Welcome, humans.
Almost exactly 1 year ago, an AI-generated song mimicking Drake and The Weeknd went viral. Now, it's Drake who's in hot water for using AI to deepfake the vocals of the late Tupac on a diss track aimed at Kendrick Lamar (he yanked the track after a lawsuit threat).
Here's an app idea we definitely hope no one builds: a tool that whips up personalized diss tracks for your enemies...
Here’s what you need to know about AI today:
- ChatGPT can now remember facts about you in new convos.
- Writing emails is the fastest-growing use case for AI.
- 61% of GenZ report using AI tools in place of search engines.
- Claude now has a “Stop Generating” button.
On the podcast: Apple x OpenAI, Meta’s business model with AI, and The Verge’s state of AI survey (Apple Podcasts, Spotify, YouTube).
ChatGPT can now keep track of info from one chat to the next.
see how Memory works
Back in the day, each time you booted up ChatGPT, you’d need to rehash everything about you and your goals (like, "I write a newsletter about AI yada yada.").
We’re waiting for GPT-5 to address all of this. In the meantime, OpenAI has cooked up a few solutions.
- Initially, it rolled out Custom Instructions, allowing users to customize the ways ChatGPT responds.
- Now, it’s launching Memory for Plus users in the US, designed to let ChatGPT remember facts about you across conversations.
For example, we've instructed ChatGPT to always offer up 5 synonyms whenever we request a synonym:
Here’s our how-to on using Memory (first published here):
1. Go to Settings → Personalization → Memory → Manage.
2. Add any details you want ChatGPT to recall, like:
- Past jobs, current skills, and future goals.
- Key elements typically included in your emails.
- Your preferred tone, voice, and style of writing.
- Jargon specific to your profession.
- Tools frequently utilized in project management or software.
3. When mid-convo, just say, “Remember this”
4. For hotkeys and shortcuts, do the following:
Why it matters: Memory slightly outperforms Custom Instructions, but we view it as a mostly stopgap measure.
It gets us one step closer to personalized AI assistants that truly remember everything, and I mean everything, about us: what projects we’re working on, preferred software, our brand voice, and more.
FROM OUR PARTNERS
Help your company become Enterprise Ready in the AI era.
What do AI unicorns Perplexity, Copy.ai, and Jasper all have in common?
They all use WorkOS to land more enterprise deals.
Large organizations must ensure the safety of their sensitive information when using AI platforms.
WorkOS helps high-growth AI companies become Enterprise Ready by offering:
- Easy-to-integrate auth and SSO.
- Self-serve SSO onboarding for customer IT teams.
- Free for up to 1M monthly active users.
Discover how WorkOS can prepare your company for enterprise success here.
A lot of people are turning to AI for a lot of basic things.
The Verge just published the results from its survey on how American adults use and think about AI.
Among the major takeaways is that Millennials and GenZ are AI’s power users:
- 61% of GenZ and 53% of Millennials use AI tools in place of traditional search engines like Google for researching topics.
- 53% of Millennials and 49% of GenZ are more drawn to products when it’s labeled as “powered by AI”.
In related news: An Axios survey of its Communications readers shows that people are more likely to have a favorable view of colleagues who incorporate AI into their work.
Here’s more from the fast lane:
- Writing emails is the fastest-growing use case for AI.
- 55% of consumers trust the accuracy of AI-generated information.
- 55% are moderately worried about the impact of AI tools on the 2024 Presidential election.
Hear Pete’s extended take on The Verge’s survey results on today’s episode of The Neuron podcast (Apple Podcasts, Spotify, YouTube).
Around the Horn.
- Speculation is rife that a new chatbot, termed “gpt2-chatbot,” could be OpenAI's next big release, possibly GPT-4.5 or GPT-5.
- The Financial Times snagged a licensing deal with OpenAI that allows ChatGPT to spit out summaries, quotes, and links to FT content. The licensing deals roll on!
- Claude finally has a “Stop Generating” button.
- Gradient released a version of Llama 3 8B that supports up to 1M+ tokens, potentially handling up to 750K words of text.
- Bill Gates is reportedly still very hands-on at Microsoft, specifically with its AI happenings.
Treats To Try.
- *Sana AI is your ultimate AI assistant for work—it helps analyze documents, summarize meetings, automate repetitive manual tasks, and more. Work faster AND smarter—try Sana AI for free today!
- GitHub Copilot Workspace is a new workspace environment that helps devs brainstorm, plan, build, test, and run code.
- Truva is like an expert AI chatbot on your website that actually helps guide users to the right tools, features, and info within your app.
- Play AI is building conversational AI assistants that are trained to talk like your brand.
- EdwarMax’s AI diagram generator creates diagrams from text descriptions.
*This is sponsored content. Advertise in The Neuron here.
Tuesday Ticker.
Here are the results from last week’s poll:
- A.G. chose Very worried: “There are so many people that already believe fake articles, can you imagine what it will be like when they "see" someone endorsing whatever it is they want to believe in?!”
- G.S. chose Very worried: “Very difficult to legislate against, will become ever more convincing. How on earth will the average person (me) know what is real?”
- P. chose Not worried: “It's a perception thing. The more fakes out there, the more we realize that we can't trust what we see and hear.”
Would you read news that's been summarized by an AI chatbot like ChatGPT/Perplexity?
Login or Subscribe to participate in polls.
A Cat's Commentary.
