Hello Reader,
Promptcraft is a weekly curated newsletter on AI for education designed to elevate your AI literacy.
In the 50th issue 🎉, you’ll discover:
- “The king is dead”—Claude 3 surpasses GPT-4
- Now you can use ChatGPT without an account
- UK and US sign landmark AI Safety agreement
Let’s get started!
~ Tom Barrett
PERFORMANCE
.: “The king is dead”—Claude 3 surpasses GPT-4 on Chatbot Arena for the first time
Summary ➜ Anthropic’s Claude 3 Opus model became the first AI to surpass OpenAI’s GPT-4 on the Chatbot Arena leaderboard since its launch in May 2023. This marks a notable achievement for Claude 3, suggesting it may have capabilities comparable or superior to GPT-4 in certain areas like natural language understanding. The user-based ranking system of Chatbot Arena reflects their actual use on day-to-day tasks. The leaderboard aims to capture subtle dimensions of chatbot quality missed by numerical benchmarks.
Why this matters for education ➜ As I mentioned in issue #47 when the new Anthropic models were released, the benchmarks used for marketing are always a little misleading. Actual use by people integrating these models into real tasks might tell a different story. And that story so far, is Claude-3 Opus is better than GPT-4.
While GPT-4 remains a strong contender, especially with a major update expected soon, Claude 3’s rise underscores the increased competition in the AI big model space. Anthropic has major backing from Amazon’s investment, and their model for guardrailing is very interesting.
Constitutional AI (CAI) is an Anthropic-developed method for aligning general purpose language models to abide by high-level normative principles written into a constitution.
I hope this news encourages more educators to become curious about these other big tech and research companies driving AI innovation.
There is more than just Google, Microsoft and OpenAI.
|
ACCESS
.: Now you can use ChatGPT without an account
Summary ➜ OpenAI has removed the requirement for an account to use its popular AI chatbot ChatGPT. This change opens access to anyone curious about ChatGPT’s capabilities, rather than just registered users. Overall this represents a notable shift in how OpenAI is positioning ChatGPT as an AI for the masses versus a restricted product.
Why this matters for education ➜ The removal of login requirements by OpenAI expands access to AI tools like ChatGPT, making them more widely available to users, including communities that were previously excluded due to limited access to technology or inability to provide stable account credentials. While this increased accessibility is a positive step towards democratising AI, it also raises concerns about the potential risks associated with improper use, particularly if users lack sufficient understanding of the tool’s limitations.
|
|
AI SAFETY
.: UK and US sign landmark agreement
Summary ➜ An agreement to collaborate on guidelines for the development of artificial intelligence. The principles aim to foster safe, ethical, and responsible AI that respects human rights. Key areas of focus include AI explaining its decisions, minimising bias, ensuring human oversight, and not being used for harmful purposes like mass surveillance.
Why this matters for education ➜ This agreement builds upon commitments made at the AI Safety Summit held in Bletchley Park in November last year. It is essentially a partnership between the US and UK AI safety institutes to accelerate their research and progress. For education, we might see clearer ideas about how to build teacher AI Literacy or pathways for implementing student chatbots in classrooms. Guidelines for responsible AI implementation ensure that all students, regardless of background or socioeconomic status, access safe and ethical AI tools in their learning environments.
|
.: Other News In Brief
🔍 Anthropic researchers wear down AI ethics with repeated questions
🚀 Microsoft upgrades Copilot for 365 with GPT-4 Turbo
⚠️ AI Companies Running Out of Training Data After Burning Through Entire Internet
🗣 OpenAI’s voice cloning AI model only needs a 15-second sample to work
🤝 US, Japan to call for deeper cooperation in AI, semiconductors, Asahi says
🇮🇱 Israel quietly rolled out a mass facial recognition program in the Gaza Strip
📚 How do I cite generative AI in MLA style?
🤖 Now there’s an AI gas station with robot fry cooks
🎵 Billie Eilish, Pearl Jam, 200 artists say AI poses existential threat to their livelihoods
💡 Gen Z workers say they get better career advice from ChatGPT
|
:. .:
What’s on my mind?
.: Positive Augmentation
The video below was shared with me during a webinar for our AI for educator community, humain.
It features students at Crickhowell High School in Wales using an AI voice tool to augment their language skills.
It was published by the British Council. Here’s the description:
Our latest Language Trends Wales survey reveals a declining interest in language learning at the GCSE level in Wales. Amidst all the talk about Artificial Intelligence disrupting the language learning scene, can we instead leverage it to inspire students to learn a language? We conducted an experiment with students at Crickhowell High School in Wales. Watch what happened.
Although not referenced, I am fairly sure the AI tool HeyGen was used to translate and augment the speakers. I could be wrong, as there are so many of these tools now.
Last week, I shared that HeyGen was set to close a USD$60 million funding round, valuing the company at half a billion USD. The valuation demonstrates the growing interest and potential in AI-powered language media tools like HeyGen.
The technology is very impressive, and you can try it for free. Here is one of my Design Thinking course videos, translated into Spanish.
|
What do you think? The changes are almost imperceptible.
This augmentation tool is part of a family of image filters and style generators that have long been integral to social media tools.
The young people in the video, having grown up in an era where selfies and filters (augmentations) are commonplace, understand this technology better than most people.
If you listen back to the comments in the final part of the clip, as they reflect on what they have seen, you can sense a general sentiment that while these tools are impressive, they will never replace the need for authentic human communication.
It is interesting to reflect on how these new, powerful media tools portray us with new skills and capabilities.
I can watch myself speak Spanish, and although it feels like a trick, it is amazing not just to imagine yourself with a new skill but actually to see a synthetic version of yourself demonstrating that skill. This experience provides a tangible representation of the potential for personal growth and acquiring new abilities.
:. .:
~ Tom
|
Prompts
.: Refine your promptcraft
There was always something peculiar and subversive about the Fighting Fantasy books. I think I enjoyed the page-turning as much as the fantasy gameplay.
Have you had a chance to generate your own with a chatbot?
Fighting Fantasy books were single-player role-playing gamebooks created by Steve Jackson and Ian Livingstone.
They combined fantasy novels with role-playing games where readers played as heroes, made choices that determined the story’s outcome, and rolled dice in combat encounters.
Back in December 2022, it was one of the first prompts I was playing around with ChatGPT, and it was fun to generate your own game:
You decide to try to find a way out of the darkness and escape the danger. You search your surroundings, looking for any clues or hints that might help you navigate your way through the shadows.
Let’s try Claude-3-Opus – the most powerful model available – and see what we get. Here’s a prompt you can try, too.
And here is the opening of The Labyrinth of Lost Souls generated by Claude-3-Opus.
I am not sure if there is good mobile phone coverage in the labyrinth, but I will try to stay in touch.
And these locals look friendly…right?
Image generated with Midjourney 6.0 |
:. .:
Remember to make this your own, try different language models and evaluate the completions.
Do you have a great prompt you would like me to share in a future Promptcraft issue? Drop me a message by replying to this email.
|
Learning
.: Boost your AI Literacy
AFRICA
.: The State of AI Regulation in Africa – Trends and Developments [PDF]
There’s a varied approach to AI regulation across the continent, including the adoption of national strategies, policies, establishment of task forces, and adoption of AI ethics principles.
Africa faces unique challenges in regulating AI, such as infrastructural and governance hurdles, absence of specialised laws, and outdated legal frameworks.
The Tech Hive report suggests several opportunities to strengthen AI regulation, including global engagement, sector-specific regulation, leveraging existing laws, and promoting a multi-stakeholder approach.
Also of note is the impending Continental AI Strategy, which is expected to catalyse the development of more regulatory measures at the national level.
CHINA .: Generative AI in China
A helpful short article by Professor Mike Sharples reflecting on his experience visiting Shanghai. He briefly outlines how GenAI is being used in practice for business and education.
China has been developing AI for business, government and the military for many years, with notable success in data analysis and image recognition. But it lags behind the US in consumer AI, notably language models. One reason is a lack of good training data.
BASICS .: Non-Techie Guide to ChatGPT- Where Communication Skills Beat Computer Skills
In this video, I’m setting out to debunk the myth that ChatGPT is exclusively for those well-versed in technology or that it requires special training to use. I emphasise how anyone, especially educators, can use this tool effectively through the simple art of communication.
|
Ethics
.: Provocations for Balance
Do Language Filters Homogenise Expression?
If AI translation tools smooth over cultural differences and localised slang, does this promote harmful assimilation? What diversity is lost when all voices conform to a single standard? Should cultural preservation outweigh frictionless communication? Can both coexist in our increasingly global society?
Inspired by some of the topics this week and dialled up.
|
:. .:
.: :.
The more we invest in our understanding of AI, the more powerful and effective our education ecosystem becomes. Thanks for being part of our growing community!
|