Loading summary
Transcript20 lines
- [00:00]
Comcast Business Announcer
With leading networking and connectivity, advanced cybersecurity and expert partnership. Comcast business helps turn today's enterprises into engines of modern business. Powering the engine of modern business Powering possibilities. Restrictions apply.
- [00:19]
Julie Chang
Welcome to Tech News briefing. It's Friday, October 10th. I'm Julie Chang for the Wall Street Journal. Artificial intelligence is transforming how companies manage risk. We'll hear from two executives about how they're leverag AI to navigate legal and regulatory challenges. And then Microsoft wants its AI chatbot to gain independence from ChatGPT, and it thinks healthcare is the way to do it. We've got the exclusive details on that effort. But first, advancements in AI promise to help companies mitigate risk. But what does that actually look like in practice? At the recent Journal House Singapore event, Dow Jones's Nicholas Elliott sat down with Chi Kin Lam, group head for legal and compliance at DBS bank, and Kelly Geer, managing director for legal and compliance in Asia, the Middle east and Africa. At FedEx, they talked about different AI use cases that their workplaces have already put into practice. Here's part of that conversation.
- [01:21]
Nicholas Elliott
Kelly and Chikin, welcome. Kelly, let's turn to you. So I know one of the things you mentioned to me was the using AI in investigations, which an important part of your job. Can you tell the audience how exactly you've made use of it in that way?
- [01:35]
Kelly Geer
The EMEA region for FedEx is a little over 100 countries and territories. So when you're conducting an internal investigation, you're dealing with a number of different languages. And what AI has helped us do is use recording interviews in investigations. And the transcription function has been particularly helpful. We also have a number of investigators across the region where English is maybe their second or third language. And so writing that investigation report in English is difficult. So our teams have saved so much time in going back and forth with the investigator, like, is this really how you want to say this? Are you sure this is the word? And so we started having them write the reports in their local language and then we translate it into English. And it's been phenomenal, the results. And then several years ago, the legal team created a chatbot. These are lawyers, they are not coders. And they got together and developed a chatbot that can answer questions related to our compliance policies. And so it directs them based on the question, the decision tree logic to go to that policy, and then by default, a SharePoint site. And we're hoping to improve on this with more AI in the future. And then my third example is external facing. We've recently developed A tool that helps our customers with the harmonized tariff schedules. If you live in Asia, you might wake up to a new regulation. You never when you're shipping your goods to to the US so we're really proud of this tool. It's on FedEx shipmanageredex.com and so whatever item you're shipping into the US specifically, you can describe the item in this chat bot within this tool will help you describe the item more accurately so that you can get the right harmonized tariff code and then give you a link directly to the US Government site. And of course, this helps with customs clearance for our customers so that their shipments are compliant and get through quickly and don't cause any delays.
- [03:26]
Nicholas Elliott
I did want to get on to the very important topic of agents. This is AI that has some autonomy. So you set tasks within certain limits and then let it go. It seems to be a very active, innovative frontier of AI, but it also carries quite a lot of risk. Chen, what are you doing in that area? In the area of agents, your AI.
- [03:48]
Chi Kin Lam
Governance policy probably needs to be able to distinguish between high risk use cases and low risk use cases. So a low risk use case, hey, would you like this for lunch? And I'll buy the lunch for you, that's okay. But filing a suspicious transaction report, you are effectively calling somebody a criminal. You really don't want the AI making that decision. So the AI policy that helps you to make those decisions at scale 46,000 employees has to be set up right? Proper examples, proper guardrails. Okay, now then, secondly, don't always call the LLM because you pay for that. Calling a token and calling tons of tokens and getting all your people constantly calling tokens. Your bill is going to be pretty high. So watch out for when you don't need the LLM when just an API or robotic process automation step would work. Don't Hyper Engineer prompt every single thing that you do. And if you have a prompt, then you do take the risk of that probabilistic calculation spitting out some garbage. Right? So don't take that risk if you don't have to.
- [04:56]
Julie Chang
That was Dow Jones's Nick ELLIOTT Speaking with DBS Bank's Chi Kin Lam and FedEx's Kelly Geer at the recent Journal House Singapore event. You can find the full chat@journalhouse.WSJ.com we put a link in the show notes coming up. Microsoft has increased staffing at an internal AI lab that competes with OpenAI. What will this mean for their partnership? That's after the Break.
- [05:29]
Viking Cruises Announcer
Committed to exploring the world in comfort. Journey through the heart of Europe on an elegant Viking longship with thoughtful service, cultural enrichment and all inclusive fares. Discover more@viking.com.
- [05:47]
Julie Chang
As we told you about in yesterday morning's Tech Minute, a new version of Microsoft's Copilot is expected to launch this month. Aiming for greater independence from OpenAI's ChatGPT. We report Microsoft is teaming up with Harvard Medical School to enhance Copilot's ability to provide reliable medical information. So chatting with it is more akin to speaking with a healthcare professional. Microsoft hopes this will make Copilot more competitive in the consumer market. WSJ reporter Sebastian Herrera broke the story. Sebastian, people already ask chatbots like ChatGPT all sorts of medical questions. So how exactly would Copilot be different?
- [06:29]
Sebastian Herrera
Well, Microsoft wants to have the most accurate and helpful medical advice, and that points people in the right direction. And there is room for improvement in the chatbot industry for this. And that's because there are plenty of examples of people using ChatGPT and getting bad medical information. So, for example, there's a study led by researchers at Stanford University from last year that said that out of a bunch of queries that they asked for medical advice, that they got bad advice about 20% of the time. And there's also been reported instances of ChatGPT fueling mental health crisis for people. So Microsoft wants to be the opposite of that, and they're hoping to make Copilot something that people go and use, and if they can figure out the medical part of it, it could be really successful for them.
- [07:20]
Julie Chang
Where does this leave Microsoft's partnership with OpenAI?
- [07:23]
Sebastian Herrera
So Microsoft says that the partnership with OpenAI is still really important to them, really valuable, and they do have this agreement in place to extend their partnership for years to come. We'll still see Microsoft and OpenAI be close partners. But as a big tech company, Microsoft is also working to build up its own AI models with the goal of not having to rely on OpenAI. Because right now a lot of their chatbot technology does rely on OpenAI. And so they want to change that. They want to have their own models be used within their chatbot. Their goal is to have Microsoft models replace workloads for Copilot, specifically to replace OpenAI workloads.
- [08:08]
Julie Chang
Yeah. One thing that caught my eye about your reporting was that Microsoft is increasing staffing at its own internal AI lab. Tell us more about that.
- [08:17]
Sebastian Herrera
What's been interesting with Microsoft's AI Lab, which they created just a year ago, is that a lot of people that have joined it are very new. So if you look at most people who have joined Microsoft AI, a lot of them have been in the past six to nine months. And sources that I've talked to that are familiar with the lab say that they're increasing headcount a lot. And we've also reported on a lot of new hires that they've gotten from Google DeepMind, which is also an AI lab. The CEO there at Microsoft AI, Mustafa Suleiman, he was formerly at Google and so he's been able to attract a lot of people from the Google ecosystem to come work at Microsoft AI. And they've recruited people from startups that are really experienced. They have clinicians on staff with some of this healthcare initiative.
- [09:05]
Julie Chang
Sebastian, what challenges is Microsoft up against here?
- [09:09]
Sebastian Herrera
They're up against a lot of challenges. First of all, trying to innovate in the healthcare industry is extremely challenging. And it's not the first time that Microsoft has tried to do something in the health space. They've tried to manage and organize health data for years and they haven't really cracked healthcare. A lot of companies haven't as well. So on one end, healthcare is a very tricky space to get into and they have a huge challenge in trying to innovate in that space. And on the other end, Copilot is very behind OpenAI and other players like Anthropic in terms of people using this product. Not to mention, not only is OpenAI their partner, but it's also their competitor. You know, Microsoft is known as a really big, great enterprise technology company, but they haven't really done a lot in the consumer space that's noteworthy. So by no means is it promised for Microsoft to succeed here in any way.
- [10:10]
Julie Chang
That was WSJ reporter Sebastian Herrera. And that's it for Tech News Briefing. Today's show was produced by me. Julie Chang, Jessica Fenton and Michael Lavalle wrote our theme music. Our development producer is Aisha Al Muslim. Chris Zinsley is the deputy editor and Falana Patterson is the Wall Street Journal's head of News Audio. We'll be back later this morning with TNB Tech Minute. Thanks for listening.
- [10:40]
Fiona Gregg
We've reached what's called the Silver tsunami moment when 4.2 million people will turn 65. Top of mind for them is avoiding shortfalls in retirement income. Here's Fiona Gregg, global head of investor research and policy at Vanguard, on some important steps to take.
- [10:55]
Unidentified Vanguard Speaker
You have these small pots strewn across the earth at these different employers. Social Security, how do they all add up? A first order opportunity is consolidate and simplify.
- [11:04]
Fiona Gregg
Get more insights from Vanguard on how you can navigate an uncertain market@vanguard.com.