Beyond chat | Issue 11

Welcome to Issue 11 of unfolding ai. It’s a bit of a strange week for AI news, lots of ‘news’, but not lots of ‘available’. Most of the big Tech players have announced new features ‘available soon’ or ‘early next year’.

It’s easy to get carried away around this excitement, but remember at these announcements the examples are very ‘curated’. Let’s wait and see how they perform in the real world.

Best regards,

Paul, Co-founder (and newsletter editor)

AI as a wearable

We have become used to wearable technology, the apple watch is the most popular watch on the planet. Behind its sensors and data gathering tools is a wide array of machine learning (apple marketing term for AI) detecting everything from health to falls (personal experience is this works!).

Meta has partnered with Ray-Ban to enter into the ‘smart sunglasses’ arena, whether this will make a larger indent than ‘google glass’, ‘snapchat glasses’ is yet to be seen. A lot of this still for me revolves around 3rd party privacy, ie people in the field of view that don’t really want to be ‘googled’.

More interesting is the idea that computers (and AI) can become much more discrete, this is the angle hu.ma.ne are pursuing.

Of-course where all this leads to is, how will it be monetised for the mass market. What will be the advertising strategy? How will the content fit into the experience.

Imagine asking ‘what restaurants do you recommend near here?’, how will you get into that response?

Moving beyond text, multi-modal

2024 (as a prediction) will see the move towards adding voice / text / image and video as both input and output into general purpose AI, especially on the in phone application versions of the platforms.

This is a key enabler to make ‘a personal AI assistant’. Many lifeMin (life-admin) tasks could be solved if the input was a picture, think of calendars, menus, opening times, to-do lists. Assuming that the Ai assistant can have a wide range of integrations to then perform tasks on them, this is the consumer product.

Bill Gates described it as who-ever solves this, will win the AI race for consumer adoption. I also suspect the AI will be ‘behind the scenes’, which is exactly how Apple are moving in this area.

If you have not tried just using chatGPT as a transcription tool, you really should. It’s already excellent. Transcription tools for meetings, including context of the assets being shared has significantly advanced in the last 12 months. This is the most advanced multi-modal use todate.

You should be asking yourself why you are taking meeting notes anymore.

We use otter.ai, although it’s likely this will be absorbed into teams/zoom/google meet. So check what you have access to already.

Productivity, stepping into automation

The current estimates around productivity gains for completing tasks assisted with AI such as chatGPT are between 17% and 40%. If you take a moment to think about the time taken to complete a task with chatGPT, how much of that time is spent ‘prompt engineering’?

If like me you now have 1000’s of chatGPT sessions, it’s pretty hard to find that prompt you wrote a month ago, to re-use. That’s why we advise creating a personal playbook (copy paste) of repeatable prompts.

The next big area of progress in AI and productivity will be in ‘process automation’ or chat Agents. These are combinations of low-code (think of them as flow charts that run code), and calls to chatGPT and other applications.

This is a really simple one (not using any AI) from my zapier setup. Zapier is even available as a chatGPT4 plugin, though I actually find that harder to use than zapier.com native interface.

The difference between an automation (like zapier) and an agent (sorry none really available yet to show you) is that the agent, or more likely combination of agents, runs behind the scenes, or ‘is dispatched’ to do tasks. A bit like a personal-ai-concierge.

Zapier is very approachable to learn, and has 1000’s of preset templates and integrations before you have to get into the complexity of AI and coding. You might be surprised how any activities in your day to day productivity, just a little automation can help with.

We will be exploring automation in more depth over the coming months, mainly in premium content

Quick things to check out

A fun google tool

Ideal to keep the kids (or yourself) amused in the coffee shop!

Also from Nvidia

A prompt to try in chatGPT (and other LLMs)

Forgotten a birthday? Yep, we have all done it. As usual our standard format where anything in [..] can be edited. I wonder if I should automate this to my calendar…

write me a short, friendly, personal email to [persons name], apologising for missing [their birthday], offer to [catch up for a coffee] [next week]

Help us grow, every referral helps

Subscribe to Premium to read the rest.