Future of AI in education

From: redpointai

The future of education is poised for significant transformation with the integration of AI. Salman Khan, founder and CEO of Khan Academy, envisions a classroom 20 years from now that amplifies the best aspects of current great classrooms, while also introducing new immersive learning experiences [02:04:00]. Khan Academy, serving over 150 million learners across 190 countries, has already deployed their AI-powered tutoring assistant, Khanmigo, to over 1.4 million students and teachers [00:30:00] [00:40:00].

Envisioning the Classroom of Tomorrow

In the next 5 to 20 years, classrooms are expected to evolve significantly, moving beyond passive lectures [02:06:00].

• Student Engagement [02:18:22]: Classrooms will prioritize engagement, interaction, problem-solving, group work, and presentations [02:22:00].
• Teacher Empowerment [02:33:43]: Teachers will be more mobile, actively working with students. AI will provide “superpowers” by assisting with lesson plans, grading, progress reports, and offering insights into student needs and classroom management [02:43:00] [03:11:00].
• Ambient Technology [03:50:00]: Generative AI could become ambient, observing classroom dynamics and providing insights without requiring constant screen interaction [03:48:00] [03:51:00].
• Immersive Learning [04:15:00]: The mainstream adoption of virtual and augmented reality, combined with generative AI, will enable immersive simulations and virtual worlds, allowing students to “go back to ancient Rome” for interactive learning experiences [04:17:00] [04:26:00].

Khan Academy’s Approach to AI in Education

Khan Academy’s fundamental mission is to provide personalized, high-quality education, a goal that aligns perfectly with the capabilities of AI [05:08:00].

Khanmigo: The AI Tutor and Teaching Assistant

Khanmigo, launched after gaining access to GPT-4, aims to approximate tutoring and teaching assistance [06:08:00] [06:20:00].

• Features [06:30:00]: It includes guardrails for teachers to monitor student activity, prevents cheating, ensures safety and privacy, and employs a Socratic method to promote good pedagogy [06:34:00].
• Adoption [06:55:00]: Initially expecting 100,000 users by 2025, Khanmigo is now used by approximately 1.3 to 1.4 million teachers and students [06:57:00] [07:00:00].
• Proactive AI [07:33:00]: The next phase involves a more proactive AI, known as Khan Academy Classroom, which acts as a concierge, welcoming students back and suggesting relevant activities [07:54:00] [08:00:00]. For teachers, it will offer better insights and management tools [08:09:00].
• Reducing Errors [01:03:00]: When anchored to Khan Academy content, Khanmigo’s error rate is about 2%, split between math errors and evaluation errors [01:03:00] [12:57:00]. This is considered better than many human tutors [13:43:00].

Writing Coach: Combating Cheating with AI

Khan Academy’s Writing Coach addresses concerns about cheating by integrating AI ethically [15:41:00]. Teachers can assign tasks through the AI, which acts as an ethical writing coach, helping students brainstorm, outline, and draft [15:46:00]. When a student submits work, the teacher sees not just the final output but the entire process [15:57:00]. The AI identifies if content was copied from external sources like ChatGPT, thereby undermining all forms of cheating [16:05:00].

Teacher Integration and Impact

Teachers are leveraging AI to streamline their daily tasks and enhance classroom activities [14:10:00].

• Lesson Planning [14:16:00]: AI helps create and tweak lesson plans, making them more engaging and appropriately sized [14:18:00].
• Content Generation [14:34:00]: Through a partnership with Blooket, Khanmigo can generate question sets for in-class games in minutes, a task that previously took teachers half an hour to an hour [14:36:00] [14:40:00].
• Immersive Learning [15:22:00]: Teachers are using AI simulations, such as talking to historical figures like Harriet Tubman or George Washington, to create highly engaging class openings [15:25:00].

Challenges and Opportunities in AI Adoption

While the potential of AI in education is vast, several challenges exist.

Engagement and Proactive AI

A significant hurdle is student engagement, as only a small percentage of students proactively seek out AI assistance [07:47:00]. The shift to proactive AI in Khan Academy Classroom aims to address this by making the AI a more central and interactive part of the learning journey [07:54:00]. The most effective approach involves teachers assigning AI-driven tutoring sessions and holding students accountable, which significantly boosts engagement [34:05:00] [34:40:00].

Policy and District-Level Adoption

Schools are expected to be among the first places to see mainstream adoption of AI for productivity and learning [19:12:00].

• Cost-Effectiveness [19:47:00]: AI solutions are dramatically cheaper than traditional interventions like live tutoring (e.g., $10-$15 per year compared to $25-$50 per hour) [19:48:00] [20:18:00].
• Teacher Benefits [21:03:00]: Districts are reporting that AI saves teachers at least 5 hours per week, making it a valuable recruiting and retention tool [21:03:00].

Technical Challenges and Future Capabilities

Building effective AI for education requires significant development beyond simple prompting layers [21:32:00].

• Guardrails and Moderation [21:48:00]: Implementing robust safety and moderation features, including managing false positives, is crucial for educational settings [21:48:00].
• Math Accuracy [22:07:00]: Ensuring high accuracy in math, particularly with complex evaluation errors, is an ongoing area of focus [22:07:00].
• User Interface [22:35:00]: Developing natural and intuitive user interfaces that integrate AI seamlessly into various aspects of learning (e.g., brainstorming tools, outlining tools, drafting assistance) is vital [22:37:00].
• Desired Capabilities [23:46:00]:
  • Improved Memory [23:58:00]: Enabling AI models to retain context and memory over longer interactions is key for personalized learning [24:07:00].
  • Advanced Voice Integration [24:26:00]: Integrating advanced voice capabilities for more natural student-AI interactions [24:34:00].
  • Higher Quality Questions [24:58:00]: Models need to generate more reliable and high-quality questions [25:01:00].
  • Vision Capabilities [25:30:00]: Allowing AI to “see” and understand student work (e.g., handwritten math problems on a tablet) to provide immediate, natural feedback [25:34:00].

Model Evaluation

Khan Academy uses a rigorous evaluation framework for AI models [29:08:00]:

• Tough Test Cases [29:16:00]: A battery of several hundred “tough test cases” (e.g., specific math problems, nuanced evaluation scenarios) are used to assess model performance [29:24:00]. Initially, models failed on 70% of these, but this has improved to under 10% [29:46:00].
• Machine and Human Labeling [30:10:00]: AI is used to flag potential errors in interactions, which are then verified by human labelers to determine accuracy and conversation productivity (e.g., student engagement vs. disengagement) [30:13:00] [30:44:00].

Impact on Skills and the Workforce

The rapid pace of AI advancements necessitates an adjustment in the skills taught to students for the future workforce [42:40:00]. While foundational skills like writing, reading, and math remain crucial [43:10:00], entrepreneurship – the ability to combine existing resources in new ways to create value – will become increasingly vital [43:26:00]. This means individuals will need to continually experiment with new tools, integrate outputs, and use their critical thinking skills to refine AI-generated content [43:58:00] [44:17:00].

The "Blank Screen Problem" [08:19:00]

Users often face a “blank screen problem” where they can prompt anything but struggle to know what to prompt effectively [08:21:00]. This highlights the need for AI to be more proactive and guide users, rather than simply waiting for commands [08:57:00].

Broader AI and Education Market

The AI in education market is currently characterized by a lot of “noise,” with many startups offering thin prompting layers over existing models [38:25:00]. While AI enables faster app development, it also lowers barriers to entry [38:46:00]. Non-profits like Khan Academy have an advantage due to their ability to take a longer-term view on development and build trust with the education community regarding pedagogy, efficacy, and ethical use of AI [39:06:00] [39:51:00].

Other areas ripe for AI innovation in education and workforce development include:

• Interviewing and Assessment [40:33:00]: Streamlining the resource-intensive and often broken interviewing process to create more standardized, efficient, and equitable assessments [40:37:00].
• Corporate Training [42:11:00]: Using AI simulations for more engaging and interactive corporate training on topics like cybersecurity or harassment [42:13:00].

Evolution of AI in Education

The perception of AI in education has rapidly evolved [44:50:00]. Initially, there were concerns about errors, hallucinations, bias, and cheating [49:30:00]. However, the focus has shifted to how AI can empower teachers and engage students in productive ways [45:08:00]. Organizations like Khan Academy are now self-identifying as “AI-first organizations,” indicating a rapid adaptation within the education community [48:16:00] [48:36:00].

The Future of Learning

“I think there’s a world where artificial intelligence can give more teachers, all teachers some time back on their planning side of things, giving them better insights on how do they uh well some ideas and insights on how they can better manage their classroom, better insights on where the students are at any point in time, better interactivity with their students.” [03:11:00]

The vision for AI in education is to move towards a future where it not only supports learning but also enables new forms of credentialing, such as high school and college credits and diplomas, making world-class education more accessible globally [37:39:00].