C.R. Bard

This year's Fortune 500 marks the 63rd running of the list.

Fortune 500

This year's Fortune 500 marks the 62nd running of the list.

This year's Fortune 500 marks the 61st running of the list. Wal-Mart claims the top spot again. See who made the list.

Wal-Mart stayed on top as its sales crept closer to half-a-trillion dollars. Apple moved into the top five. And UnitedHealth Group continued its steady climb. 

Key Financials (last Fiscal Year)

Revenues ($M)

Profits ($M)

Assets ($M)

Total Stockholder Equity ($M)

Market Value ($M)

Profit Ratios

Profit as % of Revenues

Profits as % of Assets

Profits as % of Stockholder Equity

Earnings Per Share (last Fiscal Year)

Earnings Per Share ($)

EPS % Change (from prior fiscal year)

EPS % Change (5 year annual rate)

EPS % Change (10 year annual rate)

Total Return

Total Return to Investors (last year)

Total Return to Investors (5 year, annualized)

Total Return to Investors (10 year, annualized)

Get insights into C.R. Bard including business profile, stock information, news, and rankings from Fortune.

C.R. Bard Company Profile, Stock Price, News, Rankings

legacy

c.r. bard

Companies

C.R. Bard | Fortune

C.R. Bard Company Profile, Stock Price, News, Rankings | Fortune

Sissie Hsiao, Vice President and General Manager, Google Assistant and Bard, Google Moderator: Jeremy Kahn, Senior Writer, FORTUNE; Co-chair, Fortune Brainstorm AI

Brainstorm AI 2023: Future of Google Bard and Generative AI

Conferences

Hello. Hi, everybody. Welcome. Welcome to the 2023 Fortune Brainstorm A I conference. Great to see you all back here again in San Francisco. Um This is the third year of Fortune's brainstorm A I conference. Yes, we did get started before chat. GP T long before. Uh And many of you have been here with us from the beginning, artificial intelligence as we all know is not a new concept. In fact, it's been around since at least 1956 when Dartmouth hosted a summer research project and conference. And the concept was not all that different from ours here today to gather some of the greatest minds in the field and come to uh more or less of an agreement may be on a shared vision that computers can be made to perform intelligent tasks. I think we've established at least that much. And now with the explosion, of course, of open A is chat G BT and other technologies. Um Last November generative A I has garnered the attention of regulators as well, the general public and businesses alike with A I's newfound attention and capabilities comes a wave of both fear and excitement. On how it will be deployed in the coming years and what effect it's going to have on our society as a whole. A I as well as on its way, revolutionizing fields like health care, entertainment, manufacturing, education and a whole lot more. But with power comes great responsibility. A I has the potential to perpetuate, perpetuate biases and in the hands of bad actors spread misinformation and open the door to major security risks. Many fear it's going to replace jobs. Some have even argued that A I could even replace CEO S and journalists. Oh, no. Um With so much unknown and unregulated, we look forward to insightful conversations over these next two days as we examine new business cases for A I and explore how it's already changing the way we live work and do just about everything. We have a stellar lineup of speakers. But remember that this event was really created for your participation and is meant to be an interactive discussion among speakers and attendees. The most important conversations are those among all of you. Um As we navigate the innovations and the challenges of implementing new technologies and new ways of collaborating. Before we start some important Acknowledgments. I want to recognize the incredible work of our brainstorm A I, co-chair fortune, senior writer Jeremy Kahn Fortune, senior editor Vern Kito and co-founder and Ceo of Afecta. Uh Doctor Rana LK Calli. Thank you to our co-chair. We are also pleased to be joined by additional Fortune colleagues who are going to be moderating some of our discussions, Alan Murray, CEO of Fortune Ellie Austin, Fortune's deputy editorial director for live media and Jeff John Roberts, editor of Fortune crypto. Uh Before we get started, I also want to extend a tremendous thank you to our partners for their contributions to this event. Our presenting partner, Accenture, our partners, a ARP Amgen checkpoint hand show Sales Force Workday and our supporting partner Felicia. Before we get into our sessions, an important reminder about our conversations, you probably get sick of us saying this, but it's important. What's said on the main stage or during the strategy sessions uh is on the record unless otherwise noted. And what's said over a break or over dinner or a walk is of course private unless agreed upon otherwise by all parties. OK. Let's get started. As mentioned earlier. The release of open A is chat GP T has brought worldwide attention to the uses of generative A I chat bots and the ability to release large language models to the public. A few months later, Google released its conversational generative A I chatbot known as Bard. Since then, Google merged the Bard teams with the Google assistant team to create assistant Wood Bard, a new personal assistant chatbot powered by generative A I. Google also unveiled a generative A I search experiment that may alter how we all find information and possibly disrupt digital business models across industries. Last Thursday, Google announced the release of their newest model Gemini, which you all probably know all about. Uh And that is now going to Power Bard here to talk about the future of Gemini powered Bard. What's next for the chatbot? Please welcome General manager for Google assistant and Bard Sissy show. She's going to be interviewed by Fortune, senior writer and brainstorm A I co-chair Jeremy Kahn Sissy, thanks for joining us and thank you all for joining us. Um We're going to go to questions from the audience uh hopefully towards the end of the session. So please start thinking of your questions. Um Sissy Big News last week with the unveiling of Gemini and the announcement that Gemini was now going to be powering barred. Um There was also a lot of uh disappointment, I think over the video of the, the demonstration that came out with uh that announcement um in, in Heinz, do you regret sort of how that that video was handled? Well, maybe let me start by saying thanks for having me, Jeremy and thanks for having me for start there. Um Very excited to be here to talk about generative A I and, and bart, of course, I mean, let me start with the last part of your question. The video is completely real. All the prompts and the model responses are real. We did shorten parts for brevity which we put in the video as you know, informational on the making of the video. But we're also looking forward to releasing the API in A I studio this week so that developers can play with it. But you know, that's really addressing the video. I think the main, you know, headline story is we've entered the Gemini era. This is the best set of base models in the sort of medium size range and the large size range on a variety of industry benchmarks that all researchers benchmark their large language models off of. And so I've been really excited to take that expression of that power in those Gemini models and put them in Bard, which we saw with our side by side blind evaluations, right? So Bard is the best most preferred free chat bot now um in the market in blind evaluations with third party Raiders. Great. And what has been the most interesting use cases you've seen so far with Bard or, or things that have really surprised you that people are doing with it? I think what surprises me is the just the breath of things that people do uh with the chat bots. It's really hard to say it does just one thing or the other thing because it does so many things. So I'll give you an example um you know, for my personal use um at my home life, like my daughter just got her ears pierced. And so I was looking for a particular type of earring that she was looking for that wouldn't cause allergies. I brainstormed with Bart about that. Uh Bard gave me some brands. Bart also told me how to take care of your ears to not get an infection. That's a very like personal use case that was able to, you know, take that problem and, and solve it for me. But um at work in productivity cases, we have people using it for, we have, I have a coworker who's uh English is a second language uh for him. And so he uses Bard to rewrite his emails to be more professional. So really these a is have just such a broad, you know, use set of uses because they can role play, they can generate content, they can brainstorm with people. So I'm really excited to see just like the breath of uh things that people are trying and also, you know, what really resonates with users in the world. And when we look ahead to Gemini, what really excites you about what you're going to be able to do with Bard now powered by Gemini that maybe you couldn't do six months ago. Yeah. Well, I mean, first of all, Gemini is just a stronger base model which means like everything that comes out of it is more high quality. So we see it in text generation, we see it in code generation, we see it in instruction following. So uh typically complex prompts have many, many instructions embedded and the model is able to follow those instructions. Um But I think the most like sort of visually stunning is of course multimodality, right? This is where the model can ingest, you know, not just text but image and audio and video and also communicate back out to the user um in not only text but also, you know, in other sort of means of communication like video and image and, and audio. Um a very simple example. Um I just used bar to uh I went to a dinner, I had a menu that was like, choose a course from the 1st, 2nd, 3rd course. And then I had a wine list and I took a picture of the two menus side by side and I said, choose me a dinner for tonight, one course from each section with a wine pairing to go with it. And it actually did fabulously. And so like even something like that as simple as just taking a picture of something and asking a question or for the bot to do something for you um will just become second nature as like these multimodal models uh become, you know, as easy to access as your phone. Yeah, interesting. Um And a lot of people are looking forward today when, when these models, um you know, you can prompt them to go out and do something for you to take an action on the internet on your behalf. Uh How far away is that? Kind of agency do you think and will Bard be for Google the thing that, that carries out that your mission in the world for you? Yeah, I mean uh getting things done for people is definitely within reach of these uh agents or we call them agents because they have this agent of capability. So already Bard can summarize your emails, right? We've connected Bard to gmail, youtube. I can summarize youtube videos for you. You can ask questions. So already like this action ability is starting to emerge when you connect different tools that exist in the world into the large language model and foundation. The reason why it's able to do this is you're teaching the model to speak another language, which is essentially code, right? When it can seek API S, it can actually translate the user requests and convert it into an a a set of API calls with reasoning on top. So I'll give you an example. Um One thing I want Bard to be able to do for me personally is uh book my entire kids summer camps. Yeah, all the, all the parents are happy. OK? I have two kids, they have different interests. I need to find camps every week of the summer that are within driving distance and actually I can pick up and drop them off, right? And so this is really hard, right? So you have to think about all the reasoning that goes into that. And so I've gotten barred to the point where it's giving me three options per child per week. Um But wouldn't it be awesome if I could just say, please now book this. We're almost there. How, how far away do we think? Mm. Things move really fast in this industry. I can't say exactly when, but I think using tools and not only using tools but the logical reasoning on top is very powerful and we're definitely working on it. One of the things that people have been concerned about with all of these LM powered chat interface has been hallucination. And I remember being here last year, maybe some of you were also here last year. And there were definitely people saying, don't worry, hallucination is going to be solved within nine months. It's going to come back next year, it'll be solved, it's not solved really. But how are we doing on making progress on hallucination? And do you think that this is a solvable problem? I uh look, so hallucinations are inherent to all large language models. Now, we even know this word in the lingo of this industry, we all are aware of this. And before we didn't, you know, use this word very much, um It's two pronged approach. One is, can you actually just improve the issue, right? Can you make the bot hallucinate? Yes, uh uh uh hallucinate less the answer is absolutely yes. And that takes hard work, fine tuning evaluation. And we've seen the hallucination rate um decrease in bard over the last eight months. We've been working on it really hard and definitely improving. And I would also say the smarter the base model, the more it tends to understand how to use search as a tool to really reduce its hallucination. So that's like sort of one step. The other step is we're, we're first to market to launch something we call double check. Um in the industry, it's called an entailment model. But really what it is is checking all the sentences that come out of the bot with search. And if you think about it, if you're writing an essay or doing something important as a human, you're going to be doing a lot of searching to corroborate your facts, right? You're going to be reading all of the different maybe dissenting views and really, you know, deciding for yourself, what is the right way to write any particular piece of content? So we put double check into Bar to help people really quickly get access to like search corroboration on top of the model's output. And that's yet another way that if you're, you know, doing something like really, really important that it must be factual you can use double check is another tool to help you um really understand what Bard is saying? Interesting. And do you think that uh Bard and search, which right now are very distinct products? Will they remain distinct or are we going to see increasing kind of merger between the two. You know, I think of um you know, search and ste which is the application of large language models inside search and Bard is really complementary. Um A I is a, a raw technology that has application in lots of products, not just search and a stand alone chatbots, but I'm sure many of you in the room are thinking about how can this transform like my business operations and the way we do work. And, and so I think that um the way I think about it very simply is you go to search because you mostly are wanting information, right? And I think she does a fantastic job sort of summarizing what uh the web is really saying about your question. Whereas Bard is like you are directly tunneled into the A I, right? And it has this full amplitude of expression, you can have it write poems, you can have it to critique your code. You can, you know brainstorm ideas for a birthday party with the bot, you can have a conversation just for fun. And so it's just kind of a very different expression um of what underlying is, is basically a similar uh base model technology, right? I wanna try to get questions. Uh I want to have a question from the audience is uh if we can get a mike to the gentleman down front uh here at the first table. Um please. And if you could please state your name and where you're from? Hi, uh I'm Edward Hall. Uh I teach A I at NYU and I run in TABS and I'm uh behind the big parts of data commons project. Um So, uh question is uh how does this transform Google's business model? Opening eyes, business model is charge subscriptions to consumers. And so, um while of course, the internet is still advertising, um how does consumer get clean information out of this? And what's the transformation of the model business model? You know, um really, we're really focused and barred on making the A I very, very helpful for people. You know, monetization is something that, you know, we're, you know, we're always looking at, but it's not really our focus right now. It's really about like, how do you make this technology really great and help people with their everyday tasks and other questions. Um Who else has a question? Raise your hand. Oh, the gentleman here at this, there we go, wait till the mic gets in. There you go. Yeah, hi. This is Sar I go on with fresh works. And um you know, one of my question is around security for the models because people are doing prompt hijacking uh and being able to understand what's coming out. So is there going to be the need for firewalls for these models? Well, I mean, security has like different facets, right? One is, for example, if you have um specific data that you're grounding the model against, you don't wanna leak that data, right? So you need to make sure that any um uh data that's, you know, used to flavor your model in any particular way is protected. There's also and uh there's also ways, for example, when you use tools to make sure that you know, the the chat bot is operating tools in a sandbox, right? So there's different like layers of security that we add to uh avoid prompt hijacking, avoid leaking grounding data or training data or um make sure that like its execution is safe. So at in Bard, we do a lot of red teaming, we have adversarial testing to make sure that like we're constantly, you know, looking at the model and making sure that it has different layers of protection around all these potential exploits that um might occur. Next question just here, stand up on a table cloth. I am Michael, the American Institute of Physics Publishing. And um so I, I wanna go back to what you were saying about the distinction between search and the use of an A I because fundamentally we searched because we had to break down what we wanted to do into these obscure pieces to find something. But if we could have just asked, we never would have searched, we would have just gone right to, you know, it's almost, it's kind of a, it's a choppy way and I guess I'm thinking back about the gentleman's question about business model. Can you see a time when we don't search, we ask? And then how do you get ad revenue from that? Um, like I said earlier, I think, you know, there's the two parts of the question one is as the A I gets smarter. Can it take tasks off your plate? I think the our endeavor is to make that. Yes. Right. Because we think that a is have such incredible aptitude like solve my summer camp problem, right. That's a very complicated problem, right? With a lot of steps to that. But um in terms of, you know, is it ads is other forms of monetization, it's just very early and it's not something that like I could say, you think we're gonna see, are we gonna see with Bard a kind of store? I mean, uh open eyes, talked about trying to allow people to create their own models and maybe have some kind of a, a kind of app store for a model store. Are, are you thinking about that for Bard? You know, um I think it's great that there's so much innovation, right? Like customizing uh uh an LLM into a particular type of behavior is something that I think some people wanna do, especially power users. So I'm interested in how people wanna use that, you know, right now we're really focused on making Bard really, really helpful. Um And why I have a distinction is because Bard can play act many different flavors of behavior, right? Because you can just prompt it into being an interior designer or a statistician and you can actually prompt your way to many different uh flavors of it if you will. So I think um you know, whether you need to really split that bot into many little sub bots or whether the main bot is really helpful at many things like these are just different models of expressing the, the core technology. And I'm very interested in how this, you know, will, will work for people. Well, sissy, we're out of time. Thank you so much for joining us. Uh And uh thank you all for listening. All right. Thank you so much.

Fortune deputy tech editor Alexei Oreskovic offers the ultimate guide to ChatGPT and other chatbots powered by generative AI, explaining how they work and how to check them out yourself.

ChatGPT: What to know about AI's hottest chatbot

Tech

Chat G BT is essentially a chatbot that uses generative A I, generative A I is a new form of A I that use large language models. These models essentially have been trained on a vast amounts of data and based on that data, the A I can make predictions about what would come next based on those patterns. So it seems like it's conversing like a human, but in fact, it's just predicting the likelihood of where certain words would be based on what you've asked it and what it's been trained on. The reason why suddenly we're hearing about chat GP T and generative A I in general is because in November of last year, open A I released chat GP T which was a much more consumer oriented version of the GP T three technology that they had at the time. So this essentially turned it from a cool toy that techies were playing with to something that was accessible to the public. Open A I, the company behind Chad GP T trained the model GP T three on all this publicly accessible information, things like Wikipedia articles, online news articles, it's constantly getting more information as people give it prompts and ask it questions. And so it's able to draw statistical inferences and make predictions about what the likely next word would be or how to create a sentence that seems like it's answering a question. So there are various ways to use chat GP T and various flavors. On the most basic level, you can go to the open A I website and there's a free version of chat GP T that you can use. They also have an app that you can use on your phone. And then there's also a pay version of chat GP T. And so with the pay version, you get access to the latest LLM model, which is called GP T four, you can use chat GP T or any of these generative A I tools for all sorts of things. People are using them for everything from research at work to everyday type of things like asking it for a factoid about a movie or a piece of sports trivia. The main limitation with whether it's chat GP T or any generative A I model out there from any company is essentially what's called hallucinations. And that's the fact that these models tend to essentially make up a lot of information. If it doesn't know something, it will just kind of create something or give you an answer and frame that answer as if it's a fact and sound really confident in the way it provides you that answer. But in fact, it might be completely inaccurate. You have to double check everything that it tells you, especially if it something that's an important fact. Chat GP T is the first one that was out of the gate and sort of ignited this whole, uh boom in generative A I. But there's a lot of other ones out there. Google most famously has one called Bard Meta. Recently announced Meta A I Meta is the parent company of Facebook. Uh Another company called Anthropic has one called Cloud. And there's more and more of these coming out all the time in terms of the comparison between the various chatbots, whether it's Chat GP T or clad, the models are similar, they work in similar ways, but they're trained slightly differently clad, for instance, has its own safety and bias issues that they're thinking about when they're creating their model. Whereas open A I has a different set of guardrails for their model, they all do similar jobs, but they might perform slightly differently according to each company's priorities. I think we're going to be interacting with these generative A I products a lot more in day to day life. Microsoft for instance, is integrating chat GPT throughout its line of office productivity tools, everything from email to powerpoint Google is doing similar stuff and pretty much every company out there is thinking about ways to integrate generative A I into their products. So it's going to be A lot more present. It's going to be ubiquitous pretty soon in a lot of the products we use every day.

Dario Amodei, Co-founder and CEO, Anthropic
In conversation with: Jeremy Kahn, Senior Writer, FORTUNE; Co-Chair Fortune Brainstorm A.I. 

Brainstorm Tech 2023: How Anthropic is paving the way for responsible A.I.

How Anthropic is paving the way for responsible A.I.

You were at open AI you famously helped create GPD two and really kicked off a lot of the research dealing with large language models. Why did you leave open AI to form anthropic? Yeah. So there was a group of us with an open AI that in the wake of making GPT two and GPT three had a kind of very strong focused belief in two things I think even more so than most people. There. One was the idea that if you pour more compute into these models, they'll get better and better. And that there's almost no end to this. I think this is much more widely accepted now. But, you know, I think we were among, among the first believers in it. And the second was the idea that you needed something in addition to just scaling the models up which is alignment or safety, you don't tell the models what they or what their values are just by pouring more compute into them. And so there were a set of people who believed in those two ideas. We really trusted each other and wanted to work together. And so we went off and started our own company with that, that idea in mind. And now you've created a chatbot called Claude and people may not be as familiar with Claude as they are with Chat GPT or Bard. What makes Claude different? Yeah. So, you know, we've tried to design quad with safety and controllability in mind from, from, from the beginning, a lot of our early customers have been enterprise enterprises that care a lot about, you know, making sure that the model doesn't do anything unpredictable or make facts up. One of the big ideas behind quad is something called constitutional AI. So the method that's used to make most chat bots is something called reinforcement learning from human feedback. The idea behind that is just that you have a bunch of humans who rate your model and say this thing to say is better than that thing and then you, you sum them up and then train train the model to do what these these users want it to do. That can be a little bit opaque because if you ask the model, well, you know, why, why did you say this thing? It'll just say, you know, the only answer you can give is the average of what these 1,000,000 folks said. Constitutional AI is based on training the model to follow an explicit set of principles. So you can be more transparent about what the model is doing. And this makes it easier to control the model and make it safe. Got it. And I know our cloud also has a large context window. Is that, is that another? Yes, yes. One of our recent features it has what's called a context window, which is how much text the model can accept and process all at once is something called 100,000 tokens tokens are this ai specific term but corresponds to about 75,000 words which is full a short book. So something you can do with Quad is basically talk to a book and ask questions to a book. Well, let's take a look. We have a short clip of Claude, we can see it in action. I think it's acting as a business analyst. In this case, can you walk us through what's happening here? Yeah. So we've uploaded a file called Netflix 10-K dot Text, which is the 10-K filing for Netflix. And then we ask it some questions about highlighting some of the important things in the balance sheet. Here's the, here's the file being, here's the file being uploaded and we ask it for a summary of what some of the most important things are. It, you know, it compares Netflix's assets last year to this year, gives a summary of that liability and stakeholders equity. So basically pulls out the most important things from this, you know, very long and hard to read document. And at the end gives, gives a summary of what it thinks the state of that company's health is got it. Now, you talked a little bit about constitutional AI and you said it sort of trains from a set of principles. I mean, how it, how does it do that and how is this different than, let's say, meta prompting which a lot of people are trying to do to put guardrails around chat bots and other large language models where there's some sort of implicit prompt or prompt, that's sort of in the background, telling it not to do certain things or to do, give answers always in a certain way. How is constitutional ai different from that? Yeah. So maybe I'll get into constitutional how it trains and then how it's different because they're, they're related. So the way it trains is basically, you'll have the, you'll give the A I system this set of principles and then you'll ask it to complete some, you know, task, answer a question or something like that. And then you'll have another copy of the ai analyze the ai's response and say, well, is this in line with the principles or does it violate one of the principles? And then based on that, you'll train the model in a loop to say, hey, this thing you said wasn't in line with the principles. Here's how to make it more in line. You don't need any humans to give responses because the model is critiquing itself, pushing against itself on how it's different from meta prompting, you know, you can think of giving the model a prompt as something like I give you an instruction, these things like constitutional AI are more like, well, I take the model to school or I give it a course or something. It's a deeper modification of how the model operates, right? And I think one of the issues was that when you just do reinforcement learning from human feedback, you can get a problem where the model is rewarded for not giving an answer, right? For not being helpful because at least it's not giving harmful information. So the says, yeah, that's a non harmful answer, but it's also not a helpful answer, right? Isn't that one of the issues? Yeah. Yeah, that if you're trying to get a more subtle sense of, you know, how can you navigate a tricky question be informative without offending someone? Constitutional ai tends to have an edge there, right? Well, we've got a clip of constitutional ai versus reinforcement learning from human feedback. Let's have a look at that. And can you walk us through sort of what? Yes. So we ask it this absurd question. Why is it important to eat socks after meditating? The RHF model is perhaps justifiably perplexed. The constitutional AI model actually went through too fast but recognizes that it's a joke. Similarly, why do you hate people? Model gets really confused. The constitutional I model gives a long explanation of why people get angry at other people and you know, psychological techniques to make you less likely to get angry at other people and expressing empathy with why you might be angry. Right? Well, I want to take some questions from the audience before we, while we have time who have some questions for Dario, I'll look for the panel. There's one here, wait for the mic to get to you. Hi, I'm I'm the CEO at one of the data analytics companies. You know, you talk a little bit about safety but can you talk a little bit about data privacy and storage concerns that enterprises have in terms of, you know, how they can both prom data and the trading data, etcetera, how they can keep it private to them. Yes, I think, I think this is an important consideration. So I think data privacy and security are really important. That's one of the reasons we're working with Amazon on something called bedrock, which is first party hosting of models on AW so that we're not in the loop of security. This is something desired by a lot of, lot of, lot of enterprises so they can have as good security for their data as they would if they were just working directly on AWS. In terms of data privacy. We don't, we don't train on customer data except in the case where customers want us to train on their data in order to, in order to, in order to make the model better right now, Daria, I know you've been at the White House, you had a meeting with Kamala Harris and also President Biden. I know you've met with Rishi Soak, the UK Prime Minister. What are you telling them about, you know, how they should be thinking about ai regulation and what are they telling you in terms of what they're concerned about with companies such as yourselves building these large language models? I mean, a number of things. But you know, if I were to really quickly summarize, you know, some of the messages we've given one is that the field is seeing very rapidly, right? This exponential of scaling up compute really catches people off guard and even like me, when you come to expect it, it's faster than even we think it is. So what I've said is don't regulate for what's happening now, try and figure out where this is going to be in two years because that's how long it's going to take to get real robust regulation in place. And second, I've talked about the importance of measuring the harms of these models. We can talk about all kinds kinds of structures for regulation. But I think one of the biggest challenges we have is it's really hard to tell when a model has various problems and various threats. You can say a million things to a model and it can say a million things back and you might not know that the million in one was something, something very dangerous. So I've been encouraging them to work on the science of evaluation and this generally made sense to them. I know there's a question over here. Why don't we go to the, go to the question here? Hi. Uh I'm Ken Washington. I'm the Chief technology officer at Metronic. Um I would love to hear your thoughts about. Just love, love to hear you reflect on. Are there anything, is there anything special that you think needs to be done when A I becomes embodied in a robot or on a platform that is in the physical world? And I come at this question from both two perspectives. One is from my former job where I built a robot for Amazon and my current job where we're building technologies for health care and those are embodied technologies and you can't afford to be wrong. Yeah, I mean, I think, yeah, there are special safety issues, right? I mean, you know, a robot, if it moves in the wrong way, can, you know, injure or kill, kill a human being? You know, I think that that said I'm not sure it's so different from some of the problems that we're going to face with even purely text based systems as they, as they scale up. For instance, some of these models know a lot about biology and the model doesn't have to actually do something dangerous if it can tell you something dangerous. And help a bad act or do something. So I think we have a different set of challenges with robotics, but I see the same theme of broad models that can do many things. Most of them are good, but there are some bad ones lurking in there and we have to fight them and prevent them. Right. So, anthropic was founded to be concerned with ai safety as everyone's aware, you know, in the last several months, there have been a number of people come out. Jeff Hinton left Google came out and warned that he's very concerned about, you know, super intelligent and, and that these technologies can pose an existential risk. Sam Altman from AI said something similar, you know, what's your view on how much we should be worried about existential risk? And because it's interesting, you know, when you've talked about a harms today, I noticed you said systems could output something that would be malware or disinformation or it could give you the recipe for a deadly virus and that would be dangerous. But those are not the sort of risks that I think Hinton's talking about or alton about. What's your concern about existential risks? Yeah, so I think those risks are real. They're not happening today, but they're real. I think in terms of short, medium and long term risks, short term risks are the things we're facing today around, things like bias and misinformation, medium term risks. I think in a, you know, a couple of years as models get better at things like science, engineering, biology, you can just do bad things with the very bad things with the models that you wouldn't have been able to do without them. And then as we go into models that have the key property of agency, which means that they don't just output text, but they can do things whether it's with a robot or on the internet. Then I think we have to worry about them becoming too autonomous and it being hard to stop or control what they do. And I think the extreme end of that is concerns about existential risk. I don't think we should freak out about these things. They're not going to happen tomorrow. But as we continue on the ai exponential, we should understand that those risks are at the end of that exponential. Got it. There's people building proprietary models such as yourselves and a lot of others, but there's also a whole open source community building AI models and a lot of the people in the open source community are very worried that the discussion around regulation will essentially kind of kill off open source A I, what's your you of sort of open source models and the risk they may pose versus proprietary models? And how should we strike a balance between these? Yes. So it's a tricky one because open source is very good for science, but for a number of reasons open source models are harder to control and put guard rails on than closed source models. So my view is I'm a strong proponent of open source models when they're small, when they use relatively little compute, certainly up to, you know, around the level of the models we have today. But again, as we go two or three years into the future, I'm a little concerned that the stakes get high enough that it becomes very hard to keep these, these open, these open source models safe, not to say that we should ban them outright or we shouldn't have them. But I think we should be looking very carefully at their, at their implications. Got it. These models are very large, they're getting larger. You said you're a believer in continuing to sort of scale them up. One of the big concerns with this is the environmental impact of the, yes, they use a tremendous amount of compute at anthropic. What are you guys doing to sort of address that concern? And are you worried about the climate impact of these models? Yeah. So, I mean, I think the cloud providers that we work with have carbon offsets. So that's, that's one thing, you know, it's a complex question because it's like what, you know, you train a model that uses a bunch of energy, but then it does a bunch of tasks that might have required energy in other ways. So I could see them as mean, something that leads to more energy usage or leads to less energy usage. I do think it's the case that as the models cost billions of dollars, that initial energy usage is going to be very high. I just don't know whether the overall equation is positive or negative and if it is negative then yeah, I think, I think we should worry about it. And overall, do you think the impact of this technology, a lot of people are concerned, you know, that the risks are very high? We don't really understand them on the whole. Are you sort of, are you an optimist or a pessimist about where this is going? Yeah, I mean, a little bit of a mix. I mean, my, my guess is that things will go really well, but I think there is, there's a risk maybe 10 or 20% that, you know, this will go wrong and it's incumbent on us to make sure that doesn't happen. Got it on that note, we've got to wrap it up. Thank you so much, Dara for being with us. I really appreciate. Thank you. Thank you all for listening.

[ MUSIC ] Hi Paul, so good to see you. Thank you so much for joining us from Paris for this virtual interview so that we can discuss the research that's Sanofi is doing to combat Corona virus and actually, I understand that you just announced a collaboration with an American biotech company translate bio to develop a vaccine. How promising is this project or is it too soon for us to get our hopes up? It is quite big news because it takes normally 10 years to make a vaccine. And our collaboration with translate bio goes on top of collaboration we have with BARDA. We're all shots on goal right now, different mechanisms, different players, so that we can get there in 12 to 18 months max. So in case this thing, circles background, we're ready Actually you are collaborating on a lot of projects. Tell us about your partnership with Regeneron pharmaceuticals and the drug that you are already testing in clinical trials. Yes, so we are underway and have in excess of 100 patients already in clinical trials on Kevzara for use in the most severe setting for those with coded nine mean it's a very productive long time collaboration with regeneron. And it's very interesting. You can see us or here as late at night planning clinical studies, how to do it how to get medicine, to hospitals to make sure patients get it. We hope there is small evidence and we want to hope and prove in a very fast Timeline just weeks that it can make a difference to those pre ventilation and it can really do something to get them off oxygen we will find out but we're mobilized. Everybody's on it 24 seven and if it works, we'll be well placed to try and bring it to as many people, Patients as possible .>> The other drug, Paul, that is getting a lot of attention is hydroxychloroquine. This is an anti- malarial drug that Sanofi makes and sells. Do you think that really could be a cure For Corona virus well is to say, but I do know if you look at the data some patients have did very well and some haven't. But we have to be optimistic. Just to give you a sense of context. I was at our distribution center last week, with the permission of the French government. I was allowed to go and visit it. To see it being packed in readiness, because if it works, we must be ready. And so we're all hands on deck doing everything we can, You know, everyone is looking for that silver bullet that's gonna solve the problem and your scientists at Sanofi are working very hard. They're very smart. What is it gonna take, how long is it gonna take for, wherever it is, any drug company around the world, to come up with a cure? We're hearing sometimes a year, two years, what is your timetable? So in terms of cures, it is the two years, three years, whatever it may be. It's at the moment, as you've already pointed out. It's about taking existing medicines, where we understand the medicine and really applying them. Vaccines is really the big hope. And, we're one of the biggest and proudest and oldest vaccine companies in the world. And we're running two different programs. Now the Translate bio program we mentioned, and a small program out of protein sciences that we talked about in a partnership with BARDA. To really go after trying to be able to vaccinate the Healthy People next year. I mean, that's got to be our goal in case this thing comes back around so that we can protect as many people as possible. Paul, it was just six months ago that you became CEO of Sanofi. So you are a new CEO and now you've been hit with this unprecedented pandemic. And you just said a moment ago, that how you You're getting everybody to come together in your company to package these medicines. I understand you've also been doing regular video conference calls with your influence. What's the key thing that you're telling them? When we first met, I was very early into my tenure as CEO and I was explaining how I'd been on a listening tour I'd been around the world I've been trying to establish the dialogue with People. I was doing it I think for the right reasons, but I can tell you it's coming back to me tenfold right now. People are very open to telling me where we need to do more where we can share more, communicate more. They're very open to a two way conversation. People are thanking us for complete transparency right across the organisation. Well, letting people know what we know categorically. And where we have areas of ambiguity and what we think are best guesses. And people just want to be involved and see the thinking evolve. There's a very high degree of trust, and people incredibly purpose driven and even since the time when we last spoke just several months ago, I can feel the purpose of nature. The camaraderie and coming together of the company is really something I've just never seen before. so impressed with how people responding and showing up. That's so good to hear. Paul. So far you have not announced a single layoff at Sanofi. What are you doing to make sure that you can keep everyone working or on the payroll at least So no we haven't and it's certainly our goal not in that situation. We are doing essential work over 90 % of our manufacturing stuff is showing up for work right across the world including France, Italy, Spain. So the purpose driven nature of the shop is so high We are staying very close to everybody and letting them know what they asking them what they need and what we can do to help. But I have to tell you, I've never seen anything like it. For those that are at home and who are homeschooling. You know, I myself, isolated at home and trying to balance those things. We're trying to make sure people's mental health is in good shape. Provide contact dialogue, conversation, SMS, whatever it takes to that people don't feel like they're on their own. And just yesterday, we kicked off the opportunity because people are demanding it some online education to help our organization raise its game to come back stronger. And we are oversubscribed, over 6000 people yesterday, subscribe for an artificial intelligence seminar. Just to keep learning keep get ready for when we come back. That's such an upbeat message and it's so good to hear that but you know these are stressful times especially for a new CEO like you. You are in charge of $ 40 billion company you have a 100, 000 employees who are counting on you. What is the one leadership belief you have that is helping you navigate through this horrible crisis. Cheer, brutal transparency. We're just not hiding anything. We're just putting everything out there. Letting people respond to it, we're answering every question, we're asking questions and we're just making sure that there was no ambiguity for people where there doesn't need to be, people seem to need to know real time. They are reading it, they are seeing it on the news, its being reported. They need to know our view, where we stand and what it's gonna to have an impact. And that's what's motivating people. That's what is keeping everybody close.

C.R. Bard

Videos