FESA

Create and implement a program powered by natural language AI technologies to automatically generate didactic material to improve the learning process of high school students with personalized material.

The specific problem we are trying to solve is the quality of education in Peru. Currently 9 million[1] students enrolled in the country's national education system are affected. The factors that contribute to this problem in our community include the majority of the authorities of the educational system and older teachers with old ideas, corruption, low quality teachers and teaching methodology. The consequences that derive from this problem include poor academic performance. On the other hand, schools that have decided to use more recent methods and technologies have better academic results[2]. According to the PISA Test, Peru in 2018 was ranked 64 out of 77 countries evaluated [3].

[1]https://m.inei.gob.pe/estadist...

[2]Personal experience as a student in a fully digital high performance school. Alvaro Zevallos 

[3]https://elcomercio.pe/peru/pru...

The solution consists of a system that generates personalized educational material using natural language AI. The technology uses the latest natural language AI technology which is currently OpenAI’s GPT-3. The generation of the didactic material will be done using certain inputs of information such as the course, the agenda approved by the government and the notes of the previous period of the students to generate a material for each school year. The output (in phase 1) will consist of textbooks and teachers' guides.

For example, in phase 1, a block of educational material is generated per school each year. To create this block, 2 requests are made per course, per high school year. A request for the book and another request for the teacher's guide.

Each request needs a specific set of information to generate the output.

book = Request(Course_name, is_book, This year’s Syllabus, Last year’s Syllabus, Last year grades, specific context)

professor_guide = Request(Course_name, is_guide, This year’s Syllabus, Last year’s Syllabus, Last year grades, specific context)

The response should be some sort of text and images to convert into a PDF.

Specific context parameter is extra information such as school’s name, location, rating and other considered relevant.

The Request function of the system is the core functionality and utility. It will transform al the provided input into a query useful for GPT-3 to obtain the expected result. Extra details such as fine tuning will be required and control over the results for the first few phases.

The target population is the students of the national education system in primary and secondary schools. Their main need is that the subjects they are taught, and the ways in which they are taught, could be improved. For this reason, our solution generates both textbooks and didactic guides so that teachers have a better method at hand than the one they already use. This solution is going to be implemented in 4 gradual phases to adapt to a country that is averse to change. Some institutions could advance directly to a more advanced phase due to technological advantages in their facilities.

In phase 1, an educational material of the book or printed type will be made, which will be separated by each school, so that each school will have a different book, with different difficulties but following the same national agenda, these books that are updated every year for each school using student grades as feedback.

In phase two, the focus will be changed instead of per school to educational classrooms. Each classroom will have different educational material, a different syllabus, adapted to the needs and learning speeds of the children belonging to the classroom such that one room will have certain educational material, possibly books, and another room will have another book that leads broadly the same themes but with different difficulty.

  For phase three, it is the first phase that needs devices, a large number of technological devices, in this case laptops, tablets or similar, so that teachers can manually enter student grades so that artificial intelligence can make a classification of students based on levels and allows the generation of didactic or educational material according to each student. For practical purposes, a separation by learning groups could be implemented within the classroom, such that there are three or four different educational materials and distributed individually to each student, that is, students of different classificationlevels. high level students all have the same book, middle level students all have the same book, low level students all have the same book, and so on.

  And for phase four, each student will have access to a tablet or laptop where they spend most of their academic time. In this case, the material generated by artificial intelligence consists of each question, each topic they see, practically instantly. For example, if the student is solving an exam or is doing a guided practice, the questions that appear later in the exam adapt gradually to the ideal range of learning that that student should have to accommodate as much as possible to their learning curve.

  Those would be the four phases of implementation of this long-term plan to adapt the educational system to the new technology that is the GPT-3 or more advanced AIs in the near future.

With all these adversities we have been able to learn everything thanks to the internet, being able to find out about YouTube videos, being able to read about science, physics, mathematics and current affairs. We have developed skills and we can use technology that we could never have learned in school and we have been able to stand out and find what we are passionate about, such as robotics, to be able to exercise it by doing projects to help others.

Selene Cordova made a very simple recyclable project and she proposed it in a regional contest in Callao and she was able to win 1st place.

Selene Cordova and Eduardo Vergaray were able to win the national robotics contest at the national level with a lot of effort. We were able to execute a project that had an impact on the public, gaining experience.

Eduardo Vergaray also won many more international and national competitions. Fernando Quispe is a leading member of the robotics club at his school, he created multiple independent robotics projects that unfortunately didn't win in the first place, but he is enthusiastic about learning about it. And Alvaro Zevallos graduated in mechatronic engineering from UTEC with a scholarship. He is currently doing his thesis project in a collaboration between UTEC university and Harvard university. He also has 2 years of research and development experience; in addition to half a year of work experience developing software with the latest technologies. During university he specialized in robotics and artificial intelligence and in his current job he is part of the team that is developing a technology also based on GPT-3 called Blob to automate software development.

Our team has first-hand experience on how to get ahead despite the failures in education. We feel capable of knowing the problem and being able to improve the essential problems that affect the students of our Country.

There are 9 million students studying in a deficient education system and we, using AI technology, can create a tool and an implementation plan for our solution in a country averse to change. This solution can change the lives of those students and give them the opportunity to find their passion, make this form of education globally expandable and achieve true universal education.

We very often participate in contests, above all to be able to demonstrate all the talent we possess and show the world that education is the most essential thing for the prosperous future of humanity.

Being a team that grew up studying in this system and had to fight hard to excel in our passions and use that experience to generate  solutions.

Finding out more in depth about the subject, there are more than 6 million students who have a bad education and are likely to lead society to the increase in unemployment and informal jobs, the accelerated growth of crime, and young people who are involved in crime, drug addiction, and worse.

Our solution is to apply AI chat GPT-3 to use AI as an assistant that can help teachers do a better job of teaching, and give students much more control over their learning system. Giving them a more comfortable environment to learn.

There are two main reasons why the solution is innovative.

First is the use of the new GPT-3 technology and with the latest implementation of this at the end of November, chatGPT, a high level and high complexity AI has been achieved for the first time such that we as users find ourselves in need to improve our requests in order to extract the full potential of the technology. Some have come to consider it at the level of an industrial revolution and we agree with that statement. The world is going to change faster than we expect to finally accept AI in our daily lives and it is our duty to bring this revolution to all possible fields. In this case the our country's education

In the same way, the second reason lies in the extension of our proposal. We have not only proposed the solution but also the gradual process for implementation considering a context where our country tends to reject change. The implementation phases are explained in the question “What steps have you taken to understand the needs of the population you want to serve?” . This plan was devised to initially be accepted by some highly adaptable and innovative schools to gradually advance in the private school market. So that in the future our nations accept us as the official provider of educational material.

We aspire to get our program fully functional for the next year. As a software developer i think most of the development of the platform could be done in half a year. And with half a year remaining to advertise it. 

Have an agreement with at least 1 school. And we already have the contact information of some of them from relatives and friends.

Have one full-time employee doing research to adapt our system to future GPT versions. It is important to adapt in this fast changing world.

Technically speaking this solution is rather simple. for the first two phases the stack of technologies will be GPT-3 as the core for generation of the material. And as for the service we can run this system in a React interface to access through web and domain. This react service can use an API to connect to the GPT-3 service to send the necessary prompts and receive the result and generate a file as output.

For later phases we can add a mobile app using flutter and dart to connect it to our service and work in real-time material generation. A strong database would be necessary for this last phase to save all the information of each student.

For the next year we’re planning to have an agreement with at least one highschool On average each school has 150 students per year. Extending this initially for 5 years of high school makes about 750 students

GPT-3 in its tests has good results but it still does not have 100% reliability to be able to affirm that it is reliable in everything.

In Peru, teaching is still very slow since books are used for everything and new technologies are left aside and there is no opportunity to try them.

Corruption in Peru is a problem that not only covers the field of education but also all political aspects of the country and could be an impediment as they tend to reject change and serve mostly their personal needs

A cultural impediment is the lack of education of the majority of the parents that is transmitted to their children and they exercise what they have learned and despise the study to focus on improper and illegal activities.

At the moment we have no partnerships yet.

The business model will be first to offer our service to private schools, since we have the contact to offer our service to generate them for each educational material and they would be charged a cost for each book or guide. Later on when there are materials for each student we may change/add the option to a subscription model.

Actually one of our team members has some savings for investments and has told us that he is willing to invest some of it in this solution. We consider an approximate cost of 1000$ to make this solution ready for the first two phases, and the savings even add up to go for advertisement for a few months until we end up with our first contract or agreement. We’re also willing to go into start-up boosters and look for other investments.