary word. “OK!” Tega chirps.

The girl uses her forefinger to pan around the scene. She eventually selects an image of a girl — not wearing purple. The game puts a red mark through her choice: wrong.

The girl slumps down in her chair, head dropped to her chest as Tega says, “I’m sure you will do better next time. I believe in you.”

The robot, which MIT researchers are testing with students in a Boston-area public school, tilts toward the girl, who leans in close so that her cheek is right next to Tega’s.

Now it’s the robot’s turn. “Time to perform!” it says. The scene on-screen shifts, as though the bot is telepathically controlling the tablet. “Hmm …”

Tega looks up at its partner, as though seeking confirmation that it’s doing this right, and the girl cups the bot’s cheeks encouragingly. The robot looks back at the screen. The girl rests her hand in the robot’s soft fur and murmurs, “I believe in you.”

This kind of tight connection is typical of child-robot interactions, says MIT social robotics and human-robot interaction researcher Cynthia Breazeal. Her team is investigating how this turn-taking robot can help students learn. Kids have a “special kind of affinity” with robots, she says.

Although adults might quickly become disenchanted with machines that aren’t very perceptive or don’t speak more than scripted sentences, children are liable to chat with, listen to and otherwise treat even basic robots as sentient, social beings, says Tony Belpaeme, a social roboticist at Ghent University in Belgium. Researchers like Breazeal and Belpaeme are trying to leverage that connection to create robots that engage with kids as tutors and peer learners.

These robots aren’t meant to replace human teachers, says Paul Vogt, a social robotics and language development researcher at Tilburg University in the Netherlands. But customizable, endlessly patient automatons could provide students with one-on-one attention in crowded classrooms. That extra support may be especially helpful for children with special needs or for students who are learning in a different language than they’re used to, says Belpaeme, who is studying how robots can help immigrant children in Europe pick up a second language.

Robots might also help homeschooled students, proponents say, or teach in areas where human experts are in short supply. English-speaking robots are slated to enter some 500 Japanese classrooms this year for exactly that purpose. Hundreds of Chinese kindergarten classes also have adopted educational robots. But in Western countries, these devices have yet to invade classrooms.

Just like any expensive educational technology, however, classroom robots may never make it to every classroom. Computer and cognitive scientist Brian Scassellati of Yale University and colleagues have had success with a device named Keepon that looks like two stacked yellow tennis balls with eyes and a nose. “When we produce them in the lab, they’re probably costing us about $200 total,” he says. But many researchers use the humanoid Nao robot, which costs several thousand dollars a pop, raising the question of how many schools will be able to afford the classroom helpers.

“There’s a lot of hype about robots,” says Goren Gordon, a natural and artificial curiosity researcher at Tel Aviv University. At this point, most testing has been short-term in small groups of children. So little is known about the potential risks involved when young kids keep close company with automatons. Yet early testing suggests that robots could help students learn new skills and promote good study habits and positive attitudes toward learning. Researchers still have a lot to figure out about best practices and potential impacts if educational robots are going to achieve tenure.

Here to help

Before grading robots on their teaching abilities, consider why automated educators might work better as physical rather than virtual entities. It turns out that a robot’s body may be just as important as its brain. A review of 33 studies that examined how adults and children respond to physically present robots, videos of the robots and animated versions of those same robots revealed that people generally view physical robots more positively and find them more persuasive than videotaped and animated robots. Jamy Li of the University of Twente in the Netherlands reported these results in 2015 in the International Journal of Human-Computer Studies .

“There’s something about robots that sets them apart from a computer,” Belpaeme says. “The exact same content delivered by a robot somehow makes our brains sit up and pay attention…. We don’t yet know why that is.” Still, roboticists have exploited that attention-grabbing edge to build machines that relay information on everything from math to nutrition and sign language.

Of course, a well-rounded education is about far more than learning facts. It’s also about developing good study habits and attitudes toward education that will make students lifelong learners. In this area, robots have proved useful.

On a very basic level, robots can make schoolwork more fun, proponents assert. “If kids enjoy learning, they’re going to learn more,” Belpaeme says. “It’s really as simple as that.” Researchers at the University of Wisconsin–Madison witnessed robots’ power to make schoolwork fun when they designed a bot named Minnie to support children’s reading at home. Minnie, described last August in Science Robotics , comments on a book as the child reads aloud, shows emotional responses to stories and summarizes plot points to support reading comprehension.

Roboticist Bilge Mutlu and learning researcher Joseph Michaelis randomly assigned 24 students ages 10 to 12 to either two weeks of reading aloud alone or with Minnie. Afterward, the solo readers gave the activity more mixed reviews, reporting, for example, “I didn’t not like it, but I didn’t, like, really enjoy it.” Only four said the activity motivated them to read more. Kids in the robot group said reading to Minnie was “fun” and “a cool experience.” Seven students said they felt more motivated to read. 

Robots can also encourage specific reasoning strategies, such as thinking aloud, which is supposed to help students craft more deliberate, organized plans for multistep problem-solving. Computer scientist Chien-Ming Huang of Johns Hopkins University and colleagues programmed a Nao robot to nod along with a child’s speech and remind students who lapse into silence to keep going.

More students who read aloud with a robot companion said that the activity motivated them to read and increased their reading comprehension than students who read aloud alone. 

A robot helped with reading

Source: J. Michaelis and B. Mutlu/ Science Robotics 2018

To test whether this supportive robot helped students learn, researchers randomly assigned 26 kids who were about 11 years old to solve math word problems while thinking aloud with or without the robot’s encouragement. From a pretest to a posttest taken about one week later, the robot-trained children increased their own scores an average of 52 percent ; solo students self-improved by an average of 39 percent, the researchers reported last March in Chicago at the International Conference on Human-Robot Interaction, or HRI 2018.

For a more deep-rooted effect on students’ educational experiences, robots can model certain beliefs about learning, like a growth mind-set: the idea that success comes from effort and perseverance, rather than inherent ability.

In one experiment, 33 children ages 5 to 9 solved geometric puzzles called tangrams with a Tega. Half the kids partnered with a robot that made growth mind-set comments, such as, “You are not afraid of a challenge. I like that!” Other students worked with a bot that stated facts only: “You solved the puzzle.”

Before and after working with the robot, each child completed an assessment that rated growth mind-set from 0 to 10. The growth mind-set cohort’s scores, on average, increased a small amount , 7.63 to 8.06, but the neutral bot group’s scores dropped from 6.94 to 6.59, Breazeal and colleagues reported in Vienna at HRI 2017.

Personalization problems

Although robots show the potential to positively influence students, tailoring a bot’s behavior to an individual is still a major challenge. Roboticists have created machines that can make some simple decisions, like choosing when to encourage a student to take a break . In a study presented at HRI 2017, Scassellati’s team found that when robots offered breaks as a reward for good work, or an opportunity to refocus if a student was struggling, children learned more than if the robot called time-outs at regular intervals.

Designing robots that track student performance to adjust pacing and choose what to teach next is trickier. Some robots have been programmed to adjust activity difficulty based on student proficiency, but researchers have had trouble showing that these bots help students learn more than generic robots do.

What if robots could go beyond responding to performance by keeping tabs on how a student is feeling? Gordon and colleagues at MIT explored this idea by creating a Tega robot that analyzed facial expressions for levels of engagement and valence, which is basically “the goodness of the emotion,” Gordon says. For instance, happiness has positive valence and anger has negative. While working with students on a Spanish-language learning game that involved packing for a trip to Spain, the robot offered various types of feedback, from an excited “Woo-hoo, you’re trying so hard!” to game-related comments like, “The suitcase looks heavy.”

Read the room

One Tega robot watched students’ facial expressions after giving feedback on an educational activity and learned to offer comments that made kids happier. 

“The robot slowly learns which … behaviors result in high valence and high engagement,” and becomes more likely to use those behaviors at the right time, Gordon says. In three to seven sessions over two months, two groups of nine preschoolers worked with either this adaptable Tega or a nonadaptive Tega. From the first to final session, kids in the personalized group generally became more positive about the interaction , with their valence increasing an average of seven points on a scale of −100 to 100. In the impersonal group, valence dropped an average of 18 points, researchers reported in Phoenix in 2016 at an Association for the Advancement of Artificial Intelligence conference.

Robots attuned to students’ thoughts and feelings may make better tutors and learning companions if they can offer the right level of personalization without becoming predictable. But some educators are concerned about the amount of data machines would have to collect and store to do that job right. Human teachers may be able to get a general read on a student’s state of mind. But a robot designed to exhaustively analyze every facial expression or game move a child makes may be able to gather such detailed information on kids that it would constitute an invasion of privacy.

This concern was raised in a series of focus groups with certified and student teachers discussing educational robots. Some participants worried that companies might try to buy that student data from schools, Sofia Serholt, a child-robot interaction researcher at Chalmers University of Technology in Sweden, and colleagues reported in November 2017 in AI & Society .

Robotics ethicist Amanda Sharkey also notes that kids might feel compelled to share private information with a robot peer that acts like a friend. One remedy might include requiring robots to continually divulge what information is being collected and who the robots share it with, says Sharkey, of the University of Sheffield in England.

Social savvy

If privacy concerns about oversharing with robots could be addressed, kids’ comfort with robotic companions could be a strong force for good in the classroom. Scassellati recalls one first-grade boy who worked on English language skills with a robot. “He was so afraid to talk in class, he was so worried about making mistakes,” Scassellati says. But when the student worked one-on-one with a nonjudgmental, patient robot peer, “the first time he made a mistake … and the robot corrected him, he paused for a second, and then he went on, and it was OK.”

Gordon similarly recalls an especially shy student, who “only whispered in your ear; he didn’t talk at all,” he says. But “after the fourth or fifth interaction [with a robot], he started hugging the robot. Every three or four minutes, just stopped and started hugging the robot.”

Capitalizing on this potential for child-robot kinship could help keep students invested even after the novelty effect wears off, so that educational robots don’t end up collecting dust in a corner, Michaelis says. To that end, researchers have begun investigating how robots programmed to be more convivial can better hold students’ attention and improve learning. 

Social robotics researcher Ginevra Castellano of Uppsala University in Sweden and colleagues programmed iCat, a yellow robot with a feline face, to express empathy to test if that would help keep kids engaged with the robot for the long term. Over five weeks, iCat played weekly chess exercises with 16 children in Portugal, ages 8 and 9. The robot, described in 2014 in the International Journal of Social Robotics , monitored the game status and students’ facial expressions and offered advice or emotional support when students looked unhappy . After the first and final interactions, kids filled out questionnaires that rated their feelings of social presence with the robot — that is, how much working with iCat felt like interacting with an intelligent, emotional being — from 1 to 5.

In an earlier study with a similar setup but a nonempathetic iCat robot, kids generally rated their sense of social presence between 2 and 4, and these scores declined between the first and fifth interactions. The empathetic iCat kept the kids at a high level of social presence — between 4 and 5 — from the first through the final session.

But robots’ sociability can be a double-edged, distracting sword, as Belpaeme’s team discovered when using a sociable Nao robot to teach 7- and 8-year-olds in the United Kingdom a strategy for identifying prime numbers. Twelve kids worked with this robot, which used social behaviors, calling the child by name and making eye contact. Another 11 students worked with an asocial bot. From a pretest to a posttest, kids who worked with the asocial bot improved their scores on a 12-point test an average of 2.18 points; the social robot group improved an average of 1.34 points, researchers reported in Portland, Ore., at HRI 2015.

The socially adept bot may have diverted attention away from the lesson ; children spent about 45 percent more time looking at the social robot than the asocial one.

Machine learning

People tend to think that educational robots are ready to replace teachers, says learning researcher Joseph Michaelis. Not close. Even if robots are good at helping kids learn specific skills through highly structured exercises, the machines still need more work to handle many activities.

Robots are good at:

• Playing structured games like chess or snakes and ladders
• Teaching basic math skills or foreign language vocabulary
• Offering scripted responses to books read aloud
• Telling prerecorded stories

Robots struggle with:

• Open-ended conversations
• Dexterity for physical activities such as science lab experiments
• Being engaging without distracting from the lesson
• Keeping students’ attention over the long term

There are other reasons not to make the robots too engaging. Huang likens the dilemma to concerns about excessive screen time, which may put young children at risk for speech delay ( SN Online: 5/12/17 ). “Obviously we have good intentions for these educational robots,” he says, “but the long-term side effects … are unclear.” Some teachers in Serholt’s focus groups expressed similar concerns that kids who spend too much time chatting with robots may lose some ability to decode human facial expressions or the youngsters may adopt more robotic mannerisms.

For Sharkey, “the main concern would be that [kids] come to prefer interacting with the robot.” A robot that’s always encouraging and never disagrees would probably be easier company than other kids. A child who spends more time hanging around agreeable machines than with peers may not develop the social skills necessary to navigate interpersonal conflict, Sharkey says.

Bridges left to cross

So far, investigations of student-robot interactions have typically lasted a couple of weeks or months at most. “What we would want to get up to is a full academic year,” Breazeal says. Roboticists also need to test their technology with children from more diverse backgrounds. Belpaeme and colleagues recently ran an experiment with tutoring robots that helped about 200 children learn a second language. Compared with most educational robot studies, 200 students is a staggering number, says Huang, but “in the real world, this is like nothing.”

Amid questions about how they should or shouldn’t behave, today’s robots are still pretty limited in what they can do. Educational robots are typically designed to work on very specific tasks. The robots still have trouble understanding the high-pitched and grammatically spotty speech of little kids and don’t have the dexterity to participate in many physical learning activities such as science lab experiments.

“We are still a long way” from educational robots that can interact with students like real people, says Ana Paiva, an artificial intelligence researcher at the University of Lisbon in Portugal. Still, it’s difficult to watch a kid doting on a fluffy Tega or making small talk with a seemingly interested Nao and not imagine a future where robots might join teachers and students in class photos.

More Stories from Science News on Tech

AI learned how to sway humans by watching a cooperative cooking game

Why large language models aren’t headed toward humanlike understanding

Could a rice-meat hybrid be what’s for dinner?

How do babies learn words? An AI experiment may hold clues

A new device let a man sense temperature with his prosthetic hand

How to build an internet on Mars

‘Nuts and Bolts’ showcases the 7 building blocks of modern engineering

AI chatbots can be tricked into misbehaving. Can scientists stop it?

Subscribers, enter your e-mail address for full access to the Science News archives and digital editions.

Not a subscriber? Become one now .

MIT Technology Review

• Newsletters

This new system can teach a robot a simple household task within 20 minutes

The Dobb-E domestic robotics system was trained in real people’s homes and could help solve the field’s data problem.

• Rhiannon Williams archive page

A new system that teaches robots a domestic task in around 20 minutes could help the field of robotics overcome one of its biggest challenges: a lack of training data . 

The open-source system , called Dobb-E, was trained using data collected from real homes. It can help to teach a robot how to open an air fryer, close a door, or straighten a cushion, among other tasks. 

While other types of AI, such as large language models, are trained on huge repositories of data scraped from the internet, the same can’t be done with robots, because the data needs to be physically collected. This makes it a lot harder to build and scale training databases.  

Similarly, while it’s relatively easy to train robots to execute tasks inside a laboratory, these conditions don’t necessarily translate to the messy unpredictability of a real home. 

To combat these problems, the team came up with a simple, easily replicable way to collect the data needed to train Dobb-E —using an iPhone attached to a reacher-grabber stick, the kind typically used to pick up trash. Then they set the iPhone to record videos of what was happening.

Volunteers in 22 homes in New York completed certain tasks using the stick, including opening and closing doors and drawers, turning lights on and off, and placing tissues in the trash. The iPhones’ lidar systems, motion sensors, and gyroscopes were used to record data on movement, depth, and rotation—important information when it comes to training a robot to replicate the actions on its own.

After they’d collected just 13 hours’ worth of recordings in total, the team used the data to train an AI model to instruct a robot in how to carry out the actions. The model used self-supervised learning techniques, which teach neural networks to spot patterns in data sets by themselves, without being guided by labeled examples.

The next step involved testing how reliably a commercially available robot called Stretch, which consists of a wheeled unit, a tall pole, and a retractable arm, was able to use the AI system to execute the tasks. An iPhone held in a 3D-printed mount was attached to Stretch’s arm to replicate the setup on the stick.

The researchers tested the robot in 10 homes in New York over 30 days, and it completed 109 household tasks with an overall success rate of 81%. Each task typically took Dobb-E around 20 minutes to learn: five minutes of demonstration from a human using the stick and attached iPhone, followed by 15 minutes of fine-tuning, when the system compared its previous training with the new demonstration. 

Once the fine-tuning was complete, the robot was able to complete simple tasks like pouring from a cup, opening blinds and shower curtains, or pulling board-game boxes from a shelf. It could also perform multiple actions in quick succession, such as placing a can in a recycling bag and then lifting the bag. 

However, not every task was successful. The system was confused by reflective surfaces like mirrors. Also, because the robot’s center of gravity is low, tasks that require pulling something heavy at height, like opening fridge doors, proved too risky to attempt. 

The research represents tangible progress for the home robotics field, says Charlie C. Kemp, cofounder of the robotics firm Hello Robot and a former associate professor at Georgia Tech. Although the Dobb-E team used Hello Robot’s research robot, Kemp was not involved in the project.

“The future of home robots is really coming. It’s not just some crazy dream anymore,” he says. “Scaling up data has always been a challenge in robotics, and this is a very creative, clever approach to that problem.”

To date, Roomba and other robotic vacuum cleaners are the only real commercial home robot successes, says Jiajun Wu, an assistant professor of computer science at Stanford University who was not involved in the research. Their job is easier because Roombas don’t interact with objects—in fact, their aim is to avoid them. It’s much more challenging to develop home robots capable of doing a wider range of tasks, which is what this research could help advance. 

The NYU research team has made all elements of the project open source, and they’re hoping others will download the code and help expand the range of tasks that robots running Dobb-E will be able to achieve.

“Our hope is that when we get more and more data, at some point when Dobb-E sees a new home, you don’t have to show it more examples,” says Lerrel Pinto, a computer science researcher at New York University who worked on the project. 

Artificial intelligence

Large language models can do jaw-dropping things. but nobody knows exactly why..

And that's a problem. Figuring it out is one of the biggest scientific puzzles of our time and a crucial step towards controlling more powerful future models.

• Will Douglas Heaven archive page

OpenAI teases an amazing new generative video model called Sora

The firm is sharing Sora with a small group of safety testers but the rest of us will have to wait to learn more.

Google DeepMind’s new generative model makes Super Mario–like games from scratch

Genie learns how to control games by watching hours and hours of video. It could help train next-gen robots too.

Responsible technology use in the AI age

AI presents distinct social and ethical challenges, but its sudden rise presents a singular opportunity for responsible adoption.

• MIT Technology Review Insights archive page

Stay connected

Get the latest updates from mit technology review.

Discover special offers, top stories, upcoming events, and more.

Thank you for submitting your email!

It looks like something went wrong.

We’re having trouble saving your preferences. Try refreshing this page and updating them one more time. If you continue to get this message, reach out to us at [email protected] with a list of newsletters you’d like to receive.

To revisit this article, visit My Profile, then View saved stories .

• Backchannel
• Newsletters
• WIRED Insider
• WIRED Consulting

Will Knight

Toyota's Robots Are Learning to Do Housework—By Copying Humans

As someone who quite enjoys the Zen of tidying up, I was only too happy to grab a dustpan and brush and sweep up some beans spilled on a tabletop while visiting the Toyota Research Lab in Cambridge, Massachusetts last year. The chore was more challenging than usual because I had to do it using a teleoperated pair of robotic arms with two-fingered pincers for hands.

Courtesy of Toyota Research Institute

As I sat before the table, using a pair of controllers like bike handles with extra buttons and levers, I could feel the sensation of grabbing solid items, and also sense their heft as I lifted them, but it still took some getting used to.

After several minutes tidying, I continued my tour of the lab and forgot about my brief stint as a teacher of robots. A few days later, Toyota sent me a video of the robot I’d operated sweeping up a similar mess on its own, using what it had learned from my demonstrations combined with a few more demos and several more hours of practice sweeping inside a simulated world.

Autonomous sweeping behavior. Courtesy of Toyota Research Institute

Julian Chokkattu

Most robots—and especially those doing valuable labor in warehouses or factories—can only follow preprogrammed routines that require technical expertise to plan out. This makes them very precise and reliable but wholly unsuited to handling work that requires adaptation, improvisation, and flexibility—like sweeping or most other chores in the home. Having robots learn to do things for themselves has proven challenging because of the complexity and variability of the physical world and human environments, and the difficulty of obtaining enough training data to teach them to cope with all eventualities.

There are signs that this could be changing. The dramatic improvements we’ve seen in AI chatbots over the past year or so have prompted many roboticists to wonder if similar leaps might be attainable in their own field. The algorithms that have given us impressive chatbots and image generators are also already helping robots learn more efficiently.

The sweeping robot I trained uses a machine-learning system called a diffusion policy, similar to the ones that power some AI image generators , to come up with the right action to take next in a fraction of a second, based on the many possibilities and multiple sources of data. The technique was developed by Toyota in collaboration with researchers led by Shuran Song , a professor at Columbia University who now leads a robot lab at Stanford.

Toyota is trying to combine that approach with the kind of language models that underpin ChatGPT and its rivals. The goal is to make it possible to have robots learn how to perform tasks by watching videos, potentially turning resources like YouTube into powerful robot training resources. Presumably they will be shown clips of people doing sensible things, not the dubious or dangerous stunts often found on social media.

“If you've never touched anything in the real world, it's hard to get that understanding from just watching YouTube videos,” Russ Tedrake, vice president of Robotics Research at Toyota Research Institute and a professor at MIT, says. The hope, Tedrake says, is that some basic understanding of the physical world combined with data generated in simulation, will enable robots to learn physical actions from watching YouTube clips. The diffusion approach “is able to absorb the data in a much more scalable way,” he says.

Toyota announced its Cambridge robotics institute back in 2015 along with a second institute and headquarters in Palo Alto, California. In its home country of Japan—as in the US and other rich nations—the population is aging fast. The company hopes to build robots that can help people continue living independent lives as they age.

The lab in Cambridge has dozens of robots working away on chores including peeling vegetables, using hand mixers, preparing snacks, and flipping pancakes. Language models are proving helpful because they contain information about the physical world, helping the robots make sense of the objects in front of them and how they can be used.

It’s important to note that despite many demos slick enough to impress a casual visitor, the robots still make lots of errors. Like earlier versions of the model behind ChatGPT, they can veer between seeming humanlike and making strange errors. I saw one robot effortlessly operating a manual hand mixer and another struggling to grasp a bottletop.

Toyota is not the only big tech company hoping to use language models to advance robotics research. Last week, for example, a team at Google DeepMind recently revealed Auto-R , software that uses a large language model to help robots determine the tasks that they could realistically—and safely—do in the real world.

Progress is also being made on the hardware needed to advance robot learning. Last week a group at Stanford University led by Chelsea Finn posted videos of a low-cost mobile teleoperated robotics system called ALOHA. They say the fact that it is mobile allows the robot to tackle a wider range of tasks, giving it a wider range of experiences to learn from than a system locked in one place.

And while it’s easy to be dazzled by robot demo videos, the ALOHA team was good enough to post a highlight reel of failure modes showing the robot fumbling, breaking, and spilling things. Hopefully another robot will learn how to clean up after it.

You Might Also Like …

In your inbox: Will Knight's Fast Forward explores advances in AI

Hackers found a way to open 3 million hotel keycard locks

A couple decided to decarbonize their home. Here's what happened

A deepfake nude generator reveals a chilling look at its victims

Are you noise sensitive? Here's how to turn the volume down a little

Matt Burgess

Steven Levy

Amanda Hoover

Scott Gilbertson

Help | Advanced Search

Computer Science > Human-Computer Interaction

Title: "my unconditional homework buddy:'' exploring children's preferences for a homework companion robot.

Abstract: We aim to design robotic educational support systems that can promote socially and intellectually meaningful learning experiences for students while they complete school work outside of class. To pursue this goal, we conducted participatory design studies with 10 children (aged 10--12) to explore their design needs for robot-assisted homework. We investigated children's current ways of doing homework, the type of support they receive while doing homework, and co-designed the speech and expressiveness of a homework companion robot. Children and parents attending our design sessions explained that an emotionally expressive social robot as a homework aid can support students' motivation and engagement, as well as their affective state. Children primarily perceived the robot as a dedicated assistant at home, capable of forming meaningful friendships, or a shared classroom learning resource. We present key design recommendations to support students' homework experiences with a learning companion robot.

Submission history

Access paper:.

• Other Formats

References & Citations

• Google Scholar
• Semantic Scholar

BibTeX formatted citation

Bibliographic and Citation Tools

Code, data and media associated with this article, recommenders and search tools.

• Institution

arXivLabs: experimental projects with community collaborators

arXivLabs is a framework that allows collaborators to develop and share new arXiv features directly on our website.

Both individuals and organizations that work with arXivLabs have embraced and accepted our values of openness, community, excellence, and user data privacy. arXiv is committed to these values and only works with partners that adhere to them.

Have an idea for a project that will add value for arXiv's community? Learn more about arXivLabs .

The Tech Edvocate

• Advertisement
• Home Page Five (No Sidebar)
• Home Page Four
• Home Page Three
• Home Page Two
• Icons [No Sidebar]
• Left Sidbear Page
• Lynch Educational Consulting
• My Speaking Page
• Newsletter Sign Up Confirmation
• Newsletter Unsubscription
• Page Example
• Privacy Policy
• Protected Content
• Request a Product Review
• Shortcodes Examples
• Terms and Conditions
• The Edvocate
• The Tech Edvocate Product Guide
• Write For Us
• Dr. Lynch’s Personal Website
• The Edvocate Podcast
• Assistive Technology
• Child Development Tech
• Early Childhood & K-12 EdTech
• EdTech Futures
• EdTech News
• EdTech Policy & Reform
• EdTech Startups & Businesses
• Higher Education EdTech
• Online Learning & eLearning
• Parent & Family Tech
• Personalized Learning
• Product Reviews
• Tech Edvocate Awards
• School Ratings

10 Questions To Get Acquainted With Your Online Students

Best books for your children, public universities are failing the public, institutions must change how they handle sexual assaults, indications that your university is on the verge of shutting down, an age-by-age approach to disciplining your children, common elements of instructional systems design, great education leaders solicit feedback, great education leaders are instructional leaders, great education leaders are accomplished mediators and negotiators, why do homework, when you can make a robot do it.

Robots are infiltrating our daily lives. They are cleaning our floors, driving us around on campus or the golf course, operating on people, flipping hamburgers, and they’re teaching children in Asia. There are even robots like mBot that teaches kids about robotics.

Robots are meant to automate repetitive tasks, leaving humans free for more creative and rewarding endeavors. And what is more repetitive than having to copy text from a book several times?

The writing robot that can copy texts for students

Enter a very enterprising young Chinese teen. She used the money she received over Chinese New Year to buy a so-called copying robot to do her homework for her. In all fairness, the homework, in this case, was to copy passages from a textbook. Chinese teachers give this kind of homework, even during the Lunar Year holidays, so students can practice the thousands of Chinese characters that make up Mandarin Chinese.

The device, a metal frame, and pen comes with the marketing message that it could “imitate all sorts of handwriting”. In her case, the device performed perfectly, finishing all her homework in record time and in her handwriting. The incident was reported , including the destruction of the device by the girl’s mother. Far from being criticized for her dishonesty, the girl was praised by netizens for her initiative and the company that manufactures the writing robot has been inundated with inquiries ever since.

In an investigation, the South China Morning Post found several entries for a so-called “copying robot” on the e-commerce platform Taobao. It seems like a good idea to let a robot do something for you that is boring and repetitive, but you have to ask yourself if that’s a sensible thing to do. Surely you miss out on a learning opportunity if you let a robot do your work for you.

That doesn’t mean that robots cannot be useful when it comes to school homework. Rather than doing the work for you, your robot can help you to do the work and master the concepts in the curriculum. 

The robot that can do math homework

For kids who might feel burdened by too much homework, the future holds much promise. Christopher Yan and Derek He have come up with a concept of a homework-solving robot. Their prototype can solve math homework. What’s more, when it’s completed, the robot will be able to write the answers according to the user’s handwriting, so a teacher won’t suspect that the work was not done by the student. Their project known as Homework Help , picked up the top honors at this year’s PennApps, one of the largest student hackathons in the world, held at University of Pennsylvania.

It seems there is a need for robots that can help students with their homework. Scientists have noticed that children form close emotional bonds with social robots. It is not yet clear why this is so, but educators can exploit this to help students with their studies. Robots can, and do, add a fun aspect to learning. They can also add a fun aspect to homework.

Harnassing Edtech’s Disruptive Power

Technology collaboration results in improved success.

Matthew Lynch

Related articles more from author, stop expecting artificial intelligence to do it all.

Cutting Campus Costs By Integrating AI Tools Into The Classrooms

What is backup software.

Your Next Tutor May be a Digital Voice Assistant

How to use siri on spotify.

AI Curricula for K-12 Classrooms

The AI That Does Your Homework

Upload a worksheet or use a homework ai helper that is specially designed to provide you with all your homework solutions..

We have a number of AI bots that are specially designed to help with all subjects and all grade levels.

Just take a picture of your homework and our high powered AI will provide you with all the answers.

4 AI Tools For Students

Oddity AI is the groundbreaking new artificial intelligence that students have been dreaming of for years. This AI homework helper has revolutionized the way people approach their schoolwork, making it easier than ever to get work done quickly and efficiently. With Oddity AI, users can ask questions directly to the AI and receive instant answers in seconds. The AI not only gives accurate answers but also provides detailed explanations of how each answer was reached so users can gain a deeper understanding of their schoolwork.

AI Essay Writer

Write high quality essays, in your own writing style, in seconds!

Math Answers

No matter the math problem, simple or advanced, OddityAI's powerful homework helper can solve the problems for you.

Flashcard and quiz generator

Upload a picture of your hand-written notes and automatically convert them into flashcards and quizzes to help you study.

English Help

From grammar to literature analysis, our AI is here to assist you with any task you may be struggling with.

Science Answers

From biology to chemistry to physics, our homework AI is equipped to assist you with any topic you may be struggling with. Need help understanding a complex scientific concept? Want to know how to complete a lab report? Our AI that does homework is here to provide you with accurate and reliable answers.

Homework Help Robot Does Your Math Homework

This particular project for both, known as Homework Help, picked up the top gong at this year’s PennApps, which happens to be one of the largest student hackathons in the world, where it was held at the prestigious University of Pennsylvania. Just how does Homework Help perform such a “miracle”?

Well, it will rely on a camera that will boast computer vision in order to identify simple math problems, before it moves to solves them, while writing out the answer with a pen in a mechanical manner. It is capable of measuring the distances between characters in order to make sure whether they are part of the same number, while there is some space beneath them which will allow it to figure out just where the answer should be written. Forget about it solving complex math problems at this point in time, as it does only simple addition problems. Perhaps the future might see a more impressive repertoire.

Filed in Robots . Read more about Science .

Discover more from Ubergizmo

Subscribe now to keep reading and get access to the full archive.

Type your email…

Continue reading

What We Guarantee

• No Plagiarism
• On Time Delevery
• Privacy Policy
• Complaint Resolution

We've detected unusual activity from your computer network

To continue, please click the box below to let us know you're not a robot.

Why did this happen?

Please make sure your browser supports JavaScript and cookies and that you are not blocking them from loading. For more information you can review our Terms of Service and Cookie Policy .

For inquiries related to this message please contact our support team and provide the reference ID below.