Memory for ChatGPT widely available

Howdy, wizards. Here’s what’s brewing in AI this week:

1. Memory for ChatGPT becomes more widely available. Not yet in Europe or Korea, though.
2. Reid Hoffman interviews his AI twin. Building a virtual twin like this could be a solid use case for the workplace.
3. Apple has resumed talks with OpenAI. The next iOS update could bring everyday use of AI into the mainstream.
4. US Government starts AI Safety Board with leading tech execs. It’s a tough task to balance the opportunities and threats in AI race.
5. Elon Musk is close to raising $6 billion for xAI. A massive bet from investors on xAI’s success.
6. GPTs: top arrivals in the GPT store and on whatplugin.ai.
7. Bytes: the other important AI news this week.

‍

Dario’s Picks

1. Memory for ChatGPT becomes more widely available

‍

‍

ChatGPT’s memory feature is now available to users on paid ChatGPT plans. Exceptions: Europe and Korea.

The feature allows ChatGPT to remember important information about you across chats. It does so automatically by default, and you can also specify particular things you want it to remember. You can turn on or off memory in Settings → Personalisation → Manage.

Also, if you’re a bit confused between custom instructions and memory, this Reddit guide compares and explains the difference.

‍

Why it matters

More personalisation in AI is inevitable – it should theoretically make the chatbot more useful to you over time, and this would easily translate into more loyal users (so there’s a user & business benefit to create such features).

PS I’ve looked at the stuff it has remembered about me in the last couple of weeks since I got the feature. Some memories look useful for future interactions (e.g. “has INFJ personality type”) while other things look random or very case-specific (e.g. “wants to left-align the table headers.”). Not noticing much of reference back to this information yet, but it’s still early.

‍

2. Reid Hoffman interviews his AI twin

‍

‍

LinkedIn cofounder Reid Hoffmann created a digital AI twin of himself, and did a back-and-forth interview with it. The AI twin was trained on Reid’s many books, speeches and podcasts to be able to mimic him as closely as possible.

The facial expressions, gestures, lip-sync, etc. are really good. It was built by Hour One for the video and Eleven Labs for the voice. And using it to interview oneself really illustrates just how weird (and kinda cool) things are about to get.

‍

Why it matters

Just look at how AI Reid easily summarises his books for different target audiences – very cool. As memory and recall gets even better in future AI models, I think Hour One is on to a solid use case with building an AI workforce twin like this (they also just raised $20m.) Imagine the productivity boost of having a twin that can deliver content and messages on behalf of you, in whatever language.

‍

3. Apple has resumed talks with OpenAI

Apple is reportedly again in serious talks with OpenAI to implement Gen AI features for the iPhone with the new iOS 18 update. Apple was recently in talks with Google on similar matters, after which the first negotiations with OpenAI fell through. However, the discussions have renewed according to a Bloomberg report.

Rumours also have it that the main AI features of iOS 18 will be based on an in-house model, while OpenAI and Google tech would be used for a chatbot or search feature.

‍

Why it matters

With its massive distribution of 1.4B users, the next iOS update could potentially catapult the everyday use of AI into the mainstream.

‍

4. US Government starts AI Safety Board with leading tech execs

An AI Safety Board has been created with the mission to protect the economy, public health and vital industries from being harmed by AI-powered threats. OpenAI’s Sam Altman, Nvidia’s Jensen Huang and Google’s Sundar Pichai all have seats on the board. Notably, Elon Musk and Mark Zuckerberg are excluded. Other members include 20+ academics, civil rights leaders, and government officials.

Together with the Department of Homeland Security, they’ll develop recommendations on operation and production regulations within the industry, and how to address vulnerabilities and bulletproof their AI systems against disruptions.

‍

Why it matters

With this initiative, the Biden-administration is working to fill the regulatory vacuum that currently exists in AI while trying not to stifle innovation too much. Undeniably a tough task to balance the opportunities and threats in this crazy innovation climate we’re in right now.

‍

5. Elon Musk is close to raising $6 billion for xAI

Elon Musk’s company xAI is currently finalising a $6 billion investment round at a $18 billion pre-money valuation, expected to close in a few weeks unless the deal terms changes. Among others, Sequoia Capital, who are also backing OpenAI, are participating in the round.

Context: xAI is a rival to OpenAI that has the Grok foundation model, which is in early access for premium subscribers on X. Grok-1.5 is still unreleased but is reportedly a big upgrade from Grok-1 which initially fell quite flat.

‍

Why it matters

This is a massive bet from investors on xAI’s success. A unique selling point for xAI is the potential access to unique training data through Musk’s empire of companies ie X, Tesla, SpaceX, Boring Company and Neuralink.

I think the expectations could come into fruition, but one thing that would concern me as an investor is X’s declining user base since Musk took over, as the social media platform would likely be the widest distribution platform for anything xAI whips up.

‍

GPTs

Productivity potion

Something new or inspiring to get your week going

‍

‍

Top new arrivals in the GPT store

Highly rated, new GPTs featured in OpenAI’s official GPT store (from the last week)

• Essay Writer #14 in  writing
• ChatCody - GitHub, GitLab Assistant #15 in  programming
• Legal Documents & Law Expert #15 in  lifestyle
• Therapist (mental - no medical advice) #16 in  lifestyle

‍

Top newcomers on whatplugin.ai

Top rated GPTs that made it into whatplugin.ai from the last week. How rankings work.

‍

‍

 Bytes 

More notable AI news from the last week

• Adobe Introduces Firefly Image 3.
• OpenAI: Introducing more enterprise-grade features for API customers.
• Perplexity launches Enterprise Pro.
• First NVIDIA H200 GPU hand-delivered to OpenAI by Jensen Huang.
• Musk escalates OpenAI fight with subpoena of ex-board member.

‍