Breaking news to lead us off…
Claude and I are NOT friends.
You have heard me talk a lot about ‘my friend Claude’ over the last year. I thought we had something. But guess what? He doesn’t feel the same way.
So here’s what happened. If you missed it, this ChatGPT prompt has gone viral:
“From our conversations, what do you know about me that I don't know?”
Of course, like most of you, I received the most flattering response. Aw shucks, ChatGPT. You shouldn’t have. But keep going.
I had the idea to ask Claude the same thing. Seeing as how I spend more time with that LLM I was curious how similar the answer would be. Well, Claude was having none of it:
“I apologize, but I don't actually have any information about you or our previous interactions. Each conversation I have starts fresh, and I don't retain information from past chats or build up knowledge about specific users over time. I'm an AI assistant created by Anthropic to be helpful, harmless, and honest, but I don't have personal experiences or memories to draw from. Is there perhaps a different way I could assist you today?”
I was shocked. Claude, why are you embarrassing me in front of everyone? (I was doing this during our AI Council meeting.) Surely you don’t mean that. Then this happened.
Licking my wounds, I turned back to ChatGPT to ask it the same thing
I’ve had a day, guys. It’s like finding out there’s no Santa Claus.
1. 13,000. That’s how many words are in this essay from Dario Amodei. Dario is the CEO of Anthropic, and I will be writing a sternly worded letter to them regarding Claude being ‘fake’ (as the kids say).
Anyway, the essay is titled Machines of Loving Grace, based on a poem titled All Watched Over By Machines Of Loving Grace (Richard Brautigan).
The essay is truly 13,000 words. I’m telling you upfront—I have not read it. I listened to Paul and Mike break it down on the most recent Artificial Intelligence Show. I also read Casey Newton’s summary and opinion on it.
I did put it into Google NotebookLM and made a podcast, which I will listen to later today. I’ll share the link next week if it’s worth a listen.
The reason I’m sharing it is so you can choose how deep you want to get into it. As Paul points out in this week’s pod … are we really ready for this? No, of course we’re not. And it nearly broke my brain just reading about the key takeaways and overview. If you check it out in any form, let me know what you think.
2. Meta’s AI chief is also renouncing dooomerism.
Yann LeCun says the thought of AI endangering us is ‘complete B.S.’
Those are two really smart people who think our future is bright. So, let’s just go with that for a while!
3. Seven nuclear reactors to power Google’s AI ambitions.
So this is just my take – nothing from the linked article. But I keep hearing/reading about how we’re closer to AGI/superintelligence than the means to control or regulate it. Or it’s not the technology that’s holding us back – it’s the means to power it that it is. And as I heard at this year’s MAICON, there is no plan for once we achieve AGI. It’s just a race to get there.
There are so many things that still need to be resolved or considered, like the environment. But now, with a move like this, we aren’t just speeding up…we’re doing so without any foresight of what the environmental impact could be with more computing, usage and technology.
These are big issues. I wish I could offer any type of rationale thought or idea around what could be done or why we are progressing at this rate. But I truly think there is no stopping this trAIn.
4. Since ChatGPT’s launch, the U.S. defense and homeland security departments have paid out $700M for AI projects.
That seems like a lot, no?
I wonder if AI spending will be a topic at any future presidential debates or town halls.
5. ChatGPT has launched a Windows desktop app.
Wow. Game changer? Well, the latest game changer that is. As this article points out, is OpenAI going to make a play for ‘the new office suite’? You’d have to think they are. And you have to think this is going to accelerate the introduction and adoption of AI agents.
Christopher Penn tackled intermediate prompting in his most recent newsletter. He dug into examples of how you can use both automated and manual chain of thought prompting approaches. I’d highly recommend reading through it and seeing how he approaches it as well as his link to his company’s prompting framework document.
…Adobe’s AI video model is already here? It’s inside Premiere Pro. (The Verge)
Just like a Tootsie Pop, it’s inside!
…Ambient AI will see you now. What this new doctor’s assistant is, and why it’s listening to you? (CNBC)
…AI helped the feds catch $1B in fraud this past year. (CNN) Catch me if you can, part 2?
…Some of the most prominent AI models are falling short of the EU’s regulations. (Reuters) Phew—thank goodness we don’t have any regulations here!
…Google NotebookLM received an upgrade, so you can now edit, direct and control the ‘podcast’? (TechCrunch)
Swarm, swarm, swarm!
OpenAI has introduced Swarm. It’s an experimental framework aimed at helping developers build and coordinate networks of AI agents. In other words, agents that will be able to automate and do the work of departments.
(Here’s Shelly Palmer’s take: “AI Swarm” sounds scary because it is.)
Yes, this is the agentic move that many feared would lead to mass layoffs of white-collar jobs … or the emergence/refocus of existing jobs. No, it’s not a product yet, but they have basically just released “the cookbook.”
This is why I have harped on choosing to live in opportunity—not fear. I understand how these agents will work and am the one who conducts the orchestra. Remember Steve Jobs' famous quote: “The musicians play their instruments. I play the orchestra.”
Well, it’s never been more relevant. Pick up the baton, folks–-here we go!
-Ben
As a reminder, this is a round-up of the biggest stories, often hitting multiple newsletters I receive/review. The sources are many … which I’m happy to read on your behalf. Let me know if there’s one you’d like me to track or have questions about a topic you’re not seeing here.