LA Times’ Tough AI Launch

Plus: Google wants DOJ leniency; TSMC’s $100B deal

Author

Joey Padgett
March 06, 2025

In partnership with

Welcome back to Forests Over Trees, your weekly tech strategy newsletter. It’s time to zoom-out, connect dots, and (try to) predict the future.

Here’s the plan:

  • Tech News Takes — super-short analysis and commentary

  • Tool of the Week — tools you’ll find useful

  • Strategy Tips — strategy nuggets (for business and life)

  • F/T Shoutouts — sharing launches, tech events, and other reads

LA Times’ Tough AI Launch

Plus: Google wants DOJ leniency; TSMC’s $100B deal

⚡ Tech News Takes ⚡

  • What’s up: Earlier this week, the L.A. Times launched AI-generated counterpoints — called “Insights” — next to opinion pieces. They’re trying to build trust with readers and avoid bias by covering multiple sides of a given topic. Under the hood, the Insights feature uses Perplexity, the AI app known best for its search and citation functionality. But there have been some issues since launch. Critics accused Insights of having misleading citations and pro-AI bias. And in one instance, the AI feature crafted a counterpoint that defended the KKK.

  • So what: Trying to increase trust and reduce bias is well and good, but this is bad implementation. Instead of an unconstrained version of Perplexity, they could use customized models with more filters/safeguards, specifying out-of-bounds viewpoints. Sure, that’s a watered down version of “avoiding bias”, but anything beats defending the KKK. Alternatively, they could selectively enable the feature for some opinion pieces but not others, prioritizing it for topics where Americans hold a range of acceptable views, even if still a minority opinion for LA Times’ readers.

  • What’s up: Last week, Google met with the Trump DOJ, asking them to abandon efforts to break up the company. The DOJ has two ongoing antitrust cases against Google — one for search, and one for adtech. For the search case, which Google already lost, the next step is an April trial to evaluate remedies. The DOJ could potentially require Google to divest Chrome and end default search agreements with Apple. For the adtech case, the judge hasn’t ruled yet, although most analysts see the DOJ’s adtech case — and likelihood of winning — as even stronger than the search case.

  • So what: This is wild. On the one hand, it’s not surprising to see Google making its case to the Trump DOJ to see if they can steer clear of serious remedies. You can’t fault them — shooters shoot. But on the other hand, Google does seem to have unfair advantages, as we’ve covered here for search and here for adtech. It would be unfortunate for competition — and a huge waste of taxpayer dollars that have gone into the investigations/cases — if these end with a slap on the wrist.

  • What’s up: Earlier this week, TSMC and Trump announced that the chip-maker plans to invest $100B+ to build five more chip plants (”fabs”) in the US, adding to the $65B+ they’ve already committed/spent to build fabs in Arizona. The deal is still subject to Taiwanese government approval, and they’ll likely insist that the most advanced chips (2nm and 1.6nm) still come from Taiwanese fabs, so the US will only produce 3nm+. Even with those restrictions, TSMC will still be producing the most advanced chips of any fab in the US.

  • So what: It’s smart for Taiwan and TSMC to continue using advanced chips as a protection mechanism. As long as the US needs those and they can only come from Taiwan (which is true today), the US has a vested interest in deterring China from invading/absorbing Taiwan. It’s also fascinating to see the difference in circumstances and outcomes between the US and China. The more the US depends on TSMC for advanced chips, the weaker our own chip production gets. So we throw dollars at that problem the same way we do with AI model development — creating super-clusters using the most powerful, most expensive chips to advance our AI. In contrast, China can’t get as many advanced chips (thanks to trade restrictions imposed by the US and allies), so they invest in 1/ domestic chip-making (to be more self-sufficient) and 2/ AI innovation within chip constraints (resulting in AI models like DeepSeek!)

🛠️ Tool of the Week 🛠️

Build a successful marketplace business. Fast.

  • Launch a marketplace like Airbnb, Etsy, or Fiverr in a day with Sharetribe’s no-code marketplace builder.

  • Customize your platform with code as much as you want (or hire a dev from our Expert Network to do it for you).

  • Scale your marketplace to millions of users—never worry about infrastructure.

🧭 Strategy Tips 🧭

The LA Times Loses Trust

Today's strategy tip is all about trust.

We’ll dive deeper into the LA Times’ AI-generated counterpoints, using the Trust Equation framework to help break it down.

Let’s meet the framework.

The Trust Equation

This framework is a mental model for understanding how people and organizations can build or damage trust with each other. And the Trust Equation has four terms:

  • Credibility: Are you accurate, fact-based, and believable in what you say?

  • Reliability: Can I count on you to do what you say? Do you walk the talk?

  • Intimacy: Do I feel connected to you? Do I feel like you care about me?

  • Self-orientation: Are you mostly saying/doing things for yourself, or for others?

The first three terms have a positive relationship with trust – when those go up, trust goes up. But self-orientation has a negative relationship with trust – when that goes up, trust goes down.

For my visual learners, ze visual:

Ok, now let’s bring the LA Times story back into this.

Done well, the “Insights” feature for AI counterpoints might have helped increase trust, but it was done poorly.

Where the L.A. Times Went Wrong

Here’s how it affected each term in the Trust Equation:

  • Credibility down – citations for some of the stories were wrong.

  • Reliability down – unpredictable outputs mean that you have to be careful what you believe on the LA Times website.

  • Intimacy down – trust gets built when readers recognize writers they like, but no-name AI feels cold and commoditized.

  • Self-orientation up – AI summaries create a perception of being self-serving (cutting costs, on-trend company, etc.)

So it’s not surprising to see the reader backlash on this one – trust was damaged in multiple ways.

Wrapping Up

In the short term, they’ll probably abandon the feature. Sure, the traffic spike from everyone trying to see the issues firsthand is probably enticing…

…but not all press is good press (pun very much intended).

In the longer term, there are probably opportunities to roll this out in stages as hallucinations go down and guardrails become more effective.

🌲 F/T Shoutouts 🌲

  • Robot lineup — Loved this graphic from Visual Capitalist. Pretty fascinating to see the latest humanoid robots next to each other.

  • Collison Brothers — The Stripe co-founders (and brothers) made an appearance on the All In podcast a few weeks ago. I never get tired of hearing from them — worth a listen.

Want to suggest a shoutout? Send it here.

The forest is growing.
Feel free to share this post.

If you have feedback or want to discuss advertising,
reply to this email or send me a note.