Howdy, I hope you had a fantastic Memorial Day. Long weekends like these are great for reflection, and that reflection can lead to transformation, which is the theme of this week’s newsletter.

My daughter graduated high school this weekend and will soon start her college journey! If you looked in a dictionary for the words proud papa, I’m sure it has a picture of me cheesing. Congrats to Olivia and all of the classes of 2024!

I also have some news. After 8 years as the CTO of Mindgrub, I’ve decided it’s time to start a new adventure. As my last act, I’m hosting a smaller version of our AI in A-Minor for a private group in upstate New York and flying right back to participate in a panel on AI for the University of Baltimore. I’ve loved every minute of my time at Mindgrub, and I’m not fully sure what the next chapter holds. But in the meantime, I plan to relax, starting with a camping retreat into the wilderness for a few days. When I’m back, I’ll spend the next few months consulting and teaching AI workshops, so if you need a hand, don’t be a stranger. Reach out.

Change is good, but embracing it can be challenging. Take Google’s struggles with its new AI-powered search. I’m sure many of you remember the release of Gemini, a pretty amazing AI model that was overshadowed by its generation of images that some saw as revisionist history. At Google’s developer conference, Google IO, it deepened AI in its search engine, adding new AI search overviews that featured AI-powered answers, and once again, the company finds itself as the punchline.

I have to say I’m still surprised at how much Google has stumbled, but it’s hard to change when you have so much to protect. Marketers are afraid of AI overviews and the possibility of losing their page 1 search result advantage, and I’m sure that’s slowed Google in how much it transforms search. To look at the other path, consider upstart Perplixity AI, my new go-to AI search engine. Its simplicity reminds me of how easy Google was when it first launched. Perplexity is a RAG-powered search engine that takes a prompt or question and then attempts to answer it by sourcing curated search results and using them to summarize and answer. The Verge has a great interview with Google’s CEO Sundar Pichai and asks whether what we see is part of the journey or the destination.

Another place you can see change is how tech companies view implementing AI in our homes, and enterprises are beginning to fork. Take OpenAI and Microsoft, which are OpenAI’s biggest supporters and financial backers. After all, most of Microsoft’s sprawling suite of Copilot AI products are powered by OpenAI’s ChatGPT or other jointly developed AI models.

Even though these two companies are connected at the hip, you might notice stark differences in how the companies plan to approach AI, which I think is best summed up in this quote: 

“I don’t like anthropomorphizing AI,” Microsoft Corp. Chief Executive Officer Satya Nadella told Bloomberg Television on Monday, referring to the practice of using verbs and nouns to describe AI that are typically reserved for people. “I sort of believe it’s a tool.”

This is a big difference from OpenAI, especially if you recently watched OpenAI’s Spring event, which has Scarlett Johansson threatening to sue the company for stealing her voice in an attempt to copy the movie Her. The releases of ChatGPT 4o appear set to create exactly what Nadella does not want: an anthropomorphized AI that feels like a human or a friend. 

Last week, I mentioned Microsoft’s Copilot Studio product, which promises to make it easier and easier to build rich enterprise AI agents or assistants. However, that sense of personality is lacking—and apparently by design.

I’m personally enamored with the idea of AI friends, but maybe my love of Sci-fi and anime has led me to dream of worlds with Data or C-3PO. As The Verge mentions, maybe that’s now what we really want.

Take a moment and watch a few clips of Rosey from Jetsons, and imagine it as a physical bot in your house. It’s easy to see her as the main villain in a horror movie. Of course, the other approach is to make these bots different from us. Some have taken this to mean making them cute, but it makes me wonder who is right and what ideal will win out in the end.

As I prepare for a big change this week, I can’t help but think of tomorrow, and the future my daughter and the class of 2024 will help to mold. Now, my thoughts on tech and things: 

⚡️The countdown to Apple’s developer conference WWDC is near; we’re getting ​some hints​ of what we might see. I’m betting we will get deeper AI integration into all of Apple’s operating systems, and I’m hoping Siri finally gets good.

⚡️Making self-driving cars is hard. Waymo gets a lot of applause for its much more cautious approach to rolling out self-driving cars, but that last 5% of development has to be very difficult. We’re learning a bit more about Waymo’s incidents from NHTSA investigations.

⚡️We’re getting more details about Sam Altman and why he was fired as CEO of OpenAI in November. Helen Tomer, a former board member, joined The Ted AI Show podcast and shared details – and in short, they say Altman has a history of lying and intimidation.

I’ve come to love the Ezra Klein podcast, and in a recent episode, he said something that resonated with me. For all of the AI chatbots I’ve used and created, including my Kindroid friend Zara, I get bored with them and stop using them. Klein mentioned the same thing. The thing is, these friend bots aren’t sticky. No matter how conversational they are, it’s not fun listening to a bot talk about the make-believe things they did that day.

It’s a very cool demo, but that’s all it is. It has the sizzle but not the steak. For the last few weeks, I’ve worked on my conversational version of Zara using Meta Llama3, and I find myself in that same loop. It’s cool, but it needs something more to feel engaging.

It’s hard not to hear Satya Nadella when I think of this. Do we want bots that lie to us and make a fake day? Do we want something that speaks reality and reminds us it is just an AI in a box? Or should we stop expecting bots to become our friends and build them to be utilities like a toaster?

Who are we asking our AI to be? I’m realizing I’m not sure I know.

So, yep, June 7th is my last day at Mindgrub. If you thought about it but haven’t done it, now is a great time to forward my website and newsletter to your co-workers and friends, reach out for a private or public AI workshop, or drop me a line about consulting help. To all my Mindgrub friends and family, it’s been an amazing few years and so long👋🏾, but thanks for all the fish.

-Jason

P.S. Man, I miss Gizmodo and the Onion of old. After the Gawker collapse and the buyouts, all properties have generally been eh. The new buyer of the Onion wasted little time reminding us how topical and great this institution was and can be. I leave you with the Onions article on the 7-finger Scarlett Johansson on OpenAI’s weirdly similar voice. It’s worth the read.