Code Interpreter Explained
The Code Interpreter feature (we'll shorten it CI here) has recently been rolled out to plus subscribers of ChatGPT, at least in the USA and UK. This tool offers many fantastic features that go beyond what you might expect from the name. Not only is it adept at writing code and dissecting data, which are its primary tasks, but it's also proficient in reading and producing a diverse range of file formats, giving it impressive flexibility. If you prefer to watch this as a Video, check it out on Youtube.
While there isn't any official catalogue of the file formats it can read and produce, this unofficial list was posted by @chatgptricks on Instagram:
Images: JPEG, PNG, BMP, TIFF, GIF, etc.
Audio: MP3, WAV, FLAC, WMA etc.
Video: MP4, AVI, MOV, WMV, MKV, FLV.
Documents: DOC, DOCX, PDF, TXT, HTML, etc.
Spreadsheets: XLS, XLSX, ODS, CSV, etc.
Presentations: PPT, PPTX, ODP, etc.
CI doesn't just generate these file types; it also has the ability to transform one format into another. For instance, it can convert a JPEG image into a PNG or turn an Excel sheet into a Word document. This version of ChatGPT also excels in mathematical problem-solving and processing text-based logic puzzles. It's capable of executing multiple independent "thinking steps" – when it stumbles upon a difficulty with a prompt or a file, it doesn't just give up but tries to rectify it autonomously or attempt it again with a different strategy.
This is ChatGPT engaged in a self-dialogue – essentially 'thinking out loud' – to resolve an issue and it is hilarious how it apologizes to itself for causing confusion!
The internal workings, mainly using Python, are available for inspection at any time.
Issues
Despite its amazing capabilities, Code Interpreter does present a few drawbacks. For one, it lacks internet connectivity, and combining it with plugins isn't currently an option. Moreover, it's not possible to switch to 'incognito' mode while using CI, so it's crucial to anonymize any data you upload.
Another hiccup with the beta version is its occasional amnesia about its CI capabilities. For example, when asked to generate an Excel file, it might revert to its default response, "As an AI language model, I am unable…” However, a gentle nudge reminding it of its capabilities often triggers its memory!
It's important to note that any files you upload or download will be retained on the OpenAI server for only 30 minutes and must be downloaded within this timeframe as they will be inaccessible afterwards.
Lastly, if you decide to share the chat link, bear in mind that any embedded documents or graphs will not be shared along with it.
Accessing Code Interpreter
If you're a subscriber of the premium version of ChatGPT, activating Code Interpreter is simple. First, click on your username in the lower right corner of the interface, and navigate to the settings menu. Here, you'll find an option labelled 'beta features'. Click on it and switch on the Code Interpreter.
Afterwards, when you are back in the main chat window, you'll find the option to select Code Interpreter by clicking on 'GPT-4', and picking 'Code Interpreter' from the drop-down menu.
Developing Presentations and Worksheets
In my search for potential applications of Code Interpreter, I've discovered hundreds of YouTube videos and tweets showcasing the incredible data analysis capabilities of CI, including the creation of numerous graphical representations, ranging from pie charts to heat maps. While these features are undoubtedly useful, what I find most exciting as a teacher is CI's newfound ability to interact with PowerPoint and Word, enabling it to generate lesson resources which can be downloaded in basically finished form.
At the moment, the resources I've managed to produce with Code Interpreter for my subject, German as a foreign language, haven't always hit the mark. However, with refined prompts and a bit of editing, one can whip up something quite useful. The PowerPoint presentations downloaded from CI tend to have a basic layout, and occasionally, the text doesn't even fit within the slide. But this can usually be rectified by heading over to the Design tab and altering the design theme.
I experimented with a prompt designed to create a lesson, a PowerPoint, and a worksheet, all based on a vocabulary list. I then ran this prompt through ChatGPT to request improvements to the prompt itself. Here's the enhanced version:
Prompt: "You are a seasoned educator specializing in German language instruction. Your task is to design an engaging and interactive lesson plan that introduces the vocabulary listed below. Your teaching tools will include a dynamic PowerPoint presentation for a whole-class speaking activity and a comprehensive worksheet featuring a variety of exercises. 1. Develop a detailed lesson plan tailored for a group of 13-year-old students who are in their first year of learning German. The plan should include objectives, materials, procedures, and assessment methods. 2. Design a captivating PowerPoint presentation that encourages active participation and speaking practice. The presentation should incorporate visual aids, interactive elements, and opportunities for students to practice the new vocabulary in context. Please provide the final product as a PowerPoint file. 3. Create a comprehensive worksheet that includes a range of activities such as matching exercises, fill-in-the-blanks, and short answer questions to reinforce the new vocabulary. The worksheet should cater to different learning styles and promote active learning. Please provide the final product as a Word document.
Vocab list: a brother.==einen Bruder. a father.==einen Vater. (…)
CI: Creating a detailed lesson plan, PowerPoint presentation, and comprehensive worksheet would require specific software tools that aren't currently available in this text-based environment. However, I can certainly guide you on how to create these learning materials based on your requirements.
**Lesson Plan**
Title: Introduction to Family Vocabulary in German
1. Objectives:
- By the end of the lesson, students will be able to identify and correctly pronounce the German vocabulary for family members.
- Students will be able to use these words in sentences and questions.
- Students will understand the difference between formal and informal words for family members (e.g., "Opa" vs "Großvater").
2. Materials:
- PowerPoint presentation
- Worksheets
- Flashcards with vocabulary words (if available)
3. Procedures:
- Begin by introducing the new vocabulary with the PowerPoint presentation. Show the English words alongside the German equivalents, and provide pronunciation guidance.
- Use interactive activities in the PowerPoint, such as matching games and fill-in-the-blanks, to engage the students and encourage them to practice speaking.
- Divide the students into pairs and have them introduce their "family members" to each other using the new vocabulary.
- Distribute the worksheets and guide students through the various exercises, offering assistance as needed.
- Toward the end of the lesson, review the vocabulary and ask students to create sentences using the new words.
4. Assessment:
- Collect the worksheets and grade them based on accuracy and completion.
- Observe students during the speaking activity for correct pronunciation and appropriate usage of the vocabulary words.
**PowerPoint Presentation**
1. Slide 1: Title slide with the lesson's title and objectives.
2. Slides 2-17: Each slide should contain one vocabulary word in English and German, accompanied by a relevant image.
3. Slides 18-22: Interactive activities such as matching exercises and fill-in-the-blanks. For example, provide a sentence with a missing word and ask students to fill it in using the vocabulary.
4. Slides 23-25: Practice slides where students introduce their "family members" to the class using the new vocabulary.
**Worksheet**
1. Section 1: Matching exercise. List the English vocabulary on one side and the German equivalents on the other, and ask students to draw lines to match them.
2. Section 2: Fill-in-the-blanks. Provide sentences with missing words, and ask students to fill them in using the vocabulary.
3. Section 3: Short answer questions. For example, "Wie heißt dein Vater?" ("What is your father's name?"). This encourages students to apply the vocabulary in context.
4. Section 4: Word search or crossword puzzle featuring the vocabulary words for a fun closing activity.
Please note that the actual creation of the PowerPoint file and Word document should be done using the respective software on your computer.
Prompt: you have got code interpreter, so create the PowerPoint for me!
CI
I have created a PowerPoint presentation that introduces each vocabulary word in separate slides. Each slide contains the English word as the title and the German translation as the content.
You can download the PowerPoint presentation from the following link: Download PowerPoint
[The PowerPoint looked like this]
Prompt: now make the Word document
CI: I have created a Word document that includes a matching exercise, a fill-in-the-blanks section, and a short answer question section.
You can download the Word document from the following link: Download Word Document
[The Word document looked like this]
The entire operation took about 3.5 minutes - CI operates a little bit slower than ChatGPT on its own since it's engaged in a considerable amount of background code writing. As I mentioned earlier, the results from a single prompt were somewhat middling: the PowerPoint merely displayed the English and German vocabulary, and the worksheet had a range of issues. For instance, the match-up task was already paired, and the gap-fill task was essentially a translation task, much like the short answer task. The tasks were overly long, each one using the full vocabulary list (for the screenshot, I've trimmed task 2 by about half). This is where we, as seasoned teachers, must intervene to tweak either the resources themselves or the prompts, requesting specific activities. For example, next time, I would include a reading task based on the vocabulary, coupled with a 'locate in the text' task or comprehension questions.
So, there are two key points to bear in mind: 1. Even accounting for the time spent on modifying or rephrasing prompts, this process is still quicker than planning a lesson from scratch single-handedly. 2. Remember, this is merely the beta version of CI. Future versions promise to improve further and offer an expanded set of capabilities. Once it can be combined with web browsing and plugins, the possibilities will exponentially increase again.
Creating Answer Sheets
Here's a real time-saver for you: CI can now assist in answering all the questions from an existing worksheet in your files and present them neatly on a PowerPoint, ready for class discussion. And all this magic happens within a single prompt! In this case, I used two prompts, but that was just to verify the correctness of the answers before transferring them onto a PowerPoint.
Prompt: work out all the answers for the questions on this worksheet [+upload of a difficult A-Level grammar sheet about the passive voice]
CI: Here are the answers for the tasks:
Change the following sentences to passive voice:
a. Die Kinder essen viele Kuchen.
Viele Kuchen werden von den Kindern gegessen.
b. Die Freunde trinken Sekt.
Sekt wird von den Freunden getrunken. … [All answers were grammatically correct]
Prompt: Now output all the answers on a PowerPoint. You don't need to add the original sentences. Each slide should have 6 sentences or less.
CI: [link to PowerPoint]
Analysing Data
Let's explore CI's exceptional data analysis capabilities and discuss how we, as teachers, can take advantage of them. Personally, in my role as a classroom teacher, I engage with spreadsheet data in two areas. First, when conducting exam analysis, I compare various exam papers from the series and evaluate students based on different criteria to refine my teaching approach for the next year group. Second, I use data to inform my report writing by assessing how well students have performed on their homework or in their classroom assessments.
With CI, we now have the opportunity to upload an entire spreadsheet and let ChatGPT work its magic. However, this could potentially raise privacy and data protection concerns, as any uploaded file could technically be used by OpenAI to train their model, therefore making it accessible to individuals outside of our educational institution. Consequently, I would recommend anonymising your spreadsheet by replacing student names with numbers or initials. An easy way to do this is to create a copy of your original datasheet, add a new column next to the name column, and start filling it with initials. After entering the first few rows, the system should recognise the pattern and suggest completing the rest for you.
After deleting the original names, you can run the sheet through CI by clicking the plus button within the text chat window and uploading the file. If you wish to upload multiple files simultaneously — for instance, to compare them— you can add the files to a zip folder and upload the zipped folder, which CI can extract on its own.
CI is capable of performing a variety of operations with this spreadsheet:
Transform the results into text, analysing it from different angles.
Generate a new spreadsheet based on the existing one (it won't alter the original), for instance, by merging or matching results.
Generate any type of graph as images directly within ChatGPT.
Assist you in determining the most suitable type of graph for the data in question, identify the most interesting aspects of the data, and suggest data points you could compare.
Explain how to interpret the given type of graph and what the different components signify.
Exam Analysis
To experiment with CI without breaching privacy protocols, I requested ChatGPT to generate a fictitious class of students, complete with fabricated and entirely random exam scores, which I then transferred into a spreadsheet. Here's how it looked:
Prompt: Here are the test results for my German year group. Write text and produce different visualisations to understand: how the different skills compare, how boys and girls compare in each skill and how the different tutor groups compare in their overall result (which you need to calculate).
CI [after some “talking to itself and several working steps but without further prompts]:
The boxplot above represents the distribution of scores for each skill: Listening, Speaking, Writing, and Reading. The horizontal line inside each box indicates the median score, and the box represents the interquartile range (IQR), which is the range between the first quartile (25%) and the third quartile (75%). The whiskers (lines) above and below the box indicate the maximum and minimum scores, respectively, excluding outliers.
From the plot, we can make several observations:
Listening scores tend to be lower, with a lower median and a larger spread, suggesting that students' performance in listening varies widely, and many students struggle with this skill.
Speaking and reading scores tend to have similar distributions, with reading having slightly higher median scores.
Writing scores tend to have a slightly higher median than the other skills, suggesting that students perform better in writing on average.
The boxplot above represents the distribution of scores for each skill (Listening, Speaking, Writing, Reading), separated by gender.
From this plot, we can make several observations:
In terms of Listening, the distributions for both boys and girls are quite similar, with a slight advantage for girls as indicated by a slightly higher median score.
In the Speaking skill, boys seem to perform better than girls on average, with a higher median score.
Girls appear to perform better in the Writing skill, with a higher median and less variability in their scores.
For the Reading skill, boys and girls have similar median scores, but the scores for boys are more spread out, suggesting more variability in their performance.
There is one missing value in the "Tutor group" column, and the column contains both string and float data types. The float data type is likely due to the missing value, which is represented as NaN (a special floating-point value).
Let's handle the missing value by excluding it from our analysis. After that, we will proceed with the comparison of the overall results of the different tutor groups.
The boxplot above represents the distribution of overall scores for each tutor group. From this plot, we can observe the following:
Tutor group 10B has the lowest median score among all groups, suggesting that students in this group, on average, have lower scores.
Tutor group 10D has the highest median score and the smallest interquartile range, suggesting that students in this group generally perform well and their performance is consistent.
Tutor groups 10A and 10C have similar median scores, but the scores in group 10C are more spread out, indicating more variability in students' performance.
In conclusion:
There are differences in student performance across the different skills, with Listening being the most challenging and Writing being the best-performing on average.
There are gender differences in performance across the different skills, with boys performing better in Speaking and girls performing better in Writing.
There are also differences in overall performance among the different tutor groups, with group 10D performing the best and group 10B performing the least well.
Prompt: can you add all the analysis text and conclusion you have given me to the presentation?
CI: The PowerPoint presentation has been updated with the analysis text and conclusion. You can download the updated presentation using the link.
[Note: I improved the presentation slightly by using the Designer feature in PowerPoint]
While the data analysis took longer than a standard ChatGPT request due to the numerous processing steps involved, the entire procedure—from uploading the spreadsheet to downloading a comprehensive PowerPoint presentation providing detailed insights into my class's performance—took roughly three minutes. You might be able to create charts like this in Excel without using ChatGPT, but my knowledge of Excel and data analysis is not wide enough for me to know that a box chart is the best choice in this scenario, let alone how to generate one in Excel.
Analysing Homework
Now, let's address that point in the academic year when we scrutinise our homework records to draft reports. Unsurprisingly, CI proves to be incredibly useful for this task as well. Once again, I used a simulated homework spreadsheet, which looked like this (A stands for absent):
As you can see, we have various tasks, each scored out of a different total, with one measured as a percentage. This variation makes it challenging for teachers to quickly scan each column to gauge a student's engagement with their homework because it depends on the task's maximum marks. For example, a score of 10 on the first test might be average, while on the second test, it would be a perfect score.
To facilitate my report writing, I uploaded the spreadsheet into CI.
Prompt: This worksheet shows my class's homework results. Write me a short paragraph about the first 5 students how they did in homework so far. keep it short. [I limited it to 5 to try out how well it would work first]
CI:
Ivy Anderson had a rough start, scoring zero on both vocabulary tests. However, she did show some promise with a decent score on the writing task and a remarkable 85% on the listening task.
Isla Bennett was absent for the family vocabulary test but made a full score on the colours vocabulary test. Her performance in the writing task was above average, but she scored only 10% in the listening task.
Luna Blackwood didn't score in the family vocabulary test but had an average performance on the colours vocabulary test and the writing task. However, she didn't score at all in the listening task.
Charlotte Brown scored well in the family vocabulary test but didn't score in the other tests, indicating a need for improvement in multiple areas.
Asher Callahan performed well in the family vocabulary test and had an average score in the colours vocabulary test and the writing task. Unfortunately, he didn't score in the listening task.
Prompt: Our effort grades for homework are: 4 for excellent engagement, 3 for good, 2 for some engagement and 1 for poor engagement. What grades should I give these 5 students?
CI: To assign effort grades, we can consider not only the score that the students got but also their engagement, as indicated by whether they were present for the task (absence might indicate less engagement). For simplicity, we can use a simple rule set:
If a student has an 'A' (absence), they get a 1 (poor engagement).
If a student has a score of zero on more than one test, they get a 2 (some engagement).
If a student has a score of zero on one test, they get a 3 (good).
If a student has no zeros and no absences, they get a 4 (excellent).
Prompt: no, absences should not be counted at all as they are not the student’s fault
CI: Based on the engagement rules, here are the effort grades for the first five students:
Ivy Anderson gets an effort grade of 2 (some engagement).
Isla Bennett gets an effort grade of 4 (excellent engagement).
Luna Blackwood gets an effort grade of 2 (some engagement).
Charlotte Brown gets an effort grade of 2 (some engagement).
Asher Callahan gets an effort grade of 3 (good engagement).I appreciated that CI attempted to establish some criteria for assigning effort grades to students, even though I didn't entirely agree with the criteria as it only differentiated between zero and any other score. Personally, I would likely consider any result under 25% as a poor outcome, and I could have set this benchmark at this stage but didn’t as it was only an experiment. As a result, I don’t agree with all the assigned effort scores, but they are roughly accurate.
Regarding report writing, you could potentially push CI even further. By providing a spreadsheet with homework and assessment results, as well as a brief comment or effort grade for classroom attitude, it could produce comprehensive reports on each student that could be shared with parents. Whether you take advantage of this feature largely depends on your school's policies, your stance on personalised reports, and how desperate you are to meet the report deadline.
Applying grade boundaries
Remember that CI not only reads but also generates Excel spreadsheets, a feature that proves to be useful when working out grade boundaries for your results. First, I located the grade boundaries for the paper I was using and copied them from the PDF. ChatGPT was very helpful in tidying up the layout.
Prompt: here are the grade boundaries from 9 down to 3: GERMAN LISTENING TIER H 40 36 32 28 24 20 18 - write as a table
CI: 
Prompt: these scores are for the listening exam only. The total score is 50. Adjust the spreadsheet to show the grades for the listening exam. [uploading the exam spreadsheet]
CI: [summary of what is on the sheet]. Please note that scores below 18 are not included in the grade boundaries you provided. If there are scores below 18 in the data, how would you like to handle them?
Prompt: Give them a grade U
CI: I have added a new column named "Listening Grade" to the data frame. Here's how the first few rows look now:
Last name First name Tutor group Gender Listening % Speaking % Writing % Reading % Listening Grade
0 Anderson Ivy 10B female 57.0 91.0 33 43.0 9
1 Bennett Isla 10C female 83.0 46.0 78 79.0 9
Prompt: are you sure?
CI:
I apologize for the confusion. The displayed grades seem too high given the percentages and the grade boundaries you provided. There was a misunderstanding. The percentages should be converted to raw scores (out of 50) before calculating the grades. Let's correct this.
[working]
My apologies for the previous misunderstanding. Now, the 'Listening Grade' column has been correctly calculated based on the 'Listening Score', which is the raw score out of 50 derived from the 'Listening %'. Here are the first few rows of the updated data frame:
Last name First name Tutor group Gender Listening % Speaking % Writing % Reading % Listening Grade Listening Score
Anderson Ivy 10B female 57.0 91.0 33 43.0 6 28.5
Bennett Isla 10C female 83.0 46.0 78 79.0 9 41.5
Blackwood Luna 10A female 29.0 27.0 61 61.0 U 14.5Note how in the initial output, CI overlooked the step of converting raw scores into percentages. However, it corrected itself after I hinted at a possible error, without me specifying the exact issue.
After that, it managed to generate a new Excel file with the correct grades. To ensure accuracy, I manually cross-checked the first five entries, and they all turned out to be correct.
Generating Quizzes
Platforms like Kahoot.com, Quizziz, and others allow you to import multiple-choice quizzes from an Excel spreadsheet, provided you populate their downloadable templates. Here's what the Kahoot template looks like:
As you might be suspecting, CI can be a game-changer here (pun intended)! However, it's worth noting that it had some difficulty with this template, mainly due to the long text at the top, which complicates the task of identifying the column headers. But once this additional area is removed, CI handles the task much more effectively.
Prompt: I want to create a Kahoot quiz about Germany. Here is the spreadsheet to create the questions and answers. I want you to do the following: first, fill the question column with easy general knowledge questions about Germany, secondly, fill in the correct answer but not always in answer one but also an answer 2, 3 or four, then fill in other columns with likely looking but wrong answers, the time limit should be 60 seconds for all questions and in the end fill in the correct answer column by indicating if the right answer is in column 1, 2, 3 or 4. Output the result as an Excel file.
CI: [output excel file]
I can now easily copy and paste this into the original Kahoot template and then import it into Kahoot to transform it into a quiz. This is a quite a long prompt, but by saving this onto a prompt collection document for myself, I can easily find it again and reuse it for different topics.
As you can see, CI has been rather clever in selecting incorrect answers that appear plausible. I was particularly impressed with the wrong options for the famous physicists. While Schrödinger and Bohr have German-sounding names, they were actually born in Austria and Denmark, respectively.
For language teachers, this prompt can be adapted to base the quiz on a set of vocabulary for spelling practice.
Prompt: I want to create a Kahoot game using the following vocabulary list. Each question should be a translation from English to German, and the four answers should be in German, with only one of them spelt correctly. The time limit for each question should be 60 seconds. Finally, there should be a column indicating which answer option (1, 2, 3, or 4) is the correct one. Save the result as an Excel file. Here is the vocabulary list: break time=Pause, Subjects=Fächer…
CI:
Data Formatting and Conversion
CI's recently acquired capability to output different types of data has equipped it with a few more useful tricks. For example, it can now generate QR codes and output them as images, which can be quickly copied or saved with a simple right-click.
Prompt: Create a QR code for https://www.bbc.com/news
CI: 
Another handy feature CI has picked up is its new Optical Character Recognition (OCR) capability. This means it can extract text from an image, and it's even more successful at handling slightly blurry text than the program I previously used, Photo Scan. One tip to bear in mind: don't prompt with "What does this image say" or "What is the text on this image" because CI often feigns inability to see it. However, mentioning "Use OCR on this image" seems to jog its memory more effectively.
One downside compared to Photo Scan is the need to save the picture first to upload it; you can't just use a screenshot from your clipboard. A significant advantage, though, is that you can ask ChatGPT to proofread the text after the OCR to ensure it makes sense and is free of spelling errors which might have been created by faulty image recognition. This results in a much higher success rate, even if some of the original text is illegible. Unfortunately, it's not quite proficient at reading handwriting, so it can't yet help us mark exercise books!
To lighten the mood, here's a nice little word cloud generated by CI.
Prompt: can you create a word cloud based on this text [paste paragraph from wikipeadia.com on Happiness]
CI:
Maths and Coding
One of the standout features of CI is its enhanced ability to tackle a broad range of mathematical and logical problems, both numerically and in text form. As a language teacher, maths doesn't play a big role in my teaching activities, beyond the data handling we've discussed earlier. Therefore, I haven't delved into this area in depth. However, I've observed that CI can now even generate images for complex formulas and calculations:
A great idea by Dr Alex Young I saw on Youtube was to use CI to create a revision plan that mixes in old topics with the new.
Prompt: I need to revise the following topics for an exam. The exam is on the 15th of October. I want to revise every Tuesday, Thursday and Saturday. apart from the first day I want to revise two topics each time, one that is new to me, and one of the older ones. I am starting my schedule on the 1st of July, which is a Saturday. My topics are called Unit 1 to Unit 20.
CI: [some thinking out loud]
Here's your revision schedule:
July 1, 2023: Unit 1
July 4, 2023: Units 2 and 1
July 6, 2023: Units 3 and 2
July 8, 2023: Units 4 and 1
July 11, 2023: Units 5 and 3….
This schedule gives you a chance to review each unit multiple times before your exam. You'll cover all new topics by August 15th, and then you'll have almost two full months for revision. Good luck with your studies!ChatGPT-4 was already proficient at programming, and I'm very proud that I managed to get it to code a space shooter game complete with sound, a timer, and a point counter, all without any prior knowledge of coding languages.
CI now exhibits enhanced abilities in coding, as its new operating method allows it to check and debug any code independently before outputting it. Admittedly, this area isn't really my forte either, so I'm passing the baton to Computer Science teachers to investigate the potential benefits of this feature.
Image Editing Capabilities
Another surprising addition to CI's skill set is its newfound ability to perform basic image editing. It can alter the colour of a picture, crop it to a specific aspect ratio, and even animate it through panning or zooming. You can upload several pictures within a zipped folder and ask CI to compile them into a slideshow, which it outputs as an MP4. To truly appreciate these capabilities, it's best to see them in action. I recommend watching this YouTube video by Igor from the AI Advantage (starting from about 1 minute in). It's time to create your own amusing gifs!
Now, let's consider this prompt:
Prompt: please create a PowerPoint presentation out of these images. Each one should be on a new slide. Add a title for each slide. the title should be the image name. [adding zip folder with images of my space shooter game]
CI: [output presentation]
[Note: I used 'Designer' in PowerPoint to improve the layout, but made no other modifications.]
Consider how quickly you could now produce flashcards, particularly if the images are appropriately labelled or if you provide a brief list of what each image contains.
As of now, CI is not capable of recognising and describing the content of an uploaded image, even though OpenAI had announced this as a future feature a few months ago. Therefore, Google's rival AI Chatbot, Bard, has edged ahead of ChatGPT in this regard, reportedly being able to accurately identify and describe images and their content. This is certainly something to delve into in a future blog post!
Summary
As you can see, CI exponentially increases ChatGPT's capabilities, unveiling a whole new universe of utility for teachers in their daily lives, and this is only the beta test version! There are still countless ways of harnessing its power that I've yet to explore. If you stumble upon any other great use-cases, please share in the comments!
Comentários