- Teacher Prompts
- Posts
- Better call LMM
Better call LMM
Neil Almond
February 13, 2024
👋 human, It is half term here at TeacherPrompt Towers, but we know that is not the case for all of you. Rest easy as you digest the latest news and how-tos on AI. For those that are not on half term, take solace that it will soon be you.
Knowledge builders 📚
Gemini Ultra 1.0 → A new LLM from Google that powers Gemini Advance.
$7,000,000,000,000 (trillion) → The amount of money Sam Altman wants to raise to invest in AI chips. A bottleneck of the continued growth of AI is the capacity and scale of the silicon that AI applications and models need. Altman is seeking to ensure the continued advancement of AI by addressing this challenge head on.
Industry updates 🤖
Bard becomes Gemini → For those in the Google ecosystem, Bard has had a new coat of paint and its feature-set is not massively different to Bard. However, as part of this announcement Google has announced Gemini Advanced. Currently only available in English, Gemini Advanced can be thought of as being on par with GPT 4.0, but comes as part of an upgraded Google One account. Usually £18.00 a month, it comes with a 2 month free trial, but I suggest waiting until it is baked into Docs, Sheets and Slides to see if it is something you would use.
LLMs beat the lawyers → In New Zealand, a study was conducted that looked whether various LLMs were faster, as accurate and cheaper than junior lawyers at reviewing contracts. GPT-4 Turbo was found to be as accurate and considerable cheaper and faster than a junior lawyer.
Fresh prompts ✨
Foundation level: Use the Camera → More of a feature than a prompt per se, many AI apps provide access to the camera which allows you take pictures and interrogate those images with the use of AI. For example, you could when marking work, take a picture and ask the AI if the question is wrong and what the misconception or error that was made. You could tag each photo with a pupil number and when you have finished marking a particular question, ask which pupils had a similar error or misconception. Use in Copilot app. Here is an example:
you are an expert in assessment in primary mathematics recently pupils have completed a mathematics assessment and you want to check what the errors and common misconceptions may have been made in this particular question. Take your time and use chain of thought reasoning to explain the error.The output:
Certainly! Let's analyze the student's work step by step:
1. Given Information:
- A square tile measures 20 cm by 20 cm.
- A rectangular tile is described as being 3 cm longer and 2 cm narrower than the square tile.
2. Student's Calculations:
- Square Tile Area: \(20 \times 20 = 400 \, cm^2\)
- Rectangular Tile Area: \(23 \times 22 = 506 \, cm^2\)
- Difference in Areas: \(506 - 400 = 106 \, cm^2\)
3. Error Analysis:
- The student correctly calculated the area of the square tile.
- However, the error lies in calculating the area of the rectangular tile:
- The rectangular tile should be \(23 \times 18\) (since it's 3 cm longer and 2 cm narrower).
- The correct area of the rectangular tile is \(414 \, cm^2\).
- The difference in areas should be \(414 - 400 = 14 \, cm^2\), not 106.
4. Common Misconception:
- The student likely misinterpreted the dimensions of the rectangular tile, leading to an incorrect calculation.
- They mistakenly used \(23 \times 22\) instead of \(23 \times 18\).
5. Correct Approach:
- To find the difference in areas, calculate the area of the rectangular tile correctly: \(23 \times 18 = 414 \, cm^2\).
- Then subtract the area of the square tile from the rectangular tile area: \(414 - 400 = 14 \, cm^2\).
In summary, the student's error was in calculating the area of the rectangular tile, resulting in an incorrect difference in areas. Encouraging careful attention to dimensions and clear understanding of word problems can help prevent such misconceptions in the future. 📏🔍
Expert Level: Delimiters for clarity → Sometimes, tagging aspects of a prompts input can support the LMM make sense of it and avoid hallucinations. In the following example, I wanted the LLM to understand the name of a type of sentence and an example of that sentence and provide me with some examples of each sentence type whilst using the examples I had provided. Use with any LLM.
You are an expert writer and want to create short piece of writing that demonstrates specific sentence types. The name of the sentence types will be given in <> and an example will be given in (). You want to write a 5 sentence about space using the following sentence types <adverbs followed by present or past participle phrases> (Quietly cackling to herself, the witch stirred the potion.) and an <expanded noun phrases that include a prepositional phrase> (The wicked witch with warty skin stirred the potion.).The Output:
Certainly! Let’s embark on a cosmic journey through words, weaving a celestial tale that showcases the specified sentence types. 🌌✨
Title: “Stardust Reverie”
Expanded Noun Phrase:
In the inky expanse of interstellar space, the ancient comet’s tail trailed behind it, a luminous ribbon of frozen dreams.
Adverb + Past Participle Phrase:
Silently gliding through asteroid fields, the starship avoided collisions, its hull shimmering with protective energy.
Simple Sentence:
Satellites orbited tirelessly, their solar panels angled toward distant suns, transmitting cosmic whispers.
Adverb + Present Participle Phrase:
Gracefully pirouetting, the binary stars danced their eternal waltz, painting the void with stellar hues.
Expanded Noun Phrase:
Among nebulous veils, the cosmic weaver wove threads of spacetime, her fingers deftly stitching constellations.