14th issue: The first law for AI is coming ⚖️

🌯 Wrap for the week

A close look at all the main headlines in AI

Just like GDPR, the EU is leading the race to regulate AI.

The EU does like to be the first. Kudos to them for doing it again.

This week, The European Parliament has just put its stamp of approval on a major step towards regulating Artificial Intelligence across the EU. Although this is the draft version of the law the parliament members have agreed to. It is still a big step nonetheless. Here are a couple of key takeaways from the news:

1. A Clear Stance on AI Rights and Values: The Parliament's members have expanded the proposed rulebook, reinforcing that any AI developed and used in Europe must align with EU rights and values. This includes areas like human oversight, safety, privacy, transparency, non-discrimination, and social and environmental well-being. The future of AI in Europe is looking to be not just smart, but also very ethical.

2. Stricter rules, more bans: Parliamentarians have voted for an amended version of the AI Act, which includes bans on remote biometric surveillance, predictive policing, and untargeted scraping of facial images from the internet or CCTV footage to create facial recognition databases. Essentially, tech like Clearview AI is in the crosshairs. They've also banned the use of emotional recognition tech by law enforcement, border agencies, workplaces, and educational institutions.

3. High-risk AI and obligations on makers: The definition of high-risk AI systems has been expanded to include those that could harm health, safety, fundamental rights, the environment, or influence elections. Big social media platforms that use algorithms to recommend content have also been marked as high-risk. The makers of general-purpose AI (like OpenAI) will need to identify and mitigate risks before hitting the market and also ensure to publish summaries of copyrighted information used to train the model. Well, I guess OpenAI is going to have a hard to comply with such requirements.

4. Innovation isn’t what they want to reduce: In the law that was backed, the Parliament's members have added exemptions for activities in research and AI components provided via open-source licenses. So, I guess they still want to see progress made in AI.

Whether you support this law or don’t, don't pop the champagne just yet, as there's still much negotiation and compromise to be done between the parliament and EU Member States. There's even a chance for a total breakdown in the law-making process. However, the EU seems motivated to get ahead of the curve on this, given the global attention on regulating AI.

Will they iron out the creases and come up with the perfect AI regulation playbook? Only time will tell.

Main news credit:
Europe takes another big step toward agreeing an AI rulebook from TechCrunch

OpenAI API gets a new power-up!

Personally, I just love to see new things you can do with the OpenAI API.

The most intriguing thing from this week's announcement must be the Function calling. In short, it is providing a better and more reliable way for people to connect GPT’s capacity with external tools and APIs.

It allows developers to give verbal descriptions of programming functions via the OpenAI API. When the user is making requests, the AI can then understand the request, pick the most relevant function, and extract the relevant information from that request to be used for the function. Basically, the AI model would be able to understand the question and pick the right tool to solve the problem with the information provided.

Sounds a bit vague? Here are some use cases for Function calling shared by OpenAI:

1. Create chatbots that answer questions by calling external tools: e.g. checking real-time updates about the weather, or sending out an email to the person you mentioned in the queries

2. Convert natural language into API calls or database queries: e.g. show the top 10 customers from this month via generating a SQL query to find out the answer in your database

3. Extracting Structured Data from Text: e.g. extract all people mentioned in the Wikipedia article

To top it off, OpenAI has also

• updated and more steerable versions of gpt-4 and gpt-3.5-turbo

• introduced a new 16k context version of gpt-3.5-turbo, i.e. extending the memory of the AI

• reduced the cost for the embedding model and the input token for gpt-3.5-turbo

With these improvements, it seems like OpenAI is opening the door further to a world where we can talk to our machines, have them understand us and do things for us. AI might soon be just another coworker in the office. We'll just have to make sure to invite it to the next office party!

Main news credit:
OpenAI intros new generative text features while reducing pricing from TechCrunch

Other top AI headlines for the week

Sir Paul McCartney says artificial intelligence has enabled a 'final' Beatles song from BBC

I am not a huge Beatles fan but I hope this would be a piece of good news for their true fans.

Just a quick note, the key element that AI plays here is the extraction and purification of the voice of John from old recordings. So if you are concerned if AI helped to somehow wrote this final song, it is not. The creative brilliance behind this song is still from the band members.

Google Lens can now search for skin conditions from TechCrunch

Another piece of news about how AI could make people’s life better, I just love it! Although this is not a replacement for a proper medical diagnosis, it is a useful tool to help you decide if any further medical check is needed.

Mechanical Turk workers are using AI to automate being human from TechCrunch

I think this image from TechCrunch summarised it nicely.

From its humble beginning to do the things that computers can’t (e.g. solve a CAPTCHA or identify the sentiment of a sentence) to about 1/3 of the Mechanical turkers using LLMs to support their tasks, AI has truly advanced a lot and changed a lot of human behaviours.

Talking about people loving to adopt AI, I must point out that developers also share some of the love. According to TechRadar, of over 90,000 developers surveyed, more than two-thirds collectively either using AI tools (44%) or planning to soon (26%). Also, it is interesting to point out that only 3% highly trust AI tools. So I guess, trust is not the sole factor for AI adoption in this case.

🎬️ Guess that movie

Adding a bit of fun to your weekend reading, let’s guess the famous movie from the 10-emoji description created by ChatGPT:

👨‍⚖️🏛️1️⃣2️⃣🧑‍🤝‍🧑💬⚖️😡🔄✅

If it is a bit hard to figure out, you can find the answer with some explanation from ChatGPT at the end of this newsletter.

🧰 Toolbox

Check out this week’s tools and resources that give you more superpowers with prompt crafting!

1. 📚️ Exploring opportunities in the generative AI value chain
If you are exploring ideas in this AI era and got a bit stuck, this article might help you think about the possible areas you can explore further.

I am personally quite keen about the application area, hence just copying the table and share with you all via the newsletter as well.

2. 🔧 MusicGen
Another new model from Meta! And you can try via Hugging Face for free. Just a note, the generated music would be just 12 seconds.

By the way, the wait time to get your music generation could be lengthy and you can add some audio on top of the text prompt you have. I guess that is kind of like prompting Midjourney with an image.

Check out this piece of music that I got from simply this prompt👇️

a fast-pace OST music for a side-scrolling video game created in early 2000

3. 📺️ Generative AI learning path
If you are looking for video content to learn more about Generative AI, here is one from Google.

👋 Your next Twitter friend

Connect with great people in the industry to supercharge your growth

🖼️ Image of the week

If you have been around Twitter this week, the QR code has been all the craze. Following the trend, I have got my hands on trying to create some for myself as well. If you are curious, just try opening this newsletter on one device and scan these QR codes with your phone to see if it would work. The link would direct you to the homepage of The Prompt Craftsman newsletter.

Personally, I love the QR codes where the black bars of the QR code blend in well with the style of the image.

If you are interested in generating some QR codes like this, feel free to drop me a message at [email protected] or hit me up on my Twitter @AnInquisitiveMd.

🎬️ Guess that movie - answer

The answer is "12 Angry Men" 🥳 

Did you guess that right? Here is ChatGPT’s explanation:

👨‍⚖️🏛️1️⃣2️⃣🧑‍🤝‍🧑💬⚖️😡🔄✅

This summary captures the main elements of the story: a trial (👨‍⚖️), a courthouse (🏛️), 12 men (1️⃣2️⃣), discussions between them (🧑‍🤝‍🧑), debates and arguments (💬), the balance of justice (⚖️), anger (😡), change of mind (🔄), and finally a verdict or agreement (✅).