Latest Thoughts
-
🧠Who’s Really Driving?
The reality is that many people are getting what they think is a “robotaxi” ride—except it’s actually driven by humans. It’s a strange gray area where the lines between human-driven cars and semi-autonomous vehicles blur, leading us to ask: who’s really driving, and does it even matter?
This growing trend of Uber and Lyft drivers using Tesla’s Autopilot as makeshift robotaxis has sparked concerns about safety and the ethics of allowing this tech to stand in for a driver.
Uber and Lyft drivers use Teslas as makeshift robotaxis, raising safety concernsA self-driving Tesla carrying a passenger for Uber rammed into an SUV at an intersection in Las Vegas in April, an accident that sparked new concerns that a growing stable of self-styled “robotaxis” is exploiting a regulatory gray area in U.S. cities. -
🧠Meta’s AI Video Models Keep Advancing
Hot on the heels of Meta Connect, the AI team at Meta announced major updates to its video and audio models, though it’s still unclear when these tools will be available to the public.
I’ve had the chance to work with platforms like Pika and Runway, and while they’re impressive, they’re not quite where they need to be just yet. A few months back, Runway not only released its Gen3 model but also secured a deal with a Hollywood studio to train its AI on high-quality cinematic data. These improvements are exciting, but we’re still waiting for broader access.
As these tools get better, we might just see AI-generated moments—like Will Smith eating spaghetti—become common sooner than we think.
-
🧠OpenAI DevDay Brings Real-Time API and More
This week was OpenAI’s annual DevDay, and it arrives amid significant changes—most notably, the resignation of CTO Mira Murati and ongoing organizational shifts as OpenAI potentially transitions from a non-profit to a for-profit company.
While I’ll dive deeper into the conference in next week’s newsletter, one standout announcement is the introduction of a Realtime API that supports speech-to-speech agent implementations for developers, complete with six customizable voices. This is a massive leap toward enabling developers to create truly agentic AI experiences, integrating Retrieval-Augmented Generation (RAG) for personalized data access.
The new API also allows speech as direct input for chat completions and more. There’s still plenty to unpack, but I’m excited to dive into the keynote highlights.
-
🧠Tesla’s Supercharger Network is A Game-Changer for Other EVs
It’s no secret that charging infrastructure remains the biggest hurdle to widespread EV adoption. Kevin Purdy at ArsTechnica’s recent article highlights how Tesla’s Supercharger network has breathed new life into road trips with his Chevy Bolt, emphasizing how vital reliable charging is for long-distance EV travel.
I’ve said it before—range anxiety is a real issue. For most people, a home charger easily covers daily commuting needs. But on long-distance trips, access to a reliable, fast charging network becomes essential. In my own research, I’ve tested non-Tesla chargers, and it’s been a frustrating experience—apps, broken chargers, slow speeds, and payment issues. Tesla’s Supercharger network, on the other hand, has been a key reason I chose a Tesla in the first place.
Here’s hoping that axing the supercharger team at Tesla won’t cause this network to slide backward.
-
🧠Cloudflare Gives Creators Control Over AI Crawlers
Let’s face it—robots.txt wasn’t designed for the age of AI crawlers, which are ravenously consuming content across the web. For creators, it’s tough to swallow that their hard work is being used, often for free, to train AI models.
Cloudflare’s latest feature now allows websites to block AI models or bots with a simple click. If you’ve ever had to prove you’re human before accessing a site, that’s part of the toolkit Cloudflare is offering to help publishers stop the constant battle of restricting access.
While this might be a win for creators in the short term, there’s a lingering question: Will limiting access to AI crawlers make it harder for your content to be found in AI-powered answer engines like Perplexity AI? Only time will tell, but for now, the choice is yours.
-
🧠Meta Connect 2024 Kicks Off
Meta Connect kicks off on Sept 25th, and rumors are already circulating about potential new AR and VR devices. Leaks suggest a new Meta Galaxy Quest device, offering advanced features while maintaining its accessible price point. There’s also hope we might see updated Ray-Ban smart glasses following their surprising success and Meta’s extended partnership with the brand.
I’m especially eager to see Meta’s AI products take center stage this year. Will we hear about new Llama models or expanded open-source AI offerings?
-
🧠The Environmental Impact of AI
A growing concern in the tech world is the environmental impact of large AI data centers filled with Nvidia GPUs. Training advanced AI models requires immense amounts of power and water. For instance, ChatGPT-3 consumed an estimated 1.3 gigawatt-hours of energy—the equivalent of what 120 average households would use in a year.
In response to this rising demand, Microsoft has struck a surprising deal to reopen the Three Mile Island nuclear power plant. This will help power its Azure data centers, the same ones OpenAI uses to train its AI models, with cleaner energy sources.
The move highlights the need to balance the power-hungry requirements of training next-gen AI models (ChatGPT-4 used roughly 10x the power of its predecessor) while staying true to clean energy commitments.
-
🧠New Law Puts Limits on AI Replicas
Gov. Newsom has signed a new AI bill into law that places limits on companies creating digital AI replicas. For example, this law now requires permission from the family of deceased individuals before a company can create AI avatars, reproduce their voice, or develop AI-powered digital twins.
I recently created an AI version of myself and interviewed it. From that experience, I can say the technology to create these digital twins is evolving rapidly. At last year’s SXSW, a digital AI version of Marilyn Monroe was on display. More recently, ElevenLabs signed deals with several celebrities, allowing their voices to be used on the company’s platform. Meanwhile, CGI technology has advanced to the point where it can convincingly recreate or de-age celebrities.
These safeguards seem like a smart move, but they also raise a new question families need to consider: Should your will include instructions on whether you’d allow your information to be used for creating an AI version of yourself after you die? Would you permit your kids to license your voice or a scan of your body so that you could continue working and earning money in the afterlife?
Gov. Gavin Newsom Signs Bills Regulating AI Performance Replicas Into LawIn a symbolic move, the governor visited the Los Angeles headquarters of performers’ union SAG-AFTRA to greenlight the legislation: “We’re making sure that no one turns over their name, image and likeness to unscrupulous people without representation.” -
🧠Exploding Pagers Target Hezbollah
In a story that feels ripped straight from a movie plot, more than a thousand pagers—yes, old-school beepers—exploded, killing and injuring people in Lebanon. The attack targeted members of Hezbollah, who had resorted to using low-tech devices to avoid the tracking possible with cell phones.
It seems Israel intercepted an order for Gold Apollo brand pagers, substituting a different manufacturer to embed explosive materials inside the devices before shipping. The pagers were designed to detonate when they received a coded message.
-
🧠Mistral Expands Access with Free Tier for AI Developers
Llama isn’t the only open-source AI model on the block. Paris-based Mistral, the maker of several powerful AI models, is making it easier for developers to access its offerings with a new free tier and cost reductions on API access.
Mistral has struck an interesting balance by keeping some of its models open-source and publicly available while reserving its larger, more powerful models for licensing agreements or API access.
From what I’ve gathered, Mistral’s licensing agreements are also less rigid than Meta’s, as they don’t require developers to explicitly state that the products they build are powered by Llama.
-
📺 Conversations with My AI Doppelgänger
And the interview:
-
🧠Closing the Digital Divide: The Urgency of Internet Access for All
Marketplace has a compelling series called “Breaking Ground” that delves into the impact of the Chip Act, which aims to ensure every home in the U.S. has high-speed internet by 2030.
This series serves as a powerful reminder that the digital divide is still a reality for many, especially in rural areas where internet access is scarce. While many of us have become dependent on the internet for everything—from streaming TV to ordering groceries—there are countless communities across the U.S. that live without reliable access.
Here in Baltimore, it’s heartbreaking to hear that some students head to the library after hours just to use Wi-Fi for their homework. One librarian interviewed in the midwest mentioned finding people camped outside in the cold with sleeping bags, using the library’s connection because it’s the only high-speed option available in their community.
I often say that what feels normal to many of us, especially those deeply ingrained in the tech world, isn’t always reflective of everyone’s reality. But we can and must do better. Closing this gap is becoming as essential as providing water or electricity.
-
🧠Yelp Seizes the Moment After Google’s Antitrust Defeat
In the wake of Google’s recent antitrust loss, it’s clear that Yelp smells blood in the water. Jeremy Stoppelman, Yelp’s CEO, recently penned a blog post announcing that Yelp is suing Google, accusing it of being a monopoly that unfairly suppresses local search results.
Stoppelman makes a compelling case, arguing that Google has been propping up what Yelp calls an inferior local search product to capture more search traffic within its own ecosystem—something widely known as “zero-click search.”
As I’ve pointed out in my newsletter, this couldn’t come at a worse time for Google. For the first time, competitors like OpenAI and Perplexity AI see a path to challenge Google’s dominance in search. But AI-driven search is a different beast, something I’ve referred to as “answer engines.” Unlike traditional search, these tools don’t provide a list of links or drive traffic to the sources they pull from; instead, they deliver direct answers, posing a new kind of threat to Google’s search empire.
-
🧠EVs Are the Future, Just Not as Soon as We Thought
EV sales are up, just not as quickly as expected, but growth remains steady. Look, you know I have my issues with charging networks, but after nearly a year of driving an electric car, I can confidently say it’s one of the best decisions I’ve ever made.
Range anxiety, especially during the last mile when traveling, continues to be a challenge. But for day-to-day commuting, it’s the best type of car you could choose—assuming you have the infrastructure at home to support a charger. When I hop in my car for a meeting, it’s charged to the recommended 80% (I only max out at 100% if I’m planning a long drive), and even on round trips from Baltimore to DC and back, I’ve never felt the need to stop and charge.
If we’re talking about real car anxiety—I don’t miss the days of coming home late with less than a quarter tank of gas, only to wake up rushing to a meeting and needing to stop at a gas station on the way. An electric car at home is like waking up every morning with a gas station in your house that filled your car overnight.
The transition may take longer than we thought, and maybe hybrids will help people ease into the shift, but I have no doubt EVs are the future. I just hope these lousy charging networks get their act together and more people adopt EVs, helping solve the last mile problem that adoption will ultimately fix.
-
🧠Can We Even Trust Disses in Songs Anymore?
First, Drake used AI to add a voice clone of Snoop Dogg to a track, and now Grimes, Elon Musk’s ex and mother of his child, was impersonated to create a diss track—all with AI. It’s wild when you realize that even the disses in songs might not be real anymore!
-
đź“ş What is a Photo?
.jpg)
/cdn.vox-cdn.com/uploads/chorus_asset/file/25626085/2171810474.jpg)
/cdn.vox-cdn.com/uploads/chorus_asset/file/25596068/2168114186.jpg)
