Transcript
Micah Sargent (0:00)
Coming up on Tech News Weekly, Emily Forlini of PCMAG is here. We kick off the show by talking about Claude AI from Anthropic playing Pokemon on Twitch. Then Emily, who was at Amazon's recent Alexa event, tells us about what we can expect from the AI powered virtual assistant. Before Sean Hollister of the Verge stops by to tell us about his time with Fred Framework's new desktop PC. It's sorta kinda modular. All of that coming up on Tech News Weekly.
Emily Forlini (0:37)
Podcasts you love from people you Trust.
Micah Sargent (0:41)
This is TWiT. This is Tech News Weekly, episode 376 with Emily Forlini and me, Micah Sargent. Recorded Thursday, February 27th, 2025. Amazon's new AI powered assistant. Hello and welcome to Tech News Weekly, the show where every week we talk to and about the people making and breaking that tech news. I am one of your hosts, I am Micah Sargent and I am joined on this vast and ever changing landscape that is the Internet by the wonderful Emily Forlini. Welcome back, Emily.
Sean Hollister (1:21)
Hey, Micah. Happy to be here.
Micah Sargent (1:23)
Happy to have you. So we have some stuff to talk about today. As people are aware, we typically kick things off with our stories of the week. Emily did just get back from an interesting event, so she's agreed to stick around for a little while longer and tell us about that. So I'm actually gonna be kicking off the show this week with my story of the week and that is about a new. A new player has entered the race, so to SPE speak, because Anthropic, which makes the Claude AI decided to essentially use, I would say, kind of a newer benchmark to test the capabilities of its AI model. And in doing so, we have been blessed by the introduction of Claude playing Pokemon Red on Twitch. So I want to start by kind of laying the groundwork here because it's important to understand this isn't a case where we have, you know, a bunch of YouTube videos, right, of people playing Pokemon being uploaded into the system and saying, hey, watch, this is how people play this game. Here is the sort of game manual. Here are the in game instructions for how to play. No, instead what they've done is they have actually just dropped essentially Claude into a game and said, now learn how to play it and then play it. So there's no kind of prior understanding to how to play this game in ways that We've seen other AIs play games. Okay. This is kind of just like a. What the. The. Not Crucible, but what is the. A trial by fire. There we go. A trial by fire. Situation. And I found this incredibly fascinating and enjoyable to watch. I've had it kind of running in the background for a while now, just seeing how this whole thing plays out. Because what you have is essentially in every moment, not just the AI trying to move around, because it's got like a. It's tied into the emulator so that it can hit, you know, up, down, left and right, and a. And that kind of a thing. But what. I found it fascinating almost as a way to think a little bit about how we think and to see all of that play out and to think about all of the things that we might think about in a given moment. Because there's one. One kind of quirk of this game in comparison to maybe some other games is typically, many of your interactions in RPGs are going to involve walking up to something, be it a person or a door or a store or whatever, and hitting a button, the A button, the main button, to say, I want to talk to this person or I want to interact with this thing. One thing with Pokemon is that doors and certain other, like, entrances and exits only require just kind of moving once more toward it. So you get to the spot and then you move once more, and then you can leave the space. And so you see the AI trying to figure out how to leave. And, you know, I walked up to the door, I've pressed a. It's not letting me leave. Finally learning how to do so. And then the thing has gotten stuck at different times. In fact, just recently, it got stuck somewhere and decided that the. This. This felt very. And instead of trying to figure out what it needed to do to get out of its being stuck, it just decided that the game was broken. So it's like, hey, this isn't me. This is the game. And that is so human to me, which I found very fascinating. But I don't know, I kind of wanted to talk to you first and foremost about maybe how much do you think a. A benchmark like this matters in the scheme of things? And also just if you had heard about this and what are your thoughts on this? And then also, did you play Pokemon? Have you played Pokemon? I'd love to hear about that, too.