🤓 This Week in AI
PLUS: What Mark Zuckerberg’s Hairstyle Can Teach Us About Spotting AI Deepfakes
Happy Friday!
This week, I saw an image of the world’s top tech leaders—Mark Zuckerberg, Elon Musk, Jeff Bezos, Google CEO Sundar Pichai, and Apple CEO Tim Cook—all lined up at Trump’s inauguration.
They have, after all, donated a cool $1 million each to the event.
Take a look 👇.
Convincing? Maybe.
But I could tell it was a fake right away.
The giveaway? Zuck was sporting his old hairstyle.
AI pulls from its training data, which means it's recreating patterns from what's available online.
Since most of these images show Zuck's classic pre-2024 shorter cut—and his new look hasn't been around long enough to saturate the internet yet—that's the version AI reproduced.
The takeaway?
If you’re a public figure and want people to easily spot your deepfakes, it’s time for a dramatically new hairstyle. 😆
Ok. Here’s what you need to know about AI this week (clickable links appear in orange in emails and underlined in the Substack app):
On his first day back in office, President Trump revoked a 2023 executive order signed by former President Biden—the most significant piece of AI regulation at the federal level in the U.S. The order aimed to reduce the potential risks AI poses to workers, consumers and national security.
He then signed a new order calling for an AI action plan within 180 days and a review of old policies to make sure they aren’t slowing down progress.
Trump also announced Stargate—a $500 billion private sector investment in AI infrastructure, backed by OpenAI, SoftBank, and Oracle. Sam Altman, Masayoshi Son, and Larry Ellison joined him at the White House for the big reveal of the initiative, which aims to keep the US ahead of China in AI development. The group has already started construction of its first data center in Texas.
Here’s what’s fascinating: Stargate is entirely privately funded, without a dime of government money involved. Yet Trump was the one announcing it, using the moment to frame it as a major step in keeping the U.S. ahead of China in AI development.
🧠 The real genius here? Sam Altman letting Trump take center stage. By doing so, Altman positioned OpenAI as the American AI champion while tying the President’s reputation to Stargate’s success. With Trump’s brand now invested, he has every reason to ensure the initiative succeeds.
Meanwhile, all the funding goes straight into building new data centers—for only one company: OpenAI.
💸 Google just invested another $1 billion into Anthropic, the OpenAI rival behind the Claude models, bringing its total backing to $3 billion.
Not to be outdone, Amazon has put up $8 billion, making this its biggest venture bet yet. Oh, and Amazon is also integrating Claude into the next-gen Alexa speaker.
Why does this matter?
Under the Biden administration, the Federal Trade Commission was taking a hard look at these kinds of cozy relationships between AI start-ups and their Big Tech backers.
Outgoing FTC commissioner, Lina Khan, specifically targeted Google and Amazon’s ties to Anthropic, along with Microsoft’s partnership with OpenAI, investigating whether these mega-deals were squeezing out competition in the AI market.
But with Khan on her way out and a new administration in charge, Big Tech is betting on a looser environment to push these deals through, and the race for AI dominance is only heating up.
OpenAI’s first AI agent, Operator, is here.
ChatGPT Operator can take control of your computer to order groceries, book restaurants, and purchase tickets—all at the same time.
Using a virtual browser, Operator navigates the web, clicks, types, and even fills out forms just like you would. It works in the background and sends you notifications when it needs your input, ensuring you’re always in control.
OpenAI has partnered with major platforms including OpenTable, Instacart, DoorDash, StubHub, Uber, Target and Esty to ensure the system meets real-world needs, and to improve the user experience. But the agent can work with any website.
Operator also requires your approval for logins and payments and has built-in safeguards to prevent misuse.
For now, it’s still in research preview and only available to Pro users ($200/month) with plans to expand to Plus users ($20/month) in the coming months.
If you want to see it in action, check out the live demo here.
Apple has paused its AI notification summaries for the news and entertainment categories after they generated inaccurate alerts. They’ll be disabled until Apple works out the kinks.
AI search engine, Perplexity, has made a bid to merge with TikTok U.S. Yes, really.
Is this a publicity play by the ambitious 28-year-old CEO to bring more attention and eyeballs to the platform and position it as a major contender in the AI space? Surely.
But you have to hand it to him. He’s built a reputation for turning awareness into momentum for his platform. This move is right on brand for him—and honestly, it’s not not smart 😊.
Either way, I’ll be here with my🍿ready to see how this plays out.
Chinese AI lab DeepSeek has launched an open source reasoning model (R1) it claims rivals OpenAI’s o1 on certain AI benchmarks including math, coding and complex reasoning tasks—all at a fraction of the cost.
(If you’re wondering, “open source” means the model’s structure and code are made publicly accessible, allowing researchers and developers to study, use, modify, improve and share it with others.)
A new study found 26% of students between the ages of 13-17 now use ChatGPT with help on their homework. This sounds a bit low to me, but…
A sign that you’re becoming fluent in a new language: You’re able to think in that language, instead of translating word-for-word into your native one.
Meta’s new translation model, SeamlessM4T works the same way—translating your voice across languages in real time without needing to turn it into text first.
It’s 23% more accurate than its competitors, supports 101 languages, works well in noisy settings, and powers everything video dubbing to smart glasses.
Netflix just launched “Go with the Flow,” an AI video generator that lets users control how objects and cameras move within AI-generated scenes. The project was developed by Eyeline Studios, Netflix’s VFX subsidiary with key operations in South Korea.
OpenAI has updated ChatGPT’s Custom Instructions feature to make it easier to personalize your interactions.
The new cleaner menu lets you specify:
Your name or preferred nick name
Your profession
Traits you’d like ChatGPT to have like “straight-shooting”, “skeptical”, “encouraging”, etc.
This is what it looks like:
If you’re already using Custom Instructions, your current settings won’t change.
If you’re not, you’re missing out.
Custom Instructions let you set the context and ground rules for every conversation, to give you more tailored and personalized responses.
They’re ideal for setting baseline information or long-standing preferences that don’t often change (field of work, role and responsibilities, interests, preferred tone or response formats).
For a deep dive into making the most of this feature, check out the INSIDER TIPS section of this edition of the newsletter.
This update is rolling out now on ChatGPT.com and is coming to windows in the next few weeks. I still don’t have it yet…
Legendary screenwriter Paul Schrader (Taxi Driver, Raging Bull) put ChatGPT to the test—and it exceeded his expectations.
“I’ve come to realize AI is smarter than I am,” he said, adding that he was “stunned” by its “original” and “fleshed out” ideas and feedback.
In five seconds, it gave him notes “as good if not better than most film execs” and left him asking why writers should “sit around for months searching for an idea when AI can provide one in seconds.”
Check out his full Facebook posts 👇.
The C.I.A. has developed an AI chatbot that lets analysts “talk” to virtual versions of world leaders to improve intelligence predictions.
In case you missed last week’s edition, you can find it 👇:
That's all for this week.
I’ll see you next Friday. Thoughts, feedback and questions are always welcome and much appreciated. Shoot me a note at avi@joinsavvyavi.com.
Stay curious,
Avi
💙💙💙 P.S. A huge thank you to my paid subscribers and those of you who share this newsletter with curious friends and coworkers. It takes me about 15+ hours each week to research, curate, simplify the complex, and write this newsletter. So, your support means the world to me, as it helps me make this process sustainable (almost 😄).