DMV wins gold for innovation

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.

1/ A win for ‘little tech’

And no, I’m not talking about the ‘little tech’ that Marc Andreesen and Ben Horowitz endorsed earlier this month… I’m talking about tiny self-driving cars! I wrote about Nuro and their self-driving delivery pods way back in February 2023, and after a string of delayed launches and setbacks, they’re finally notching a win.

Nuro got approval from the California DMV last week to start testing in the Bay Area. It’s awesome to see this company get some good news, and it’s a positive for the autonomous industry too. Just the same way that Cruise and Tesla crashes dampen excitement about self-driving tech and cause things to slow down, getting permission to test (like Nuro) and expanding small betas (like Waymo) create a ton of momentum for the industry. Read more here from TechCrunch.

2/ A gift to say sorry

Over the last few weeks, Perplexity has come under fire for plagiarizing (allegedly!). Users had found paywalled content in certain responses from the AI search tool, including from Forbes and Wired.

Now, Perplexity seems to be saying sorry and bringing gifts, announcing a plan to split ad revenue with publishers. They’re claiming that they had been planning the move prior to the plagiarism drama, but the timing is just too good. Regardless of the “why”, it’ll be interesting to see this ad-sharing experiment play out, because Google offers publishers – wait for it – nothing.

One key difference between Google and Perplexity, though, is that until recently Google wasn’t’ snipping blurbs and summaries directly from publishers’ content. So if I’m a publisher, I’d be thinking “splitting your ad revenue is great, but I’d rather monetize my users directly on my site… and you’re still hurting that.” How this experiment plays out will likely hinge on how the math looks – how much ad revenue do they get from Perplexity, vs how much revenue do they lose from reduced on-site traffic. Read more here from The Verge.

3/ Voting for open models

You’re not going to believe this. Regulators are cheering for Meta. Well, to be fair, they’re supporting one thing Meta is doing. But Meta will be happy. Earlier this week, the US Commerce Department put out a report supporting open-source generative AI models, of which Meta’s Llama is one. And FTC Chair Lina Khan put out a similar statement last week. Both groups are supportive of open source because they think it promotes competition and improves model access/affordability.

But this is just a cherry on top for Meta. What’s more motivating, and likely an equally strong reason to pursue open-source, is that it hamstrings their competitors. If you can’t buy OpenAI, but they’re in the lead, then open source your models, start playing catch-up, and let innovators build on top of your early models to slow down OpenAI’s charge. Read more here from TechCrunch.

4/ TikTok mis-behaving, as expected

I saw this story in the Wall Street Journal last week, and I had two reactions. First, I thought “well of course they did!”, because we’ve heard a lot since the proposed ban about the harms that could come from China surveilling US citizens (of which this seems like a good example).

But then I thought “well of course they did!” for a different reason. At the end of the day, TikTok and its famously good algorithm are entertaining because they pay attention to what you like… and then they show you more of it... until you pick your head up and realize 2 hours have gone by. You could re-write this headline to say TikTok had collected U.S. users’ views on ponies and bread-making, and that would have been just as true.

Now look, I’m not saying TikTok is doing everything above-board, but let’s recognize that Instagram and YouTube and the other social media apps we love are doing something very similar. Read more here from the WSJ.

5/ Blockchain breakthrough

Against all odds, we have a second story in today’s post about the California DMV innovating! This time, they’re putting car titles on the Avalanche blockchain. That way, they’ll be able to digitally/easily transfer titles between owners.

This is awesome, and is actually a great use of the blockchain. Because the ledger of past transactions (title changes) can’t be altered, it will help drive down title fraud, part of the reason California chose it for their 42 million car titles.

These are the types of real-world, non-speculative blockchain use cases that crypto needs to lean into to start thawing their winter (homeowner deeds, supply chain proof of origin, etc.). Paradoxically, for those use cases, you need to convince really old institutions to adopt new tech. So it takes time, but it’s worth it! Read more here from Reuters.