Loading summary
Transcript32 lines
- [00:00]
Magda Petwarden
AI's growth is taking it from centralized systems into everyday workflows, and the enterprise endpoint is taking on a new role and new risks. At the break, AMD's Magda Petwarden will discuss how enterprises are rethinking security to protect Today's AI enabled PCs.
- [00:19]
Isabel Busquette
Welcome to Tech News briefing. It's Friday, March 13th. I'm Isabel Busquette for the Wall Street Journal. Today we're looking at how AI generated digital versions of humans are changing the for market research, companies like CVS say they are using these simulated people to get feedback on products and store layouts cheaper and faster than ever before. And then since OpenAI reached a deal with the Pentagon to deploy its AI models in classified settings, at least two of its high level employees have resigned citing values and principles. We're diving into how that's shaping the latest wave of the battle for AI talent. But first, the latest job AI is taking over is market research. A handful of companies are using AI generated digital clones of real people to replace the lengthy and expensive process. Essentially a shortcut instead of going the human route. It works like this. Real people participate in in depth interviews and share data, including their purchase histories data to train AI agents who then become their digital twins. Companies can ask infinite questions of these AI clones, making the work of customer research cheaper and faster. But does it really work? Wall Street Journal Enterprise Tech reporter Bell Lin joins us for a breakdown of how it's all going bel what kinds of companies are interested in these AI simulated people and what are they doing with them specifically?
- [01:57]
Bell Lin
Yeah, so CVS is one example in which they use these AI simulated people for things like market research, for getting a better sense of what their customers are thinking and feeling. The head of their enterprise Insights group told me that they discovered that for their pet rx, their pet medication, that through talking with their AI simulated people that essentially people don't consider administering pet medicine a chore. And so that affirmed that people have this deep emotional connection with their pets and they're also using AI simulated people to do things like test store layouts. So it's replacing what they would traditionally consider using panel research for and instead they're able to use these AI simulated people to give their thoughts and feedback. And then also Gallup, which is a historical well known polling company, they are launching a partnership with Simile in which they have a joint offering to provide polling services to their joint customers. So you can imagine polling in areas like well being and mental health, job satisfaction. Those are some of the areas they're
- [03:02]
Isabel Busquette
Starting with, so what's the upshot here? Why not just do traditional market research?
- [03:07]
Bell Lin
Traditional market research is a huge industry. It's roughly over $150 billion. The whole realm of market insights, but it's very cumbersome. You have to do the work of recruiting and collecting the people who are representative, sort of randomized sample of the population. You're trying to. And then people get tired. We can only handle so many things at once, maybe five different scenarios that are presented to us and we have to take breaks. But these AI agents, these AI simulated people don't. And that's what companies like CVS told me, that they can reach so many more people with thousands of potential scenarios in mind, and they don't get tired. So they can keep asking questions again and again and again. And then also there's an interesting idea that AI simulated people can have some benefits for a healthcare company like CVS because they're able to simulate hard to reach populations like people with chronic conditions.
- [04:05]
Isabel Busquette
But how effective is this actually compared to traditional market research?
- [04:09]
Bell Lin
Companies like CVS have told me that they're able to replicate what their agentic twins have found compared to what their real people have told them with 95% accuracy. And that's really high for an area like market research. And so there's a lot to be gained from essentially finding and creating a great digital twin of a person.
- [04:30]
Isabel Busquette
Are there any risks, though, to relying on the opinions of AI people rather than real people?
- [04:36]
Bell Lin
Absolutely. So the companies that I talk to are also concerned about AI's hallucinations about model drift. And so they've started to do things like build in some guardrails. And similarly, the startup that I talked to that's offering this service does the same thing where they filter out for harmful content or sensitive content. And so there's caution being applied to this area, but at the moment feels that there's more enthusiasm than caution because there's so much to be gained from being able to query your AI people indefinitely versus having to work with humans who are slower and have to take breaks and get tired.
- [05:10]
Isabel Busquette
So where is all this going? Is this the end of traditional market research?
- [05:14]
Bell Lin
The industry is quite slow and relies on these consulting companies, these traditional market research companies that sort of have a stronghold over the market. And so it is an industry that is very ripe for disruption. But at the same time, the companies I talk to are saying that they'll continue to talk to their customers, that they'll always have the voice of the real human in mind. And so that they'll augment with the voice of the agentic twins.
- [05:42]
Isabel Busquette
That was Wall Street Journal reporter Bell Lim. What do you make of digital twins? If you're a listener on Spotify, leave us a comment with your thoughts. Coming up, why some top OpenAI employees are resigning after the company's deal with the Pentagon and what it signals about the latest wave of the fight for tech talent. That's after the break.
- [06:10]
Magda Petwarden
AI PCs are increasingly where sensitive data is acted on in real time. How should enterprise security leaders address that? Here's Magna Petwarden, corporate vice president of product management client business unit at AMD, with some thoughts.
- [06:23]
Magna Petwarden
People are going to use AIPCs, and enterprises have to think about AIPCS as a natural extension of their zero trust policies. They have to embrace this idea of security that is distributed versus centralized with sort of a moat built around it because there are no more moats.
- [06:42]
Isabel Busquette
Last summer, the AI talent war hit a new peak as Meta tried to poach top talent with sky high salary packages. Now it's entered a new phase thanks to a recent showdown in Washington. Anthropic was renegotiating its deal with the Pentagon and looking to weave in new safeguards with when talks broke down, prompting the Trump administration to declare the company a supply chain risk. Shortly after, OpenAI said it signed a deal with the Pentagon. News Corp. Owner of the Wall Street Journal, has a content licensing partnership with OpenAI. Anthropic's unique focus on safety and responsible AI has long made its culture attractive to top talent, and that has some employees at rival firms making moves.
- [07:26]
Bell Lin
Now.
- [07:28]
Isabel Busquette
WSJ's Megan Bobrowski joins us to break down the latest. Megan, how has this showdown with the Pentagon impacted the talent war between OpenAI and Anthropic?
- [07:38]
Megan Bobrowski
We've seen at least two OpenAI employees announce decisions to leave the company in the last two weeks. One of those employees specifically said that they were leaving to join Anthropic and that they were very impressed with Anthropic's values. One woman who was sort of in charge of the hardware efforts for OpenAI's robotics team, she said she felt that the contract that OpenAI reached with the government shouldn't have happened so quickly and that the things that they agreed to, you know, deserved more time and consideration and to talk through first.
- [08:15]
Isabel Busquette
So overall, how has Anthropic's approach to talent and retention differed from other AI companies?
- [08:22]
Megan Bobrowski
Anthropic hasn't tried to match the incredibly high salaries that Meta offered last summer. OpenAI offered bonuses in kind of response to that situation. And Also in December, OpenAI got rid of its vesting cliff, meaning that employees can have access to their stock immediately without joining. Typically at a company, there's a period you have to wait. So at Anthropic, that period's a year.
- [08:51]
Isabel Busquette
So Anthropic is not matching these insanely high salaries. But at the same time, you reported that its retention rates still really stand out in the industry. What is it about the culture there that's so uniquely compelling for researchers?
- [09:06]
Megan Bobrowski
Yeah, so if we remember Dario Amodai, he actually left OpenAI to found anthropic and said that it was because of safety reasons. He didn't feel like OpenAI was taking safety into enough consideration when they were building these models. And so Anthropic has always sort of had this safety slant. It's similar to OpenAI, but it has this sort of what Dario calls a values driven culture. And so when the Meta, you know, blitz was happening last summer, he said that people could leave if they wanted to, but that anyone who was at Anthropic was there solely for the mission.
- [09:45]
Isabel Busquette
And beyond this talent question, how else has the Pentagon issue impacted public perceptions of these companies? And what do the companies themselves have to say about it?
- [09:56]
Megan Bobrowski
So Anthropic is certainly winning in the court of public perception. Downloads of Claude, its chatbot, were way up last week, even surpassing OpenAI for a bit. You had people supporting them on social media. OpenAI has sort of had to do a lot of damage control, and specifically its CEO, Sam Altman, over the weekend spent time answering questions and concerns and then internally hosted a Q and A with employees, sort of talking about everything that happened and how they would have done things differently.
- [10:29]
Isabel Busquette
How has the talent more evolved over the last year?
- [10:33]
Megan Bobrowski
The talent wars just continue to be an incredibly fascinating thing to follow. Last year it was a lot about money, and Meta was offering that. This year so far, I mean, it's still partially about money. But there's also this, this idea as AI gets closer to having more and more real world impacts, how the technology these engineers and researchers are building increasingly becomes a part of their decision and where they want to work. It's not just little things that exist in a theoretical box off nowhere anym, like these things are being used in wars. They're completely disrupting the stock market. Like increasingly, you're seeing AI get integrated into daily life. And so the decision of where to work and what they do just carries more weight than ever right now.
- [11:18]
Isabel Busquette
That was Wall Street Journal reporter Megan Bobrowski. And that's it for Tech News Briefing. If you're a listener on Spotify, be sure to leave us a comment. Today's show was produced by Julie Chang. I'm your host. Isabel Busquet, Jessica Fenton and Michael Lavalle wrote our theme music. Our supervising producer is Katie Ferguson. 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.
- [11:57]
Magda Petwarden
How can enterprise security leaders protect AI PCs? Here again is AMD's Magna Pet Ward.
- [12:03]
Magna Petwarden
You have to think about security as foundational and not something that comes after the fact and that has to be anchored in hardware. Hardware protecting software is more superior from a security strategy perspective and creates sort of an immutable trust even before the operating system drivers and AI models ever load. These protections really create the root of trust for AI enabled workloads so that they can confidently run. And as AI moves onto the device, then the more sensitive data and decision making shifts closer to the hardware and that means that software only is no longer sufficient.
- [12:40]
Magda Petwarden
Learn more about how amd protects AI PCs from silicon to software@amd.com this content
- [12:47]
Isabel Busquette
was created by Custom Content from WSJ,
- [12:49]
Bell Lin
a unit of the Wall Street Journal, advertising.