Why Apple Intelligence Isn’t That Smart … Yet

[00:00:00] OCI is the single platform for your infrastructure, database, application development, and AI needs.

[00:00:06] Do more and spend less like Uber, 8x8, and Databricks Mosaic.

[00:00:11] Take a free test drive of OCI at oracle.com slash wallstreet.

[00:00:19] Welcome to Tech News Briefing. It's Friday, October 25th.

[00:00:23] I'm Zoe Thomas for the Wall Street Journal.

[00:00:25] A few years ago, some big tech companies made some big promises about reducing their carbon footprints.

[00:00:33] But now the artificial intelligence boom is requiring a lot of energy.

[00:00:38] And that's got companies betting big on nuclear.

[00:00:42] We'll tell you about the tech giants investing billions in nuclear power and why that investment is something of a gamble.

[00:00:49] And then, Apple is late to the generative AI space.

[00:00:54] But with Apple intelligence coming to iPhones with iOS 18.1, it's hoping to make a splash with users.

[00:01:02] Our senior personal tech columnist Joanna Stern spoke with Apple's head of software, Craig Federighi,

[00:01:08] about how the company is approaching bringing Gen AI to iPhones.

[00:01:12] We'll have highlights from their conversation.

[00:01:17] But first, Amazon, Google, and Microsoft have each struck recent deals meant to bring more nuclear power online.

[00:01:26] The companies are looking for clean sources of power to meet their growing energy demand tied to AI,

[00:01:33] which is threatening to derail their climate pledges.

[00:01:36] Here to tell us more is our reporter, Catherine Blunt.

[00:01:39] So, Catherine, how much have these companies' AI ambitions changed their energy needs?

[00:02:06] Are there new technologies here that are going to help?

[00:02:13] Can you help get nuclear online?

[00:02:15] Yes.

[00:02:16] So, Google, Microsoft, and Amazon have each made significant nuclear announcements in the last few weeks.

[00:02:23] Google and Amazon in particular have been targeting the development of what are known as small modular reactors, or SMRs,

[00:02:31] which are basically tinier versions of the big nuclear power plants that we have built historically.

[00:02:36] The idea is to scale down the size of the plants and other things.

[00:02:41] They become easier and less expensive to build.

[00:02:44] None have yet been commercially deployed in the U.S.

[00:02:46] So, there's a lot of hurdles that need to be overcome in order to get to that place.

[00:02:50] But to have the tech companies backing some of these sort of prospective projects is a really big deal for the SMR companies.

[00:02:58] And Microsoft made an agreement with the owner of the Three Mile Island nuclear plant to restart the undamaged reactor that was shut down several years ago.

[00:03:09] You're reporting that the tech industry's nuclear investment is something of a gamble.

[00:03:14] Why is that?

[00:03:15] Well, the thing about building nuclear plants in the U.S. is that it has historically been really hard.

[00:03:21] We have built a number of very big ones over the last few decades, but almost all of them, if not all of them, have taken years longer than expected and cost billions of dollars more than expected.

[00:03:33] And so, the challenge here is whether some of these hurdles can be overcome with these smaller models through these small modular reactors, the SMRs.

[00:03:44] What are the stakes for the tech companies?

[00:03:46] Each of the big tech companies years ago made pledges to substantially reduce their carbon emissions.

[00:03:52] And the sort of the AI race has kind of appended those pledges in the sense that the companies are now requiring so much more power than they thought that they were going to need.

[00:04:02] And so, the idea of aggressively pursuing nuclear power is really a concerted effort to write the course, so to speak, and get back on track to help reduce emissions.

[00:04:12] But in the interim, I mean, it's going to be a while before a substantial amount of new nuclear power is added to the grid, if ever.

[00:04:20] All right, that was our reporter, Catherine Blunt.

[00:04:23] Coming up, Apple intelligence isn't that smart.

[00:04:26] Yet.

[00:04:27] We'll find out about Apple's plans to improve Siri and how it's approaching privacy.

[00:04:32] That's after the break.

[00:04:43] Exchanges.

[00:04:44] The Goldman Sachs podcast featuring exchanges on rates, inflation, and U.S. recession risk.

[00:04:52] Exchanges on the market impact of AI.

[00:04:55] For the sharpest analysis on forces driving the markets and the economy, count on exchanges between the leading minds at Goldman Sachs.

[00:05:03] New episodes every week.

[00:05:05] Listen now.

[00:05:07] Companies have to consider a lot of things when launching a new generative AI product, like privacy, moderation, and competition.

[00:05:21] Our senior personal tech columnist, Joanna Stern, has been trying out Apple's first wave of AI features coming to iPhones with iOS 18.1 called Apple Intelligence.

[00:05:33] She spoke with the company's head of software, Craig Federighi, about it.

[00:05:37] Here are highlights from their conversation.

[00:05:39] Big picture.

[00:05:40] What excites you right now about AI and this new way of interacting with our devices?

[00:05:46] Well, I think interacting really gets down to it because you think about where we as humans, what are we working with?

[00:05:52] A lot of the times we're working with language.

[00:05:55] You know, we want to express ourselves through language.

[00:05:57] People are communicating to us in human language.

[00:06:00] And in the past, it was harder for our devices to really understand, in truth, what that language meant at a deep level.

[00:06:08] And so we've had all these things that we've wanted to do in product over so many years.

[00:06:12] I think as an industry, there are things that we wanted to do since the dawn of computers that have suddenly become in reach.

[00:06:20] And it moves from language even to imagery.

[00:06:22] And so this just opens so many doors to fulfill the kind of customer experiences that we've been excited about for a really long time.

[00:06:29] Apple is late to generative AI, but Apple's never really been first to a new tech product category.

[00:06:35] What is new with Apple intelligence?

[00:06:37] Well, several things.

[00:06:39] I mean, I think we looked at this as not how do we build another chat bot and bolt it on the side of our existing experience,

[00:06:48] but how do we create something that's deeply integrated, that's meeting you where you are in the experiences throughout the operating system.

[00:06:57] But I think most importantly, we view it as personal.

[00:07:01] I think the bulk of the focus in chat bots historically have been around products that don't know you at all.

[00:07:09] They're powerful and interesting, and they have some really fantastic uses.

[00:07:14] But I think the real power of intelligence as it pertains to how our customers would use our products are intelligence that understands you.

[00:07:21] And with that comes a great deal of responsibility because to do that, that intelligence has to traffic in information that your device, that you've stored on your device.

[00:07:31] And that means privacy is a tremendously important consideration.

[00:07:35] And so we've done a lot of things we can talk about at many, many layers of the stack to both enable that personalized experience,

[00:07:42] but to make it private in a way that I think has never been done before.

[00:07:46] One of the natural questions or concerns that I think people really do have with these AI tools and these AI models are,

[00:07:53] is my data being used to train these models?

[00:07:56] And so in the case of Apple's models, have you used personal information to train the models?

[00:08:03] No. Our foundation models are trained with no customer information.

[00:08:06] Your emails, your iMessages, your notes, your photos in iCloud, none of that goes into training.

[00:08:12] I want to ask about moderation.

[00:08:14] I wrote a short passage.

[00:08:16] I put in profanity.

[00:08:17] I put in Nazis.

[00:08:19] And it said, writing tools aren't designed to work with this type of content.

[00:08:23] Yeah.

[00:08:24] And then I was still able to sort of manipulate the text.

[00:08:28] Yeah.

[00:08:28] Is that a bug or is that sort of how you're thinking about approaching this?

[00:08:32] This is a deep issue because there's a real trade-off between customer agency and you could call it safety or, you know,

[00:08:48] the potential to create content that, you know, none of us are very comfortable of.

[00:08:53] And we're a personal computer company.

[00:08:55] And people have for decades used our tools to, you know, compose their deepest thoughts,

[00:09:04] to write research papers of all sorts.

[00:09:06] And so we felt ultimately we would err on the side of customer agency.

[00:09:10] You may be writing a research paper decrying the effects of Nazi ideology on, you know, something.

[00:09:23] And you may want to draw on writing tools to help proofread that.

[00:09:27] And if we said, hey, we're going to have nothing to do with that, well, you're actually trying to accomplish a positive thing.

[00:09:33] But there's, you know, some difficult material there.

[00:09:36] And so we're going to err on the side of you're the customer and we're going to let you do it.

[00:09:42] On the other hand, we are not training on a lot of information or fine-tuning on a lot of information to deal with some of those topics.

[00:09:50] And so our output is not necessarily going to do the best job with some of those ideas or terminology.

[00:09:55] And so we're going to tell you, you're the boss customer, but results may vary if you try to use us for this task.

[00:10:04] Want to shift to talking about Siri, a new, really cool, glowing look of Siri.

[00:10:09] It's cool. It's fun. The voice sounds more human.

[00:10:12] Yeah.

[00:10:13] But I've been testing Siri still, you know, modest improvements.

[00:10:17] Sure.

[00:10:18] Where is the smarter Siri?

[00:10:20] Coming. It's coming in waves.

[00:10:22] I mean, Siri and Truth has been evolving since its introduction.

[00:10:27] I mean, it's gone through many, many architectural changes.

[00:10:31] The original version didn't use deep learning at all.

[00:10:35] More recent versions were using deep learning in the cloud.

[00:10:38] The versions for the past couple of years ran entirely on device and did use deep learning models, but not large language models.

[00:10:46] And it's continuing to evolve.

[00:10:48] So with Apple Intelligence, we've created this foundation that understands personal context and the ability to take action.

[00:10:54] Siri is adopting that in stages and will benefit in stages over the coming year.

[00:11:03] And so with the initial launch, there are some cool things that the new Siri can do.

[00:11:08] So it has more understanding of your device.

[00:11:12] And so you can ask, internally we like to call it the device expert, but something that, you know, you can ask a question about how do you accomplish something in one of the Apple apps or setting up, you know, family sharing or something like that.

[00:11:24] And it can now answer it.

[00:11:26] So it's getting new capabilities.

[00:11:28] The new look is important because of where Siri is going, which is we wanted the glow to wrap around the screen because Siri is becoming able to understand the content inside the screen.

[00:11:39] And we're kind of conveying that Siri is there around your content and going to help you interact with it.

[00:11:45] And we've given you the ability to type to Siri.

[00:11:48] So those are things that are happening now.

[00:11:50] But some of the intelligence that you're talking about, we're really excited to bring to Siri.

[00:11:55] But it's not happening all at once.

[00:11:57] Thank you for answering a range of questions here today, Craig.

[00:12:01] My pleasure.

[00:12:02] Thank you, Joanna.

[00:12:03] That was WSJ Senior Personal Tech Columnist Joanna Stern speaking with Apple's SVP of Software, Craig Federighi.

[00:12:11] We'll link the full video of the interview in our show notes.

[00:12:16] And that's it for Tech News Briefing.

[00:12:18] Today's show was produced by Julie Chang.

[00:12:21] I'm your host, Zoe Thomas.

[00:12:23] We had additional support this week from Pierre Bien-Aimé.

[00:12:26] Jessica Fenton and Michael Laval wrote our theme music.

[00:12:29] Our supervising producer is Catherine Millsop.

[00:12:32] Our development producer is Aisha Al-Muslim.

[00:12:34] Scott Salloway and Chris Sinsley are the deputy editors.

[00:12:37] And Falana Patterson is the Wall Street Journal's head of news audio.

[00:12:41] We'll be back this afternoon with TNB Tech Minute.

[00:12:44] Thanks for listening.

[00:12:51] With Ecolab Science Certified, we take cleaning off your plate so you can focus on what's most important to your restaurant, your guests,

[00:12:58] and having them switch from giving your restaurant a go to making it a go-to spot.

[00:13:01] Ecolab Science Certified.

[00:13:03] Count on the scientific clean.

[00:13:04] Learn more at sciencecertified.com.

[00:13:06] first thing you need to haveülmes. This

[00:13:06] is important to have an office for me to be in the house of 4. You

[00:13:06] can focus on your restaurant just the way to make it a go-to spot.