Subscribe | Sponsorship

Welcome!

Hey there! Have you noticed that AI news isn't spreading as quickly as it used to? This is a big plus in my view as there was simply too much hype.

Some have brought to my attention the publicised 10% drop in monthly use of ChatGPT, and a study that suggested GPT-4 was getting ‘dumber’.

While I have nothing to back this up, I suspect the drop in monthly users is more likely linked to a number of people who weren’t really using the tool for anything productive and were simply ‘testing its limitations’.

I also personally haven’t noticed any decrease in performance or quality of output, but keen to hear from those who may have.

Also, I’ve decided that rather than bombard your inbox with noise each week, I will be publishing when there is enough news to share.

• 👨‍🏫 ChatGPT Custom Instructions

• 👨‍💻 Tool of the Week: ChatGPT Code Interpreter

• 🔍 Bard Updates

• 🚦 No Traffic: An AI Startup That Makes Traffic Lights Smarter

• 📖 AI Insights from Marc Andreessen

• 🌐 From Around the Web: Learning Resources

Hope you enjoy! Any feedback, please hit reply and let me know your thoughts.

👨‍🏫 ChatGPT Custom Instructions

ChatGPT has recently introduced a new feature called Custom Instructions.

This is available only to Plus subscribers.

This feature allows users to customise how ChatGPT responds to them and what it should know about them to provide better responses.

Click on the three dots to the right side of your name:

Select “Settings & Beta”:

Then select “Beta features” on the left, and then switch on the toggle for Custom instructions on the right.

You’ll then be able to select “Custom instructions” from the menu:

When adding your instructions, tooltips will appear to the right side, giving you an idea of what to add:

Once the Custom Instructions are set, ChatGPT will consider them for every conversation going forward.

This means that you don't have to repeat your preferences or information in every chat, and ChatGPT will remember them across sessions.

Great, because I really got annoyed having to start every conversation explaining that I am a planner from New Zealand.

Another enhancement to save time and effort, and also make the AI chat experience more personalised and engaging.

👨‍💻 Tool of the Week: ChatGPT Code Interpreter

Data is fundamental to how we undertake planning in the urban context. Whether it is analysing demographic trends, mapping spatial patterns, visualising transportation networks, or creating scenarios for future development, data helps us make informed and evidence-based decisions.

However, working with data can be challenging, especially when it comes to processing, manipulating, and presenting large and complex datasets. That's where ChatGPT Code Interpreter comes in.

ChatGPT Code Interpreter is a new feature of ChatGPT, the popular generative AI chatbot that can converse with humans on various topics. Code Interpreter allows ChatGPT to access and run Python code within a chat session, with additional options to upload and download files. This means that you can use ChatGPT as a data analyst, a programmer, a tutor, or a creative partner.

For example, you can ask ChatGPT to:

• Perform data analysis and visualization on your own datasets or public data sources.

• Write programs to create unique graphics, animations, or documents.

• Edit videos, images, documents, and more.

• Convert files from one format to another.

• Learn how to code in Python or other languages.

• Generate code snippets or suggestions based on your input or query.

ChatGPT Code Interpreter is part of ChatGPT's paid tier, which costs US$20 per month.

You can access it by enabling the Code Interpreter plugin in your ChatGPT settings, noting the tool is still in Beta.

Once enabled, select ChatGPT 4, then Code Interpreter, and off you go.

As a planner, I have found this tool to be a huge step up from the normal ChatGPT 4. It has helped me with tasks that used to take hours every month, such as:

• Automating data collection and cleaning from various sources.

• Generating summaries and insights from complex data sets, including identifying the types of questions to ask.

• Reviewing different documents containing data to identify missing information or anomalies.

• Pulling accounting reports together into a specific spreadsheet format and doing so instantly without an error (this task alone has saved an hour of work per month).

If you are interested in learning more about ChatGPT Code Interpreter, you can check out the following resources:

• ChatGPT plugins - OpenAI

• What is ChatGPT code interpreter? The chatbot can now edit videos and make GIFs

• What is ChatGPT Code Interpreter and how to use it

• What is the ChatGPT Code Interpreter Plugin? Ultimate Guide

🔍 Bard Updates

Google has rolled out several new features and expansions for its conversational AI chatbot Bard.

Bard is now available in over 40 languages beyond English, including Arabic, Chinese, German, Hindi, and Spanish. Its worldwide expansion comes after proactive engagement from Google with experts, policymakers, and privacy regulators.

New capabilities allow users to listen to Bard's spoken responses, adjust the tone and length of replies, pin conversations for later reference, and share responses. Bard can also now export Python code to Replit in addition to Google Colab.

Productivity features include the ability to pin, rename, and revisit conversations in a sidebar, as well as share responses via links. Users can also now include images in prompts, with Bard analysing photos to enhance responses.

These updates aim to provide more customized, shareable responses and aid productivity. They follow Bard's launch in February as a competitor to ChatGPT.

My own experience testing Bard is that it is not yet as capable as OpenAI's ChatGPT, but the gap may close with ongoing improvements.

Users worldwide can now try out Bard for themselves across languages and compare the two AI chatbots. I encourage you to have a play with Bard now that it is available globally and draw your own conclusions by openly comparing it to ChatGPT.

🚦 No Traffic: An AI Startup That Makes Traffic Lights Smarter

Traffic is one of the biggest problems that urban dwellers face every day. It causes stress, pollution, and accidents, and wastes time and money. But what if there was a way to make traffic flow more smoothly and safely, without building new roads?

Most planners know you can’t build your way out of vehicle congestion.

That's the vision of No Traffic, an Israeli startup that has developed an AI-based platform that turns traffic lights into smart devices that can communicate with each other and with vehicles on the road. No Traffic claims that its platform can reduce congestion by up to 40%, cut CO₂ emissions by up to 30%, and prevent up to 25% of accidents.

How does it work? No Traffic installs smart sensors on existing traffic lights that can detect the presence and movement of cars, bikes, buses, pedestrians, and other road users. The sensors also have V2X chips that can exchange data with connected and autonomous vehicles. Using edge computing, the sensors can process the data in real time and adjust the traffic light timing and sequence to optimise the traffic flow at each intersection. The data is also sent to the cloud, where it is synchronised with other intersections and analysed for insights and improvements.

No Traffic's platform also has safety features that can prevent collisions. For example, if a sensor detects that a car is about to run a red light, or a pedestrian is about to cross the street, it can alert an approaching connected car to slow down or stop. No Traffic says that its platform can also integrate with emergency services and public transportation systems to give them priority and reduce response times.

No Traffic is not the only company that is trying to improve traffic management with AI, but it claims to have some advantages over its competitors. For one, it says that its platform is compatible with any type of traffic light, regardless of age or manufacturer. It also says that its platform is scalable and flexible, allowing cities to customise it according to their needs and preferences. And it says that its platform is cost-effective and easy to install, requiring minimal changes to the existing infrastructure.

What do you reckon, Auckland Transport? 🤤 

No Traffic has been growing rapidly since it was founded in 2017. It has raised $67.5 million in funding so far, including a $50 million Series B round in June 2023. It has also expanded its presence in North America, where it has signed contracts with over 100 departments of transportation in 13 states. It plans to use its latest funding to hire more talent, develop new features, and enter new markets.

If you are interested in learning more about No Traffic and how it can transform your city's mobility, you can check out the following resources:

• NoTraffic raises $50M to tackle congestion and enhance road safety

• Traffic management platform NoTraffic raises $17.5M in Series A to double team size

🌐 From Around the Web: Learning Resources

Below is a list of learning resources I have been reading. I recommend pursuing one at a time and don’t try to tackle them all at once.

1. Introduction to AI for Business Users

2. Learn GPT Best Practices

3. Prompt Engineering Guide

4. Career Essentials in Generative AI