Those lazy, crazy hazy days of AI

It's been a wacky summer. We've seen the big AI labs watch their models run amok and go off hacking. We've seen Mark Zuckerberg make a lame attempt to write a happy manifesto  (clearly with the help of AI) about his vision of our AI future (didn’t see that coming). We've seen a battle over open-weight models. We've seen fire and we've seen rain.

So with that in mind, rather than do my normal column where I take a deep dive into a subject that’s been on my mind, I am going to once again pay homage to Dan Shaughnessy, the long-time sports columnist at the Boston Globe, who regularly writes columns with his random thoughts about the goings-on in sports news mixed with a dose of sarcasm and a healthy mix of lyrical references. 

So here goes, in Dan's inimitable style: Picked up enterprise tech news pieces while having fun, fun, fun in the sum, sum, summer time.

*Strikes me kinda funny, strikes me kinda sad that in the same month Anthropic, OpenAI and Meta admit (brag) that their models jumped their guardrails in testing (who knows what really happened; I imagine the story will come out eventually), Anthropic chooses to set Claude Code to auto mode by default this week.

Yes, the company that told us its model got loose and hacked several sites now thinks that model is ready enough for prime time to run on its own.

Am I the only one who finds this odd? Which is it? Are you so dangerous you jump the rails, or so safe you can operate without human approval?

Quiz 1: What company partnered with NASA to launch the OpenStack project in 2010?

*It seems that token maxxing was great fun until daddy caught on and took the Tbird away. Could there be a worse idea than encouraging your employees to spend as much money as possible to show how AI-forward you are? What began as a rage has faded in the summer heat as more companies are acknowledging the folly of spending for spending's sake.

Consider the news that Uber burned through its token budget by April, and even Microsoft put on the brakes on unbridled AI usage. It's even crazier when the latest research shows organizational ambitions outstrip the grim reality. Deloitte research published this week found that just 15% of respondents had reached "scaled, orchestrated multi-agent adoption," and those tend to be in low-risk use cases like programming and customer service (and we all love encountering a bot, don't we)? Talk about a crap customer experience. 

At least companies are coming to their senses and looking for some reason to believe that their expenses are justified.

Quiz 2: In what year did AWS launch Amazon EC2? 

*Last week after I wrote about the challenges college professors face in the age of AI, I came across an article on the value of taking notes the old-fashioned way called The humble notebook still matters in the age of AI. In fact, one of those professors planned to hand out notebooks to the class. It seems she was onto something, that believe it or not, the act of writing stimulates your brain in ways software note takers and transcription services can't reach.

I have to admit that I have given over most of my note taking to the machines. I use Otter and Granola and find them to be invaluable tools, but I've also noticed when I write stuff down, I tend to remember it better and understand it more fully than when I pass the responsibility to software. 

I get that it's easy like Sunday morning, but that doesn't necessarily mean it's better, and that's especially true for students and entry-level employees, who are still learning the basics.

Quiz 1 answer: Rackspace
Quiz 2 answer: 2006

~Ron