Student-oriented instruction, while appealing, has been found to decrease math scores in PISA exams, while teacher-directed instruction increases them. A good teacher can distill their knowledge and make learning more efficient. aei.org/wp-content/upl…
Executive function (EF) proficiency is greatly correlated with school achievement. Early EF proficiency is related to early math proficiency and in turn, early math proficiency is related to reading achievement MORE THAN EARLY READING proficiency (sciencedirect.com/science/articl…)
Word problem solving is much more related to the ability to make accurate visual representations than to reading comprehension itself. Students who made accurate visual representations were 6 times more accurate when solving word problems: dx.doi.org/10.1016/j.ijer…
Using a digital whiteboard in a classroom but using it as a normal whiteboard isn't going to raise engagement, you must add multimedia resources, access to apps, and, if possible (if students have devices of their own) screen sharing!
Math as standard-based blocks that build on top of one another is a good starting point, we need also need problems that integrate many building blocks and give room for the students to critically think which requires proficiency in EF. Are we preparing correctly our students?
Given the high cost required to keep cheating at bay in public schools, the only solution might be AI-based too: AI monitoring, AI assessment review, and AI assessment creation (or variation). edweek.org/technology/opi…
From understanding that AIs are not real people to understanding their limitations, this articles walks us through different different developmental stages regarding use of AI for K-12 students. Interesting read! edweek.org/technology/wha…
Texas is taking an important step to democratize #AI Education. #edtech texastribune.org/2024/03/12/tex…
As AI is increasingly recognized as a necessary tool for tomorrow's (and today's) workforce, universities are starting to adopt new strategies around it. Look what happening in Arizona! forbes.com/sites/jefffrom…
As LLMs get better, they'll naturally enhance their logic capabilities. In the meantime, you can use other LLMs with same/similar results with some simple prompting (2-liner with simple, natural language instructions in GPT-4): chat.openai.com/share/40be47d8…
Three models, three different answers 😎 Claude 3 is AGI confirmed Separately, what will it take to get a model to actually ask "do you want the answer for the beginning or end of day 4"? this question as stated is ambiguous
One major concern in the use of AI in Education is the reliability of the answers provided. Often times we've seen LLMs explaining something false in a plausible way that makes sense; anyone learning will most likely not know that the explanation is wrong. arxiv.org/pdf/2402.04614…
GPT-4 can't handle lots of information at once without losing coherence. Gemini 1.5 will have almost 8 times more capacity than GPT-4 Turbo, meaning it can process more information with better accuracy. This will greatly impact AI-tutors and educational applications of AI!
Gemini 1.5 Pro's 1,000,000+ token context length is incredible. I got early access and spent my Saturday night running tests. Here are 6 impressive capabilities I found:
As this memory capacity grows, we'll be able to create really personalized tutors for each student. This enables not only an academic AI tutor but a real AI tutor that can tailor lessons and examples, and understand the complexities, nuances, and pain points of each student.
We’re testing ChatGPT's ability to remember things you discuss to make future chats more helpful. This feature is being rolled out to a small portion of Free and Plus users, and it's easy to turn on or off. openai.com/blog/memory-an…
No surprise our schools are not preparing students for an AI-driven world. The Educational system must adapt ASAP for our kids to thrive. Understanding AI's pros and cons, when and how to use it, and reshaping assessments is a good starting point! k12dive.com/news/future-is…
Pretesting (taking tests before the instruction) dramatically affects students' performance. This could be difficult to do, given it requires teachers to do extra tests. How can Educational AI solve this? I read you! techlearning.com/news/the-power…
techlearning.com
It’s Counterintuitive But Pretesting Consistently Works, Says Research
Research and tips for implementing pretesting in the classroom from Steven C. Pan, director of the Learning Sciences Laboratory at the National University of Singapore
Ever wondered how LLMs (natural language AI) learn? This amazing example shows us a practical example of the process and illustrates how LLMs, even not understanding what the words mean, can give us coherent answers to our queries. Share your thoughts! nytimes.com/interactive/20…
nytimes.com
Let Us Show You How GPT Works — Using Jane Austen (Gift Article)
An interactive explanation of how language models learn to mimic language, from Shakespeare to Star Trek.
This tool not only simplifies the usage of spreadsheets but also enhances it by adding LLM capabilities to it! It also reduces the entry barriers for spreadsheets ("coding") enabling a smoother transition of young learners to data analysis.
This is going to change how people use spreadsheets! Introducing @NumerousAI, a tool that allows you to prompt ChatGPT in bulk right from your spreadsheets! Here's what you can do with this: - Content writing based on other cells - Summarization, translation, and rewriting -…
LAMs could significantly outperform LLMs in education tasks, especially for HS and college math and science content, given LLMs tend to have lower performances in logic- and calculation-related subjects. The future is ever brighter for AI in Education!
Large Language Models (LLMs) vs Large Agentic Models (LAMs) LLMs so far have demonstrated exceptional proficiency in formal linguistic competence, excelling in tasks that involve language understanding, general reasoning and generation. However, the capabilities in functional…
Incredible time-saving tool to iterate your LLM prompts until you get the desired output! (Look at this for a reference on how to enhance the post:
Have you ever had to create multiple chats in order to tweak and find the prompt that yields the best results for GPT-4? Try the Notebook feature in Copilot and watch your prompt-tweaking problems vanish away!
Personalized teaching is the key to faster learning, but considering the student-teacher ratio in most schools, this is currently not feasible. Using AI to automate repetitive, yet complex tasks can free teachers' time for a personalized experience for our students!
United States Trends
- 1. Jeremiyah Love N/A
- 2. Massie 97.8K posts
- 3. #Varanasi 248K posts
- 4. Zvada N/A
- 5. #CollegeGameDay 2,337 posts
- 6. Charlie Becker N/A
- 7. Todd Snider N/A
- 8. Aaron Donald 2,733 posts
- 9. #MeAndTheeSeriesEP1 1.46M posts
- 10. Pat McAfee 1,289 posts
- 11. Lawson Luckie N/A
- 12. #Caturday 5,255 posts
- 13. #SaturdayVibes 5,655 posts
- 14. Brooklynn 4,433 posts
- 15. Willie Green 5,226 posts
- 16. Good Saturday 37.1K posts
- 17. Wrigley Field 1,254 posts
- 18. Northwestern 3,482 posts
- 19. James Franklin 2,334 posts
- 20. Marjorie 123K posts
Something went wrong.
Something went wrong.