Application Critique – PhotoMath

As CS3216 Students, we are interested in leveraging on existing technologies to bring innovative application ideas alive. Hence, the most interesting points that I gathered out of Group 8’s presentation on PhotoMath are lessons learnt, which I can apply to other future ideas.

Lesson #1: Applying OCR Technology to other contexts

Group 8 mentioned innovative ways to leverage on Optical Character Recognition (OCR) technology, most pertinently how it can be applied to replacing manual data entry roles. A quick search on job portals in Singapore would show the prevalence of data entry jobs today, and this is a pain point faced across even developed societies. A well implemented data entry OCR application could have the potential to disrupt and replace the data entry market. 

Lesson #2: Make PhotoMath “Smart”

Group 8 mentioned how adaptive learning could be applied to PhotoMath to deliver math content to the user that is most relevant to him, based on the questions that he scans. This could potentially be used to deliver targeted math syllabus to students, and help users reinforce their learning. With all the buzz in “smart” applications, we learn how to appropriately leverage technologies to create smart applications that is targeted at specific pain points, like how “smart” PhotoMath proposes to help students learn better.

Lesson #3: How to manage unintended negative effects of PhotoMath

PhotoMath’s main target audience of students who use their application regularly are now empowered to cheat easily. Given the prevalence of mobile devices amongst students, it is easy for students to use PhotoMath to cheat, resulting in detrimental effects to these student’s education. PhotoMath could face controls by regulatory bodies to limit its use among people who abuse it. This leads us to remember to consider how to address some of the repercussions of our proposed application.

Final thoughts: Feasibility of OCR based apps in solving manual Data Entry 

OCR technology is now actually very prevalent, with many new and existing players in the market. Some of these players like Microsoft Office focus on alternative problems like utility applications for consumers to convert PDF documents to word documents.

However, I feel that the data entry pain point mentioned is a real issue that many companies face, and with enough research and analysis in this, there lies a potential to develop a targeted application which could eliminate this pain point. Other enterprise solutions like trapeze exist [1], but they have not gained enough traction. Even large and commercially successful companies like Apple did not have first mover advantage in most of their products, but still managed to capture a huge market share, because of the fact that they are able to make user friendly products that address specific and huge pain points [2]. If anyone can somehow make use of OCR algorithms to do this, it could potentially reap huge benefits.

 

[1] http://www.cvisiontech.com/products/trapeze-for-forms-processing.html

[2] http://www.businessinsider.sg/apple-second-mover-advantage-2015-6/?r=US&IR=T#zSyRHtLSeJ40RlVz.97

6 thoughts on “Application Critique – PhotoMath

  1. Cool that you highlighted Group 8’s point about being wary of negative repercussions (i.e. cheating), although my guess is that this would be less of a problem in higher education since students at that level tend to take their work more seriously.

    Currently, the set of problems which PhotoMath can tackle appears to be composed primarily of easily solvable ones. Answers to such problems can usually be found online, without complications. On the other hand, answers to more complex, university-level problems are not always accessible (think CLRS for Algorithm analysis). Perhaps PhotoMath can focus more its technology and develop it to the extent that it can tackle questions of such sort, in order to broaden its user demographic and increase the meaningfulness of its usage.

    Liked by 1 person

  2. Nice summary of my group’s presentation. 😀

    I agree that one of the pains that MicroBlink’s OCR can solve is data entry. Data entry normally takes up man-hours and money, and could potentially save a ton of costs for companies that have to process a lot of forms.

    For PhotoMath, one of the largest concerns is the way students are using it to cheat on their homework, and that students could become over-reliant on it instead of thinking and learning for themselves. However, instead of having regulations on PhotoMath, I think the onus is on educators to rethink the way they teach and make it so that tools like PhotoMath can be incorporated into the syllabus and become an aid instead of a crutch for students. Instead of tying down innovative tools like these, it makes more sense to welcome innovation and work alongside technology to educate students.

    I think one of the most common reactions to PhotoMath was: ‘I wish I had this when I was in school!’ indicating that there’s a problem with education if students are so eager to copy their homework instead of learning and actually doing it. There could be too much homework, or students may not have interest in the topics taught. After going through the Singaporean education system, I’ve had quite a few moments like ‘Why didn’t they teach this in school?’ while reading articles or watching videos. Innovation in education is probably going to take a while, but perhaps PhotoMath will be one of the leaders. 🙂

    Like

  3. Hi Kai Yi,

    Regarding applying OCR Technology to replace manual data entry, I don’t think it is as easy as that.

    There are jobs that technology shouldn’t replace because there are people who need those jobs. While the technology might be ripe for disruption, not everything should be replaced without considering the repercussions of the move.

    Many will have their livelihoods affected if technology were to take over their job. Although it might not be the most efficient way of handling something like data entry, there are problems in society that cannot be solved if we were to replace these segment of the workforce by technology.

    There are companies and social enterprises in Singapore who employ people with disabilities and the jobs provide a way for them to spend their time in a meaningful and engaging manner.

    In my visit to Enabling Village this year, I witnessed firsthand how they were using technology to positively affect the lives of people living with disabilities. For instance, they used a 3D printer to create a pencil holder for one of the beneficiaries so that he could hold the pencil steadily in his hand for writing.

    I hope technology can be used carefully so that it does not displace a segment of the job market who need their jobs for survival and sustenance.

    Like

    • Though we should not neglect the problem of job loss, we also have to remember that humans don’t live to do mindless work. Jobs will not disappear; they will just shift. Ideally, we want to train our workforce to have a new set of skills that our advancing society requires, such as being the ones to design or build such automation. It will be up to the people and the governing bodies to ensure such social safety nets and to embrace such a future.

      Like

  4. Hi Kai yi,

    With so much discussion above on OCR technology and myself taking Natural Language Processing as well as Machine learning this sem, I can’t help but weigh in on the subject.

    The OCR promised by photomath is still based on type-written values. As with all AI-related problems, we usually have to look at the problem environment and see if we can simplify the problem. In this case, involving hand written values creates a level of uncertainty that makes the problem much more difficult. Limiting the OCR to mathematical problems however, ensures that the problem (and thus the training sets) becomes smaller and easier to handle. Having OCR enhancing data-entry jobs seems to be the goal of MicroBlink’s OCR technology.

    I agree with Ryan above that humans don’t live to do mindless work. While Derek’s concerns of replacing the job market is valid both in the ethical aspect and the development of better technology (as when people feel they are being replaced by machines they will not support development of better tech), I feel that he is approaching the solution using OCR for data-entry jobs wrongly. As you have have noticed, I used the word enhancing instead of replacing in my previous paragraph.

    OCR will unlikely ever be perfect. Even I myself can’t guarantee I will be able to read my own handwriting ten years after writing something. If OCR were ever to be used for data-entry, I believe that there will still have a need for low-skilled workers to identify texts and characters that the OCR cannot confidently detect. This is where Ryan’s point comes in. The jobs of the people who are currently doing data-entry will just be shifted to do things that humans do better than robots, making use of context as well as intuition to identify solution. In this case, low-skilled workers will still be needed to monitor the process of data-entry by the OCR.

    -Zhi Yang

    Like

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out /  Change )

Google photo

You are commenting using your Google account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s