Google's response to ChatGPT-4. It's impressive, but not earth shattering, however... It takes machine reasoning a few steps forward.
Today in health, it Google Gemini is out. We're gonna talk about it on a Friday episode. Alright, my name is bill Russell. I'm a former CIO for a 16 hospital system and creator this week health, a set of channels and events dedicated to transform health care. One connection at a time. We want to thank our show sponsors or investing in developing the next generation of health leaders, short tests are decide parlance, certified health, notable and service.
Now check them out at this week. health.com/today. For those of you who are wondering if you want more information, we have our new site this week. Dot com slash news. Curated news stories from around the industry insiders are clipping stories. Sending them to us. Our editors are putting them out there.
Check it out this week. health.com/news. Let's see what else? It's Friday. Oh, Hey, we're still doing our work with Alex's lemonade stand. Our team got together this week. Exciting. Announcements coming for next year. I'm not ready to talk about it yet, but we're upping that number. Our goal this year was 50,000.
We're up a, I think a little over 60. I haven't checked it recently, but I think we're close to 60. And that is exciting. We want to thank you for your generosity. If you want to be a part of it. Go ahead and hit our website top right hand corner. You can see a link to the lemonade. Stand, go ahead and click on that to give today. Alright, one last thing, share this podcast with a friend or colleague, you said as a foundation for daily or weekly discussions on the topics that are relevant to you and the industry. They can subscribe wherever you listen to podcasts. All right, let's take a look at this.
I'm going to go to my gosh. We have six articles out here on our website about this. ZD net. Here we go. What is Gemini, everything you should know about Google's new AI model, and then I'll give you my, so what Gemini is new powerful artificial intelligence model from Google that can understand not just. But not just texts, but also images, videos, and audio. As a multimodal model, Gemini is described as capable of completing complex tasks in math, physics, and other areas. As well as understanding and generating high quality code in various programming languages. It is currently available through integrations with Google Bard and Google pixel eight.
And we'll gradually be folded into other Google services. So you, Jim Nye is a result of large scale collaboration efforts by teams across Google, including our colleagues at Google research. According to Dennis. Hassabis CEO and co-founder of Google deep mind. It was built from the ground up to be multimodal, which means it can generalize in. Seamlessly understand, operate across and combine different types of information, including texts. Code audio images and video. I'll tell you what, the thing I liked the most about this is gives us choice anyway. Who made Gemini?
Gemini was created by Google and alphabet. Let's see. Are there different versions? Yes, there are. Google describes Gemini as flexible model. That's capable of running on everything from Google data centers to mobile devices to achieve this scalability. Gemini is being released in three sizes, Gemini nano Gemini pro and Gemini ultra. I'll tell you right now, that's probably still too confusing.
The marketing people need to make it even easier, but I will tell you what each one is Gemini nano. Model size is designed to run on smartphones specifically the Google pixel eight, it's built to perform on let's see, perform on-device tasks that require efficient AI processing without connecting to external servers. Gemini pro running on Google's data centers.
Do you have any pros designed to power? The latest version of Google? Google's AI chat bot Bard. And then Gemini ultra. Though, still unavailable for widespread use. Google describes GMA ultra as its most capable model. Exceeding current state-of-the-art results. On 30 of 32 widely used academic benchmarks used in large language models. Research and development. It's designed for highly complex tasks and is set to be released after finishing the current phases of tests. How can you access it? You can access it through your through Bard, through pixel eight and some of that stuff.
How does Gemini differ from AI models? Like GPT four, this will be interesting. G Google's new Gemini model appears to be one of the largest, most advanced AI models to date. Though the release of the ultra model will be one to determine that for certain. Compared to other popular models that power AI chatbots. Right now, Gemini stands out due to its native multimodal characteristics.
Whereas other models like GPT four rely on plugins and integrations to be truly multimodal. And they have a big old chart in here to talk about it a little bit more compared to GPT. For a primarily text-based model, Gemini easily performs multimodal tasks natively. While GPT four excels in language related tasks like content creation in complex text analysis, natively. It resorts to open AI's plugin to perform image analysis, and access to the web.
And it relies on Dolly three and whisper to generate images and process audio. Googles Gemini also appears to be more product focused. Then other models available now it's either integrated into the company's ecosystem or with plans to be as it's powering both Bard and pixelate devices, other models like GPT four and Metis. A llama. Man I've been saying lamp. That's something completely different.
Anyway. Matt as Lama. Are more service oriented and available for various third-party developers for application tools and services. Anything else in here? Nah, that's about it. Let me give you my, so what on this, which is really interesting, cause they there's a video out there. Which shows them like drawing things in front of the Jim Nye model.
And it's, it says, what do you see? What do you see? And then finally it sees exactly what it is, and it's a duck as you're drawing it. And then you make it a blue dock and they say there aren't many blue ducks in the world. And it says, but there are a series of blue dyes.
So it's really good at that. That that reasoning it's visually seeing things and reasoning. So think about that for a minute. And what that means is advances in robots, advances in automation. It can see things and reason and then be put into a production loop. So there's a chance that this actually is more capable from the perspective of it's going to look at something, figure something out and then do something. Think about this on a. I mean in healthcare, you could picture this on the robots that go through the through the hall. And they could see things reason and then make decisions instead of being really highly programmed. And deterministic kind of models of how they move through the through the halls of the hospital and that kind of stuff.
They can start to extrapolate things and see a child and respond to that child or. I see somebody who's lost and respond to them and that kind of stuff. So there's a reasoning that's going on, which I think is one of the more powerful things that I see. In this AI model over a chat GPT today, chat GBT, phenomenal with texts today.
And even the the ability to create images and those kinds of things. But this is a little different in that it's understanding its environment, seeing things it's responding to it and that kind of thing. So I see this. In a robotic sense, being very interesting. And I also see it in a automation thing.
I see it taking information and turning it into actionable knowledge. And I gosh, whose posts Allie Miller had a post on LinkedIn and I'm also gonna put that on our website as well. You can hit our news page and see that. Allie Miller talked about this a little bit here. Let me pull it up.
Here we go.
There is a reasoning and production. Here's what the followers need to know. We are moving from knowledge to action. AI is going to move further upstream and we'll be tasked with taking action. And guess what actions require reasoning, something I've stressed heavily for the last several years and is is the need for critical reasoning. In this process, that's the 2024 focus AI moving into proactive interventions.
Think about that in healthcare. That's exciting. I mean from a robotic standpoint, that's pretty interesting from an it standpoint. I think that's pretty interesting. But just a whole host of things. So if we're wondering, will machines ever take action on their own? It depends how good they are at that reasoning and how comfortable we feel at their ability to make determinations and make the right determinations on the data that they're giving. Let's see, she has a couple other things here.
We're getting more efficient. AI has become 16,500 times more efficient in the last 11 years. And we're starting to see that state-of-the-art let's see. Say a large language models. See what you will about GPT for performance, but it is too expensive for most companies, including open AI itself. To run at scale assume 2024 brings lower AI op ex that's interesting as well. I ever read anything in any of the articles about that, but I would imagine, yes, it's very expensive to run the current large language models.
And so we're going to see efficiency get brought into those models. That's not going to impact us that much. It is very important from a sustainability standpoint, very different, very important from an environment standpoint and other things. But from our standpoint of interacting with the model I don't think we see that per se. Number three, it's a multimodal from here on out.
Chad GBT is a multi-site is multimodal and doesn't require the user to pre-select. What modality. They want to use Gemini seems to be the same. There'll be no lines between images, music, text, voice. Always a big part of Google demos and video. Just one pile of data. Assume that 20, 24 is multimodal very important.
Very interesting. Let's see. And there's no one model to rule them all. We've known. We've now heard this from Google DeepMind lead and more that there will be different models for different use cases. Gemini has three tiers, pro alternate nano pros for Bard Google products, nanos for pixel and ultra is for IMD pockets. I want the highest grade possible performance. We'll be consideration with a score of 90%.
Gemini ultra is the first model to outperform you and experts on a bunch of massive multitask language. Understanding. Assume that 2024 has more multimodal orchestration and delegation. Pretty interesting. This is this is pretty exciting stuff. I don't know that it's head and shoulders above GPT four.
I don't know if it's oh my gosh, I've got to switch or, oh, I'm on this path and whatever. I think you'll see the two things coalesce around this concept of reasoning, seeing things, making determinations. Just all in all becoming smarter. Responding to us in ways that are almost intuitive to what we're asking or the environment that they're in.
And and I think that has lots of opportunities. From an automation standpoint, where would I be looking at this? I'd be looking at this first and foremost in the it world. Find that security alert and respond to it. And I know we've been a little anxious about doing that. We don't want to shut off critical machines and whatnot, but if the machines start to be able to be smarter and apply some reasoning to it.
Oh, that is an MRI machine. I shouldn't shut that off right now. And I should alert some people before I do anything. That kind of stuff, again, you reasoning. The same kind of reasoning. You would want a tech. Who is looking at that log to do. I think we're going to see machines doing that reasoning.
So again, that's where I'd be applying AI as much as I possibly could because I'd have control over as a CIO. And then I would be partnering with others. I think the call center is ripe. For this, you should have a AI. Just. Coming out all over in the call center today. I think that's one area. Obviously I think coding is going to change dramatically.
I think the administrative tasks. Can be, we can look at AI in, in those areas. My gosh, I think the sky's the limit. I would be all in on this. I'd be reading and consuming everything I can. I'd be bringing in experts. From these various companies, I'd be going to these partners locations and asking them. To give you demos and understand it, even if you've decided on the Microsoft route, I would still go visit Google. I would go visit anthropic I would go visit Amazon.
I would go talk to the various players that are out there. Even some of the sub players, because there isn't going to be one model of the rule, them all. There's going to be very specific models that run each one. It's a Friday. I'm rambling. It's exciting stuff. But that's all for today. Don't forget.
Share this podcast with a friend or colleague. We want to thank our channel sponsors who are investing in our mission to develop the next generation of leaders. Short test artist, site parlance, certified health, notable and 📍 service. Now check them out this week. health.com/today. Thanks for listening.
That's all for now.