Michael King

Follow

--

Have you ever wondered what’s going on inside the ‘mind’ of an AI? Or better yet, can these astonishing pieces of technology even reason like we do? It’s a thought that’s been keeping me up at night, and I’ll tell you why.

Just recently, Google dropped a bombshell in the tech world. Palm-2, their newest Large Language Model, was unveiled to the public, sending ripples of anticipation and curiosity through the digital universe. As an avid AI enthusiast, I felt like a kid on Christmas morning. A new player in the game, a fresh face in the AI crowd — it’s the kind of news that sets my pulse racing.

But what truly caught my attention, what truly had me sitting on the edge of my seat, was the question of reasoning. How well could this shiny new model reason? Could it analyze, infer, deduce, just like a human mind? Or even better, could it stand toe-to-toe with OpenAI’s GPT-4, the current heavyweight champion of AI reasoning?

Now, there’s a little snag in this exciting narrative. As much as I’d love to have a firsthand look at Palm-2, I don’t have access to it just yet. So our exploration is going to be a bit of a treasure hunt, combing through the publicly available information to piece together a picture of Palm-2’s reasoning prowess.

As I scoured the official Google blog, a something very intersting caught my eye. Nestled amidst the technobabble and the hype was a screenshot — a snapshot, if you will — of Palm-2 in action. A reasoning question, skillfully navigated by this new contender.

ai.google

Picture this: On a shelf, there are four toy cars. Each a different color — green, blue, orange, and yellow. The green car is parked to the left of the yellow one. The yellow car isn’t the first, it’s the second from the left. And the blue car? It’s sitting pretty on the far right, the last in the line-up. So, how do these cars line up? Can you visualize it?