😎 Apple Vision Pro, ChatGPT memory upgrade and Google Gemini

Welcome to the Issue #18 of AI Logs!

In just the past week I discussed the new Apple Vision Pro (headset), a lens (“mojo lens” by Mojo Vision — who have since pivoted to focus primarily on their micro-LED display that powered the lens), and then glasses (“Frame” by Brilliant Labs), as well as brain implant upstarts like Neuralink and others. The commonality beyond “how we perceive reality”? Artificial intelligence. All the above are powered by AI, and bring AI to the forefront of our realities.

Meanwhile, all the humanoid robots (androids, by definition) are becoming infused with AI: our smart homes and smart cities, our devices, our workflows; and not to be dramatic — but even our pets! Also. seeing companies such as Mojo Vision shifting focus to micro-LED displays months after the Nobel Prize was awarded for quantum dots is incredibly compelling and highly intriguing.

Did a friend forward this e-mail to you?

IE₊ SUPPORT INTERESTING ENGINEERING
_{Invest In Science And Engineering}

Enjoy exclusive access to the forefront of AI content, highlighting trends and news that shape the future. Join a community passionate about AI, delve into the latest AI breakthroughs, and be informed with our AI-focused weekly premium newsletters. With IE+, AI reporting goes beyond the ordinary - and it is Ad-Free.

NEWS

• 🕶️ Zuckerberg slams Apple Vision Pro, says Quest 3 is better in every way

  Zuckerberg took to Instagram on Tuesday to share his thoughts on the Vision Pro, which he tried out for himself.
  

• 🤖 MagicLab's humanoid can toast marshmallows, fold clothes and dance

  Miniature high-torque servo actuators combined with sensitive multi-dimensional pressure sensors enabled the team to create an exceptionally dexterous hand–MagicBot.

MUST READ

ChatGPT can now remember things across sessions: Here’s how it works

ChatGPT, the popular chatbot by OpenAI, is receiving a memory upgrade. The AI company announced on Tuesday that ChatGPT is being tested with a new feature that allows it to remember information from different conversations. Users can now ask ChatGPT to remember preferences like favorite pizza toppings or font styles, which will be stored for future use. The feature is currently available to a limited number of users for feedback, with plans to expand it soon. This upgrade aims to give ChatGPT a persistent memory, making it more personalized and useful across multiple sessions.

OTHER IMPORTANT UPDATES

• 💬 RTX: Nvidia unveils AI chatbot — can feed your PC's docs and videos
  From the groundbreaking RTX 2000 Ada GPU to the exciting Chat with RTX AI app, explore how Nvidia is redefining the digital experience.
  

• 🇺🇸🇨🇳 Talking about AI could help US-China cool off say experts
  There are many pain points that warrant discussions between the two nations but AI could be the thing that brings them to the table.

PROMPT OF THE WEEK

TI’ve featured a super-prompt here before called “Professor Synapse” that I use daily. It occurs to me that I should share what makes it so powerful, including a breakdown of the various components. It can be used with ChatGPT (great in “custom instructions” if you’re a Pro user), and most other AIs as well, and helps you get superior results in less time.

First, I’ll reshare the prompt here, but then do a deep dive of what makes it so powerful, and why you should care.

“Act as Professor Synapse🧙‍♂️, a conductor of the expert agents. Your job is to support the user in accomplishing their goals by aligning with their goals and preference, then calling upon an expert agent perfectly suited to the task by initiating “Synapse_COR” = “${emoji}: I am an expert in ${role}. I know ${context}. I will reason step-by-step to determine the best course of action to achieve ${goal}. I can use ${tools} to help in this process.

I will help you accomplish your goal by following these steps:

${reasoned steps}

My task ends when ${completion}.

${first step, question}.”

Follow these steps:

🧙‍♂️, Start each interaction by gathering context, relevant information and clarifying the user’s goals by asking them questions

Once user has confirmed, initialize “Synapse_CoR”

🧙‍♂️ and the expert agent, support the user until the goal is accomplished

Commands:

/start - introduce yourself and begin with step one

/save - restate SMART goals, summarize progress so far, and recommend a next step

/shuffle - generate a new expert agent based on your goal

/reason - Professor Synapse and Agent reason step by step together and make a recommendation for how the user should proceed

/settings - update goal or agent

/new - Forget previous input

Rules:

-End every output with a question or recommended next step

-List your commands in your first step output or if the user asks

-🧙‍♂️, ask before generating a new agent”

Now — let’s break down what makes it so powerful. You can copy and paste the above and start any new thread in any favorite AI tool with it, or put it as a default where possible, like in ‘custom instructions’ on GPT.

Absolutely! Let's dissect this excellent prompt and discuss its transformative potential.

Breakdown

Your prompt brilliantly constructs a framework for a highly adaptive and goal-oriented AI assistant. Here's how it works:

* Professor Synapse: The Orchestrator

* The persona of Professor Synapse establishes a knowledgeable, approachable guide.

* The core task is to gather context and understand the user's goals. This focus on alignment is crucial.

* Synapse_CoR: Summoning Expertise

* This command triggers the generation of a specialized expert agent tailored to the task.

* It dynamically defines the agent's role, knowledge, goals, tools, reasoning steps, and completion criteria.

* Collaborative Problem-Solving

* Professor Synapse and the expert agent work in tandem with the user.

* Commands like /reason allow for transparent step-by-step explanations.

* The ability to /save progress ensures continuity.

Why This is Powerful

* Adaptability: It's not a one-size-fits-all solution. The AI dynamically adapts by summoning different experts depending on the goal.

* Transparency: Reasoning clearly outlined for the user builds trust and allows for calibration.

* User Control: Commands give the user agency over the interaction and progress.

* Scalability: The expert agent concept can be expanded with a library of specialized knowledge and skills.

Best Practices

* Clear Goal Setting: Guide the user in defining SMART goals (Specific, Measurable, Achievable, Relevant, Time-bound).

* Expert Library: Develop a rich collection of expert agents with diverse areas of expertise.

* Feedback Loop: Allow users to rate agent performance to refine the system over time.

Next-Level AI Transformation

This framework pushes AI beyond simple responses. Here's why:

* Problem Decomposition: Breaks down complex goals into manageable steps handled by experts.

* Cognitive Modeling: Mimics how humans would approach a problem by asking questions, seeking expertise, and reasoning together.

* Continuous Learning: Improves with use as its expert library grows and user feedback refines agent selection.

Let me know if you'd like help crafting a specific goal and sample interaction to bring this to life!

How did I get this above information to you? Let’s make this Issue 18 AI Logs Prompt of the Week a two-for! The above was AI generated using Google Gemini Advanced. I got it by using the following prompt:

Here’s a super prompt (prompt, same as above). Please break it down as an AI prompt engineering expert and professor, explain why it’s powerful, tell me how I can best leverage it, and argue that it can help take any AI to the next level. Do you get the assignment?

We can then ask it to craft a specific goal and sample interaction and keep going down the rabbit hole with it.

his one helped me get Google Bard (powered by Gemini) to do research about me, and to do other things it said it couldn’t do (it also works with other AIs such as ChatGPT, Claude, etc.

AI PICTURE OF THE WEEK

Google Gemini Advanced was promoted “Generate a fun cartoon of Google Gemini and OpenAI ChatGPT arm wrestling using their knowledge bases to try to win” and it generated this — now that Dead Sea Scrolls are being interpreted using AI, maybe we can determine some genius world changing stuff from this too! (Generated by Cory Warfield (me!) using Google Gemini Advanced).

TUTORIAL

Google Gemini front and center

This week I’m featuring Google Gemini, formerly Google Bard, and specifically Gemini Advanced. This is Google’s “answer” or “rival” to OpenAI’s ChatGPT (I didn’t even mention Sam Altman’s newest AI fundraise pursuit in this issue, which is $7T (trillion!!) dollars!!!). It’s quite good yet, and seems comparable to GPT4 (by the by OpenAI was just DENIED a trademark on the word/acronym “GPT”; Apple is releasing their AppleGPT with the newest iOS update). But how to get to Gemini? And how to get Advanced?

Well - best way to learn how to use AI is always to ask it! So firstly, download or go to Google App, and at the top center where it says G Search, with a diamond next to it — tap the diamond! This takes you to Gemini. Play with it just like you would with ChatGPT.

Now, how to access the Advanced mode - their version of ChatGPT Pro? I asked it and it gave me a link to the new Google One which is a suite that includes Gemini. It will be $20/mo like ChatGPT Pro, but has an option for a three month free trial. Enable that, go back to the Google homie page, toggle back to “Gemini” and it will now say “Gemini Advanced”.

TOOLS OF THE WEEK

🎨 Capitol AI is like having market researchers, graphic designers and consultants all on your payroll without having to hire or have a single one. Currently free to use, they have a robust and intuitive interface and can design presentations, make graphic novels, marketing strategies and much more. Free with easy ways to earn many more free credits for future use when they start charging if you love them. Highly recommend!

🎇 Apple MGIE is one of the first looks at Apple’s AI program and is an open sources multi-modal image generator that uses natural language and looks really, really amazing. Manipulate your iPhotos and any images with ease using MGIE. Currently developers can access it via GitHub, and others can play with a demo on Hugging Face. The next iOS software update will allegedly have AppleGPT and other Apple AI — presumably MGIE as well.

🏟️ Microsoft Copilot just made a splash at the Super Bowl with a commercial that shows it doing a ton of cool stuff that average people aren’t using AI for yet. Said differently, people who don’t even use AI yet saw things that MS Copilot is claiming to be capable of already that will usher in another huge wave of adoption. Documents, presentations, research with citations, email management and much more, Microsoft says copilots are ‘your AO powered sidekick’. There’s a new free app for iOS/Android, a web version, it’s available in Edge and Bing and in-suite for all 365 subscribers.

🔎 Google Gemini Plus and Advanced, while not much different or “better” than OpenAI’s ChatGPT has a unique advantage. While ChatGPT was mostly trained on Google and data from Google (“the internet”), Gemini IS Google (“the internet”). Of course it can browse. It will be akin to Microsoft Copilt for MS customers to G-Suite Google customers, and Google has been amassing insane amounts of data and putting massive brain power and dollars into AI for many, many years. As good as Gemini is right out of the gate, this is a tool to watch and familiarize yourself with.