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
And then there are other companies that have other open sources and other things they've loaded into various large language models that interact differently because of those different probabilities in these different corpora of text. As an example, if you loaded the National Libraries of Medicine content into something else, it's going to be very different than it is looking at Wikipedia, right?
It's not going to know about Napoleon. or things around Napoleon or the context of Napoleon or those sorts of things. But it will know about gallbladder disease and other things in a more complete way than, say, a general purpose large language model. So it's also going to depend on those things. Interesting. Meta, Facebook, has its own large language model based on Facebook. Right.
So it's using these very different corpora of text to create and then layer on top of other technologies to transform those things to take an input and provide an output.
I don't know if compromising is the right word. Are we making it easier to be perceived as intellectual? OK, absolutely. Yeah, because, you know, as people trained in medicine, we took a lot of time in our careers to learn to synthesize. Absolutely. Made the the act of synthesis almost. It's very it doesn't have to be complicated. It can be a simple input and out spits this.
500-word essay on the thing you want or the speech or whatever. The storytelling that goes with that is a synthesis act. It's correlating personal experiences and things that you think might be relevant to the topic that drive a compelling speaker. But you can shortcut it. You absolutely can with these things.
And you can make someone who's very uninformed isn't the right word, but who's put no work effort into it. And then they can regurgitate. Now, is that person going to be on stage? Somebody that's going to be as compelling as somebody that synthesized it and can tell it? They're just reading cue cards at that point. It won't be as compelling.
People will not necessarily be drawn to that because there is that human element. Now, are there other examples of people creating avatars that are as compelling? Potentially, yes. So it can be very short-cutting to, it's almost the human existence, right? Yeah. And the knowledge and the thoughtfulness of our race that has taken millennia to create, could it be? Yeah, it is a real, it's a fear.
Yeah, I think there's two aspects to that. One is the longevity or how long something has been around. And then secondarily, the new knowledge sources that go to inform those things. So if we take someone that's had a chronic condition for 30 years, the summarization of that course over 30 years would take an hour of digging through a chart to figure out and piece together that thing.
AI can do that in a way instantaneously almost, right? That a human cannot. And there are things it can figure out that a human may have missed because it took them an hour as opposed to it's generated a page for me to read and consume and there are correlations in it that may become more clear in that process Now, it can also then say there are correlations here that were missed, right?
In a way that's different. So it's time-saving. Well, it's time-saving, but it's also what's the length of the time the chronic condition has been around? Correct. Now, as a pathologist, even in my time since my training, our knowledge of the genetics, the markers, and other things around cancer specifically and other pathological conditions that are new.
And we can use AI to look back on those things and make correlations as well. Now, when you think about how do you then incorporate a whole genome to the condition that is a chronic condition as well, it doesn't have to be linear or this snippet means that thing. It could be this constellation of snippets means this thing. AI and machine learning in general is good at finding patterns.
And so those patterns that may have eluded a human, in this large volume of information about this individual can be made easier.
It could also go beyond that. It could go generation. Oh, generational. Yeah, yeah, yeah.
Now, we're going to be in an era where many of our records have been digitized. Now our kids' records are digitized. Right. And you correlate those things together.
That a human might not do, but an artificial intelligence could do.
Yeah, I mean... There's been studies that have shown Google searches predict epidemics or seasonality of flu or those sorts of things. So looking at various data streams and correlating them together in a way that is forward looking to say, is this an anomalous behavior to the normal state? Right.
And correlating more varied constellation of symptoms and grouping of symptoms that says, wait, this might be unique. That can be done more readily. Now, public health infrastructure in general, I think people would largely agree, needs an uplift. It's behind many other industries in how we think about data acquisition, data sharing.
There's state and local and federal restrictions and all those problems that come with the data that you want to have that we have to battle past. But absolutely, the capability of AI to look at large data streams and say, wait a minute, where are the patterns in here? Could help.
Could help. And then that could lead to time savings in an action perspective.
Yeah. On the drug discovery side, there's already been some frightening examples of drug discovery being done through AI, right? And creating compounds that are novel and have different properties that could be potentially brought to bear sooner. So it doesn't take a human chemist to really sort through those things. Yeah. Understand those things.
Showing 61–80 of 105 · page 4 of 6
← Previous
Next →