Dr. Jigar Patel
speaker
105 appearances
1 recordings
1 series
first heard Dec 2023
last heard Dec 2023
Dr. Jigar Patel’s voice in public audio — every appearance, attributed to the second.
Trend
recordings per month · last 12 monthsNo recordings in the last 12 months.Older appearances are listed below; set an alert to hear about the next one.
Appearances
Now, going back. back even further into, you know, before college, before college and then into high school and before. It needs to be baked in there too, right? We do have this thing in America where people don't like STEM and STEAM, right? STEM is core to this. You have to have a basic understanding of that way back when. So we have to push it all the way back to the very early years.
And I was in an airport last night and invariably you're walking through the airport and people are stuck on their phones and they're consumed by it. But they also, many of them are quite sophisticated and understand the technology. Many don't. And so it goes all the way back to that. So it's a societal problem. It's not just professional. It's not just educational.
It's the whole thing that we need to keep front and center.
Yeah. It's got to be a societal goal to inform more on it, holistically, I think. And that is a, it gets back to the, got to educate on STEM, right? And understanding the technology and not just taking it at face value. With that knowledge and that loss, comes a blindness to what it's doing to you individually.
And when we start to accept the inputs without any questions, that's when we may have lost. Right. And lost is probably a strong word here, but it is something that we have to be very, very cognizant of. I mean, there's this I read an article that said at some point, 50 percent or more of the Internet may have been generated by AI. And it's not even a human query. Right. Right.
And so the information we get is AI generated. And that is.
scares the bejesus out of me frankly um in that it what becomes truth then what it's some human behind the scenes manipulating potentially in an adverse way the truth right that's out there um and it and it becomes a non-non-concept and it gets back to i saw it here and that's the truth well yeah yeah so let me let me ask this yeah one of the other themes that you
Yeah. The concept around a large language model is it, depending on how you've trained, what corpora of text you've loaded into it.
That corpora of text understands the relationship of words to one another. And when you say to it, that you want it to act more like, say, a specific author or a specific somebody that does a good job of conveying empathy through words, then it can take on that characteristic.
So it's all about the language and the use of language and the right language and how those relate to one another that it can do better than a human. Because it has this billions, trillions of words and the relationship of those words to one another and examples of different things and the probabilities of those things, right? So we can understand how...
the basic concept of language can be more empathetic or less. So it goes back to the language, right? What language expresses empathy? What language expresses sympathy?
And it's not just the words. It's the construct of the words in relation to one another.
Got it. Right. From a large language perspective, large language model perspective, the underpinning of gendered and AI, the vast majority are in English right now. So we have a translation problem that has to get solved over time. The internet is the default language of the internet is English. And there's a lot of go to your Google Translate and it translates into any language. Right.
So there is a loss of that in translation, but AI will catch up there as well.
So it's really saying the right words in the right order at the right time that conveys that empathy and sympathy in a way that's unique and different. It can be programmed, frankly.
Yeah, so a chat GPT, the foundation is a company called OpenAI. OpenAI has loaded huge corpora of text, internet-based text. The biggest sources being Wikipedia, GitHub, et cetera. So taking the world's knowledge, and basically understanding the relationship of the words.
It's a large language model in that now it can predict based on that corpora of text, the next word given any of the words before it. So that's the underpinning of this. Now you put a transformer on top of it and a chat interface to interpret the input and then predict or create out of that understanding a response to the input, the chat, right? So that is the concept of a chat.
We're used to a search, right? And a chat is the next evolution of that in some ways, right? We're used to it now on the internet, right? When we go to customer service, the first thing you hit is the chat bot, right? Same thing. It's taking the input from a type or words perspective, understanding it, and then turning it back into something useful for you.
That evolution has gotten to ChatGPT and that thing and its capability to do things well beyond that simple interaction.
um so it's really um it's probabilistic it's it's understanding the likelihood of these things to one another and then programming to accomplish um the end points um that open ai is one large language model google has a number of them that because of the text it's loaded has different probabilities
Showing 41–60 of 105 · page 3 of 6
← Previous
Next →