AI round-up: Week of March 3, 2025
Let’s just go ahead and get to it because I may have seen the most mindblowing thing I’ve seen yet…
The Heavy Stuff
Story 1: Open ‘Sesame’ (Sesame)
Have you met Maya or Miles yet? No? Well, you’re going to want to do three things:
1. Stop what you’re doing and meet them.
2. Make sure you engage with one of them for at least a minute … just so you can get past the awkward part.
3. Leave about 30 minutes after you’re done to clear your head.
NOTE: if you can’t access them (it?) because of the traffic, Ethan Mollick recorded his conversation with Maya which will do the trick.
If you’ve been looking for the uncanny valley … you found it.
This was just … too much.
Story 2: I guess everything has a shelf life … even Google’s blue links (Yahoo! Tech)
Google released an experimental version of its new AI-only search, meaning you won’t see the blue links when you search.
Let that sink in.
Instead, you will have the option to go to AI mode which generates results in an AI Overview approach. I checked and found out I don’t have it yet because they’re rolling it out in batches to subscribers of Google AI One Premium (which I am). They told me I’m ‘on the list.’ Please note – as a result of this news, it’s very important you read the next story.
Story 3: How to optimize your marketing for AI (Christopher Penn)
I mean, the headline says it all but please know … this is an intensive article. Chris really jumps in and pulls apart traditional SEO and tries to carry us all through the thinking around how we should be preparing.
It’s technical at times, but I highly recommend trying to understand it … or partnering with someone (or some agency?) to help translate and do what he’s recommending so AI can ‘find’ you.
One of my favorite parts of this blog is this very easy to understand take on how to use AI tools … to help you be found on AI tools:
For example, if your content is about baking bread, what are the expected questions someone might have after reading your content? Ask an AI to give you those questions, and then you incorporate those questions into your content.
And remember to keep your FAQ pages relevant, fresh, and beefy. The bigger they are, the more training data they provide to AI models. Make sure they're loaded up with appropriate brand references so that each question has an answer pair that contains your brand.
Story 4: OpenAI to charge $20,000/month for Ph.D level agents. (Yahoo! Finance)
Alternative headline: OpenAI needs money.
Story 5: Sometimes it’s important to ask ‘why’… (CNN)
As in ‘why’ are we doing this?
For some reason, The Los Angeles Times released an AI tool that generates an opposing viewpoint to its op-ed pieces. Well, I guess it worked too well because it defended the KKK who were called out in a recent editorial.
Now, first … I’m not sure if The LA Times knows this or not, but that tool already exists. It’s called Facebook.
And second, as I’ve said many times … just because you can, doesn’t mean you should.
Story 6: You know what we need? Another AI model. (Business Insider)
Thanks Amazon! It had been a few hours since someone announced a new one.
The not-so-heavy stuff
Story 1: Google is introducing AI to help detect spam in text messages. (The Verge)
Story 2: AI is being tested on Super Mario Bros. (and doing pretty well) (TechCrunch)
BUT, can it find those secret flutes in Super Mario Bros 3?
Story 3: Microsoft unveils new voice-activated assistant for doctors. (CNBC)
I’m not sure why it’s called Dragon. Makes me think of medieval medicine. Also, can we train it to say ‘Dammit, I’m a doctor, not an AI assistant!’
Story 4: AI podcast hosts on the rise. (TechCrunch)
As in, you now have 400+ to choose from.
Story 5: Parents are teaching their gen Alpha kids how to use AI (The Guardian)
But who’s teaching the parents?
A few that don’t fit in either category
Story 1: Should AI models be unfiltered? (Shelly Palmer)
This is kind of like asking if people should be unfiltered.
Look, I get the argument. As a matter of fact, I talked about it last week in my final note when I encouraged us to think for ourselves, verify info and not accept the creators’ version of the AI we’re using.
But an ‘unhinged’ AI … doesn’t seem like something we need right now when we don’t even know how to regulate ‘hinged’ AI. I’m not saying to censor AI or curb AI learning and training. I tend to agree with Shelly – I think this will be required at some point. But instead of serving as an accelerator, can we treat it as a warning or reminder that we should be building parameters that can scale with these models, in whatever form they exist?
Story 2: The author of last year’s California SB1047 bill has a new one to protect AI whistleblowers. (TechCrunch)
Final Note
As the goal posts continue to shift on achieving AGI (and what AGI is, exactly) the conversation continues to evolve. One conversation in particular is the one Ezra Klein had on his recent podcast with former Biden AI adviser, Ben Buchanan.
Klein brought Buchanan on to talk about AGI, specifically the fact that so many people from diverse circles think AGI will be here within the next two years and why nothing is really being done to prepare.
Thanks for reading.
-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.