Loading summary
Transcript35 lines
- [00:00]
A
Many companies are struggling to scale their AI deployments or even move them past the pilot stage. Often the problem isn't technology, but organizational misalignment around goals, processes and incentives. At the break, join Caroline Roach, Senior Partner, IBM Consulting to learn why.
- [00:19]
B
Welcome to Tech News briefing. It's Tuesday, May 26th. I'm Isabel Bisquette for the Wall Street Journal. Some AI experts say a crisis is coming thanks to a delug bad and potentially even dangerous AI generated software. We're unpacking how this so called Vibe slop is being used and what could be at stake. Then the chip powering your brand new MacBook might have a defect or two, but that's the point. We'll break down the Apple strategy, helping the company solve cheaper laptops without sacrificing profits. But first, AI coding tools have made it easier than ever to generate huge amounts of software with barely any overhead. But now some top AI engineers are warning that a reckoning is coming. They say that this code, which they call Vibeslop, is buggy, risky and won't hold up for the long haul. The bottom line is that many companies could be trading short term productivity to for long term problems like service outages and security vulnerabilities. WSJ tech columnist Christopher Mims joins us to talk about how we got here, whether AI itself is a potential solution and what it all means for the workforce. Are there any benefits to accelerating code development at this kind of extreme pace?
- [01:51]
C
There are really big benefits when AI agents are applied to coding in the right context. So for example, I talked to Anthropic and OpenAI recently and they said there's a lot of legacy code that just needs to be refactored for newer systems. And it's just a huge amount of drudge work. And you can send the AI in and kind of make a first pass and then even use AI to review that code to make sure that it's meeting best practices and to make sure it's secure. The human still has to go in at the end of that and verify that everything has been done right, but it can just save many, many hours of work in those contexts. The important thing though is that in that process the human is very much in the loop. It's more like when you turn on cruise control on your car as opposed to getting in a Waymo and trusting it to get you to your destination with no input.
- [02:51]
B
So another thing I'm curious about is the extent to which AI can do that work of checking the buggy AI. Can AI solve the problems that AI is creating here.
- [03:03]
C
So folks who want to sell as many tokens as possible insist that the solution to Vibe slop is more AI. And in some contexts, they have a point. You can have AI or automatically check code written by other AI or by humans for security holes. That's a great application of it, as we've found out from Claude's Mythos. For example, you can have IT review that code to look for best practices, but obviously the issue comes when you're just asking it to do all the work for you from a minimal prompt.
- [03:39]
B
Are there people that are in the industry that are speaking out against this?
- [03:43]
C
There are a number of programmers who are really not sold on this kind of AI maximalist future. A couple of them who I talk to are the maintainers and the creators of this agentic harness called PI, which is the heart of OpenClaw, which of course is the super viral AI agent. They believe in AI to write code, but they warn that overuse of AI to write code, over reliance on it, is leading to a flood of Vibe slop, and that within companies, at some point, there's going to be a reckoning.
- [04:24]
B
Another thing I want to touch on is the impact on the workforce and this idea that maybe companies aren't going to need as many junior engineers because more experienced engineers are becoming a lot more efficient. But is there a danger in a generation that never learns how to code or do this stuff, the quote unquote hard way?
- [04:45]
C
That is, I think, almost a trillion dollar question. And I have heard so many different answers to it. It depends on how far into the future we're trying to make a prediction. You know, in the short term, lots of companies clearly are either freezing the hiring of junior engineers or even laying them off in the long run. You know, I've heard from folks at Anthropic, for example, that when done right, I can help onboard junior engineers get them into a big, complicated code base faster. When coding tools make everyone more productive at pushing out code, pushing out features, pushing out software, will the demand for software expand to accommodate that? You know, and in the past, whether it's energy, whether it's food, land, transportation, humans have always found a way to use more of a thing when it becomes more affordable. And so in the long run, you just have more people involved in that industry. So it remains to be seen when things will sort of turn the corner in the software industry and follow that historical trend. So, you know, jury's out.
- [05:54]
B
That was WSJ tech columnist Christopher Mims. Are you using AI to help you code. If you're a listener on Spotify, leave us a comment with your thoughts. Coming up, Apple's recently released budget laptop is powered by a chip that technically didn't make the grade. Why this decades old chip industry strategy is paying off for Apple? That's after the break.
- [06:25]
A
Delivering real business results with AI starts with organizational alignment, says Caroline Roach, senior partner, IBM Consulting.
- [06:32]
D
You need to understand what problem you're solving, really commit yourself to it, and then be clear on how you're going to change the way you work to deliver those outcomes. With AI as a tool, but not the only tool.
- [06:42]
A
If deployments aren't optimized to how work gets executed, ROI can suffer as programs stall and adoption lags.
- [06:49]
D
Alignment is difficult to achieve because people are rushing to a solution without first defining the problem.
- [06:55]
C
Problem
- [06:59]
B
when chips are manufactured, not every one comes off the silicon wafer working perfectly. Many are fully functioning. Some fail entirely. In between, there are chips with slight defects that can be repurposed. It's a strategy tech companies have been using for decades, and it's become a central component of Apple's design strategy and their bottom line. Now rising chip costs and AI demand are making the approach more valuable than ever. WSJ's Imani Moyes spoke with our reporter Rolf Winkler about his reporting on this.
- [07:35]
E
So, Rolf, how is Apple repurposing defective chips?
- [07:39]
F
We should be clear that it's not so much that the chip itself is defective or broken, it's more that they have a feature that doesn't work as well and so they can be repurposed for another use instead of being thrown away. And they work just fine. They work very well, in fact. It's like chickens lay different sized eggs, right? You can sell some as jumbo eggs and you sell others as regular eggs and you price them differently. It's just a way that chip companies like Apple, because Apple designs its own chips, manage that yield so it's less
- [08:14]
E
defects and more imperfections.
- [08:17]
F
Something like that, yeah. And for instance, what we're talking about here is, you know, everybody likes to talk about graphics processor units, right? GPUs. So in your iPhone, the chip has, say, six GPU cores, right? But in a MacBook Neo, the same chip has five GPU cores. So when they were making that chip, some had a malfunctioning GPU core. Well, that means they can't go into the iPhone. But do you throw them out? It could still work fine. Why don't we put it in a MacBook Neo instead?
- [08:51]
E
How long has Apple been doing this since the beginning.
- [08:54]
F
The first chip that they designed themselves, the A4, which is, oh gosh, 2010, the iPhone4. That chip went into an iPhone. It also went into the Apple tv. The reason that's relevant is because another flaw you can have in a chip is it leaks electricity, basically. And you don't want that in a device that's powered with a battery. But what if you put it into a device that's plugged into the wall just fine? So you can basically repurpose the same chip for a different device and save would otherwise have to maybe be recycled or thrown in the trash, basically.
- [09:34]
E
Many chip companies have been using the strategy for a long time. Why is it helping Apple stand out now?
- [09:40]
F
It's relevant now because of the MacBook Neo. The Neo is this very interesting device from them. They've never really had an entry level laptop and what they do is they make trade offs. There's a lot of features in a MacBook Neo that you know are not as good as what you get in a MacBook Pro or a MacBook Air, which you're paying more money for. One of them is saving money on the chip, right? Repurposing these chips that come from iPhones. The NEO is a popular device because, look, laptop from Apple is a game changer for them and puts them into a new pricing category that all of a sudden enables them to compete with Chromebooks and PCs. Right? And that's really interesting at a time that everyone in this industry is really struggling with memory and storage costs which are skyrocketing because of AI demand. You have to pay more for the memory storage that will get put into a PC or a MacBook or any of these devices. But Apple is actually still kind of going on offense with pricing with the MacBook Neo at $599 with the iPhone 17e that they didn't raise the price on relative to the 16e to take market share. Which is just an interesting way that a company like Apple with its massive supply chain and massive economies of scale can accomplish things that rivals just can't.
- [11:04]
E
What happens when Apple runs out of this category of chips like it has for the MacBook Neo?
- [11:08]
F
That's an interesting issue. Here is there's only so many of those chips that are available and the NEO is selling pretty well that Apple, according to our sources, is having to put in new orders for more chips so that they can continue to sell the NEO to people. People want to buy it and the only problem there is the cost structure, right if they were using chips they'd already made for the iPhone 162 years ago that just didn't meet spec and they were just sort of sitting around, well, they're not paying anything for those. They're basically repurposing chips they couldn't otherwise use. So now they've run out of those or are running out of those, so they need more and so they have to buy new ones. And that kind of blows a hole in the economics a little bit.
- [11:57]
E
Does that threaten the price point?
- [11:58]
F
That's a good question. If this thing is much more popular than they bargained for and they can't meet demand for it, yeah, theoretically they might have to raise price. They just did that for the Mac Mini, which is like this little puck computer you plug into a display. People are buying those like crazy to run AI at home and they can't make enough. So one way you manage supply and demand is you raise the price. And that's what they did with the Mac Mini. So it's in theory, you could see something like that happen with the NEO next year when and if they come out with the next model of it.
- [12:30]
B
That was our colleague Imani Moise speaking with WSJ reporter Rolf Winkler. 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 with supervising producer Katie Ferguson. I'm Isabelle Busquet for the Wall Street Journal. We'll be back later this morning with TNB Tech Minute. Thanks for listening.
- [12:58]
A
Scaling AI successfully requires more than the right technology. Here again is Caroline Roach, senior Partner, IBM Consulting.
- [13:05]
D
The biggest thing that we were talking about a year ago is what model to use. And the biggest thing that I'm talking about with my clients now is how do I drive change within my organization?
- [13:17]
A
Companies able to identify correct and then avoid misalignment will be best positioned to deliver meaningful business value from AI.
- [13:23]
D
The organizations that are the most successful set very clear targets and have several priorities that are very clear across the enterprise. The technology is really good, but if you're not changing your organizational alignment, not incentivizing your people correctly, not looking at workflows, you're not going to see real value with it.
- [13:44]
A
Visit IBM.com think leadership to learn how building organizational alignment can help deliver AI deployments that scale and drive growth.
- [13:52]
B
This content was created by custom content from WSJ, a unit of the Wall Street Journal Advertising department.