AI, Explained: Why It’s Different This Time | WSJ Tech News Briefing

welcome to a special episode of tech news briefing for Monday April 3rd I'm Zoe Thomas for The Wall Street Journal you've probably been hearing a lot about artificial intelligence AI seems to be everywhere these days Google meta slash Facebook Microsoft are in a race to introduce new artificial intelligence systems the AI race for for search it's on the artificial intelligence Company open AI unveiled chat gpt4 it's the photo editing app that's taken over social media lenza AI it's not the first time AI has promised to shake up our world but something about this moment does seem to be different generative AI is changing the game and maybe a lot of other things too we are Tech news briefing wanted to know more so we've launched this series airing Mondays in April called artificially minded to look at what's new now and what it means for ai's future and our own but first to understand what's going on with artificial intelligence and generative AI we need to start with some Basics because how AI works and what it can do is complicated So today we're going back to school joining me to break down AI 101 is wsj science reporter Eric neeler hey Eric thanks for joining me hey it's great to be with you so I want to deal with the basics today when we talk about AI what do we even mean sure well artificial intelligence has been around for a while it started out more as a statistical kind of method for making decisions with computers and over time of course it's evolved but really when we talk about AI it's any kind of technology that can reason learn plan and make decisions you know these are all tasks that normally require human intelligence but something else or someone else is doing it and that's artificial intelligence I feel like another term that I'll often hear when we talk about artificial intelligence is machine learning can you tell us what that is and how it relates sure machine learning is a form of AI form of artificial intelligence and instead of being programmed to do a specific thing maybe check the temperature and give a reading or go and search for a couple things and come back it can actually learn to do a specific task on its own it's learning by figuring out patterns in data and also inference if this is so then this must be so if we add these two things together so examples of machine learning for example Uber and other car sharing apps use this to predict the demand for drivers and passengers and also machine learning is also used in the medical field to identify and predict cancer tumors by looking at Medical scans maybe early detection of these tumors or medical problems so it is just learning the way we would learn you know if if one thing happens then another thing might happen is that right yeah imagine associate and there are also different ways you can train an algorithm an algorithm is like a little program right so you can train it with reinforcement you did a good job here's a little stick or you get more data for your training and when we talk about training these programs need big amounts of data so example let's say a facial recognition program in order to get it accurate you're going to need large numbers of people's faces so that the program can tell differences and it can make matches and so forth or if the program is designed to look at Medical scans for example the larger number it has to get trained on the more accurate it's going to be so the training happens first and then once it's the person that's written this program feels that it's accurate enough then it's going to do it on its own if we're thinking about the generative AI tools that people maybe have tried out now is this kind of how something like chat GPT learns to respond I'm fine when you ask how it's doing yeah it's funny what chat GPT does is look at vast amounts of text and it sort of learns from that text it assimilates this into something called a law large language model this is another form of AI that is employed by chat GPT so it doesn't have knowledge like we do in our brains but it has the ability to process language really fast and really deep ways of making associations between your question and maybe a question that's been asked somewhere else or something it's read in a book or on a website to come up with an answer so just like students it's learning you know what's in books so we can respond with them it's getting treats like stickers or whatever the computer equivalent is of a sticker If you get something right in school Eric another term that I feel like I've heard associated with artificial intelligence is neural network what is that so our brain is made up of neurons that are connected brain cells right and it does processing it doesn't do it from the top down it sort of has these little places where processing happens with language with mathematics with music with faces and so a neural network is a computer program that sort of mimics that so a neural network can identify images patterns text facial expressions by working on several levels and also several nodes of processing and again that's why they use the term neural network because it's a little bit like how our brain works a lot of people will have heard of potentially tried out chat GPT or maybe an air generator there's more and more options for people to try their hand at this but AI has been around for a while can you just tell us some places that we might have interacted with it just in our daily lives that we haven't thought about sure there's some very simple things apps that we use on our phone to to get us places navigation apps the Google search bar they're also you know when delivery drivers have to make their routes making the best route most efficient way there are other places too where it's being used by law firms that are needing to search massive amounts of cases of case law for example to find references to find applicable language used to be clerks would do this or paralegals and now they're sophisticated AI machine learning tools that can go through years of digitized cases and legal proceedings to get just the right answer to file in your brief and that's got some people worried too so it's out there it's being used I think the most important things are that we understand it and also the people that are making these programs understand the human element what are they trying to do how are these algorithms affecting people okay in a moment we'll continue our AI 101 conversation with Eric neeler but first we asked people on the streets of downtown San Francisco how they felt about AI how do you feel about artificial intelligence I think it's necessary that simple I I know that everybody's scared about it um because it's we hold knowledge in our brains and so if something can replace that then we are very replaceable I honestly like it I think it's very [Music] um it's just an innovation and they think it's very important it's helping improving a lot of fields and a lot of companies so I think it's important to work on it right now both optimistic skeptical this moment in AI feels different from developments in the past I'm back with wsj science reporter Eric nealer Eric a lot of what we've been talking about are artificial intelligence abilities that have already existed and that have been integrated into our day-to-day lives but we're seeing a lot more about AI in the past few months so what's different about this current moment than what we've seen over the past few years I think what's happened just in the last six months or so is the introduction of these large language models the chat GPT for example that just has some remarkable human-like abilities to answer our questions to write memos to make what we think could be opinions about other people or world events or Twitter for example and this is just kind of shaken us I think a lot of us because it has appeared to be some kind of new being or new Force when it's really just more of a repetitive kind of thing it's a technology that's learning from text and responding with texts that it learns and you know I I just was seeing this on Twitter remember chatgpt is not a knowledge model it's a language model and what that means is that it doesn't know things it's repeating things it seems like what we're realizing now is that there are quite a lot of risks with artificial intelligence can you take us through what some of the big ones are you know I think the everyday risks are things about accuracy bias in facial recognition also another risk is that there's a failure in system something that hasn't been thought up before by the programmer it's important to realize though that the risks are all within sort of a structure in the sense that the programmer is telling the AI where it can work what it can do the data that it has to use these are you know in many ways the researchers I've talked to and interviewed say these are manageable risks to solve these problems or to reduce these risks we need more data to train on in order to eliminate or reduce the risk of mistaken identity for example in facial recognition you need better lighting better images more images using training sets with more black and brown faces for example that'll help that algorithm make better facial recognition decisions the risks though that we often talk about can also be this sort of existential risk that AI is going to take over the risks that are identified in Hollywood films and on science fiction books about AI sort of going beyond the boundaries of what the programmer tells it to do or the training set that it's using so far we really haven't seen that and I don't think that's going to happen anytime soon to be honest are there any rules or laws about how AI can be used or cannot be used here in the United States there really aren't that many rules governing AI we have seen it attempts to restrict or limit its use by law enforcement agencies in various states and cities on our federal level no we're not seeing it in fact many federal agencies are expanding their use of AI facial recognition Technologies to say limit access to buildings to make determinations on you know employees and so forth we are however seeing some sort of move to regulate AI in Europe there are much more concerns there and they've put in some new rules just in the last couple years we're going to see how that filters down how that affects businesses and of course how it affects society and the people that use AI all right that's wsj science reporter Eric nealer thanks for joining us for this conversation Eric thanks very much okay that's it for this first episode of our series artificially minded join us next Monday when we'll answer questions submitted by you on artificial intelligence until then you can catch regular episodes of tech news briefing for the rest of the week and if you want more Tech news check out our website wsj.com Today's show was produced by Julie Chang our supervising producer is Melanie Roy and our executive producer is Chris sinsley this show was mixed by Michael LaValle and I'm your host Zoe Thomas thanks for listening [Music] thank you

Channel: Wall Street Journal

Views: 218,858

Keywords: ai explained, artificial intelligence, open ai, chatgpt 4, lensa ai, tech news briefing, wsj, generative ai, artificially minded, what is ai, ai, machine learning, what is machine learning, patterns, data, inference, data analysis, algorithms, uber ai, medical ai, how generative ai works, large language model, neural network, ai art generator, navigation apps, ai detector, ai chatbot, ai chat, ai art, ai stock, ai risks, artificial intelligence examples, ai regulation, techy

Id: 8liUOepAO9s

Length: 13min 19sec (799 seconds)

Published: Mon Apr 03 2023