Google unveils Gemini, its latest AI chatbot
Alphabet (GOOG, GOOGL) shares closed higher on Thursday, a day after Google unveiled its newest generative AI model, the Gemini chatbot. Created to handle text, images, video, and other content, Gemini comes in three versions — Ultra, Pro, and Nano — to compete with OpenAI's ChatGPT and other rivals.
The chatbot's "natively multi-modal" feature is what Google believes will make it stand out—and will be available in Google smartphones including the Pixel 8 Pro, as well the Bard chatbox.
Yahoo Finance Tech Editor Dan Howley analyzes Gemini's features, examining its capabilities versus market alternatives, and talking about Google's broader strategy moving forward.
For more expert insight and the latest market action, click here to watch this full episode of Yahoo Finance Live.
Video Transcript
[AUDIO LOGO]
DIANE KING HALL: Google just introduced its new generative AI model dubbed Gemini. And Yahoo Finance's tech reporter Dan Howley is here with the details. So Google has entered the ring, Dan. Gemini, what do we know about how it competes, especially when you think about ChatGPT?
DAN HOWLEY: Yeah, I mean, this will blow away ChatGPT. Really what this is competing against is OpenAI's latest model, which is GPT-4. So that's what this is the big competitor to.
It's what they call natively multi-modal. So think about it this way, when you have something like a generative AI chat bot, you have them that they work with audio, text, video images, right? But according to Google's DeepMind, those are, kind of, all stitched together to make one actual AI model.
The big deal about this is that it's from the ground up, meant to work with photo, video, audio, text things like that. So in some of the examples they showed, it was, you know, showing this AI a picture of a-- a hand-drawn picture of a duck and it was able to recognize that it was a hand-drawn picture of a duck. And then when the presenter colored the duck in blue, it recognized that it was blue and said most ducks aren't blue. And then they pulled out a rubber ducky that was blue and it said, OK, maybe some ducks are blue.
And it said what material is this? And they were like, well, it looks like it's, you know, some kind of rubber. And they're like, will it float? And it goes-- so it's like it understands all of this, kind of, stuff.
It's not smart, right? It's based on things that it's been exposed to, it's learned. It doesn't have emotion. This isn't "Terminator" or anything like that just to get that out there.
DIANE KING HALL: Well, you say that it isn't "Terminator," but I don't know if you remember, there was an article out earlier this year. I think it was a Kevin Roose piece where he was interacting with Bard, right? And it started to seem to take on this persona that was just like-- it just got extreme and professing love if you remember that.
DAN HOWLEY: Yeah.
DIANE KING HALL: So like, you're saying it won't be the Terminator, but there is an issue when you think about guardrails when you think about, you know, what's happening with the generative AI space. So what would you say is the takeaway? So you say this is-- of course it's going to blow away ChatGPT, but what about, like, where OpenAI is now and then in general all the players that are developing their own, like Meta, their own answer to that?
DAN HOWLEY: Yeah, so this is going up against GPT-4, so that's OpenAI's top of the line model, right? So this is going to be a big deal. And the other thing about this is there's three versions of it.
So there's the top-end version that's going to be running in servers. That's called Gemini Ultra. Then there's Gemini Pro, which is going to be for more casual business users. And then there's Gemini Nano, which is actually out running on the Google Pixel 8 Pro smartphone.
BRAD SMITH: Oh, yeah.
DAN HOWLEY: So it's in devices. Bard is getting a Gemini update today. So they're rolling this out now.
It still needs time. It's only English language versions of this. There was a report that they were holding it back entirely because they only had the English language version. But clearly, they're rolling it out.
And I think the big thing to keep in mind is that, you know, unlike Microsoft, which is adding this as a value to its cloud service that makes up a good bulk of its revenue, Google is still wholly reliant on search and advertising, right? So they're going to add this to their workspaces. They have Duet AI, which is their workspaces AI. That's going to get a Gemini boost as well.
The smartphone market for them still pretty small, right? It's all about for them advertising. And so they specifically point out that this is going to be going into the search section, as well as YouTube and then obviously workspaces.
But it's the Google search part that's most important because right now they have something called Search Generative Experience. It's a version of Google search that runs on generative AI. So you'll search for Yahoo Finance, you know, Dan Howley or something and it'll show--
DIANE KING HALL: Of course.
DAN HOWLEY: --it'll show very handsome, great writer, things like that. So that's the generative AI version. It's also known to lie, so that's probably why it does that.
[LAUGHS]
DIANE KING HALL: Dan.
DAN HOWLEY: But then this is going to come eventually to full search. So this is basically their future proofing going down the line. They also just need to show people that, you know, OpenAI--
DIANE KING HALL: That they're in the game.
DAN HOWLEY: Isn't the only game in town. Exactly.
DIANE KING HALL: They have an answer. Exactly.
BRAD SMITH: Maybe it will surface the beard oil products that you use too.
DAN HOWLEY: Zero. Head & Shoulders--
BRAD SMITH: Really?
DAN HOWLEY: --every day.
DIANE KING HALL: Oh my, wow! Old school, all right.