Welcome, humans.

Did you see OpenAI’s Super Bowl Ad yesterday? If you missed it, here it is:

The company wisely chose not to launch anything new…instead, it spent $14M on an ad that basically introduced AI to the ~130M viewers (who certainly aren’t AI experts).

What's really interesting is what they didn't mention: No AGI, no superintelligence, no “we're going to change the world”—even though the ad was literally called “The Intelligence Age” (no doubt a callback to this).

Instead, they showed practical stuff like ChatGPT helping with business plans and language tutoring, making AI seem approachable and fun. I mean, it kinda is, right?

What do you think? OpenAI was obviously going for its “Apple 1984” moment… did it work?

P.S: Yesterday we used Deep Research to ask who would win the Super Bowl LIX. Spoiler alert: it was dead wrong. Healthy reminder not to put too much trust in AI!

In that same spirit, here are Deep Research’s picks for Oscar winners across every major category. Let’s hope GPT is a better Academy predictor than an NFL one…

Here’s what you need to know about AI today:

• We break down Sam’s “three observations” for the future.

• OpenAI plans data centers, starting in Texas.

• Apple's AI car excelled in virtual testing.

• Research group listed 20 AI-replaceable jobs.

Sam’s “Three Observations” and predictions for “The Intelligence Age.”

While OpenAI didn’t announce anything new at the Super Bowl yesterday, Sam DID share “three observations” in an understated new blog post that laid out his vision for what happens as we approach artificial general intelligence (AGI).

Sam defines AGI as a system that can “tackle increasingly complex problems, at human level, in many fields.” And he said these systems are “coming into view.”

The piece included some wild numbers about AI economics, plus Sam’s predictions for the next decade.

First, the three observations:

1. More money = smarter AI (predictably and continuously).

   1. Sam says the intelligence of AI models scales up with resources (compute, data) in a way that's surprisingly predictable. It's like a cheat code—spend more, get smarter AI. Every. Single. Time.

2. Costs are in free fall. 

   1. Sam says AI costs are dropping 10x every year, which means what costs $1 today will cost $0.10 next year.

   2. Here's Sam’s proof: GPT-4's token costs dropped 150x between early 2023 and mid-2024.

   3. For context, Moore's Law (which revolutionized computing) was only 2x every 18 months. This is way faster.

3. Intelligence returns are super-exponential 

   1. Sam says that each small AI intelligence boost creates exponentially more value. This explains why companies keep increasing their AI investments—the returns just keep getting better.

The most interesting part? Sam thinks we're about to see AI agents that work like virtual coworkers, starting with junior software engineers, rolled out at top companies.

How does he know? OpenAI’s building such a tool right now. This version won’t replace senior devs, but it should be able to handle tasks that take 1-2 days at a time with human supervision.

Now imagine having 1,000 of these AI junior devs. Or a million. Working 24/7. Get ready, cause Sam says these systems are coming to every field of knowledge work.

And that’s why he thinks the world in 2035 will look very different:

• Most goods will get dramatically cheaper.

• Scientific progress will accelerate.

• Land and luxury goods might get way more expensive.

• Jobs will change, but won't disappear.

The wildest prediction? Sam thinks by 2035, anyone should be able to access “unlimited genius”—either through some kind of universal “compute budget” or by making AI so cheap that access isn't an issue. Y’all know that’s 10 years away, yeah?

FROM OUR PARTNERS

TOMORROW: Want AI assistants that actually know your business?

Learn how to build GenAI systems that can access all your company data (Drive, Salesforce, Confluence, etc.) without breaking the bank or burning out your engineering team. This live tutorial from Unstructured will show you how it's done, step-by-step.

Register here + get the replay even if you can’t attend live.

Prompt Tip of the Day

Here’s a prompt to get ChatGPT to stop agreeing with everything you say:

And here’s a more refined version someone else shared in the thread.

Treats To Try.

1. *Incogni removes your personal data from the open internet so scammers and identity thieves can’t access it. Protect yourself online with Incogni—get 55% off with code NEURON.

2. Every has a suite of pretty cool tools to try out:

   1. Spiral breaks up long content like podcasts or articles into dozens of social posts that sound exactly like you wrote them.

   2. Sparkle automatically sorts Mac files into personalized folders as soon as you save them.

   3. Lex helps you write better by suggesting improvements and alternatives while preserving your unique style.

   4. Cora, the newest tool (waitlist only rn) that cleans up your inbox by summarizing and filing your emails, keeping the important ones visible.

3. 2Pr interviews you about your ideas and turns your voice recordings into personalized LinkedIn posts.

4. Fitlock blocks your distracting apps until you complete daily health activities (Apple only RN).

5. CheckinBee sends daily text check-ins to your loved ones living alone.

6. Want something analog to offset your digital diet? Video to Flip turns your videos into pocket-sized flipbooks that animate when you flip the pages ($25).

See our top 51 AI Tools for Business here!

*This is sponsored content. Advertise in The Neuron here.

Around the Horn.

• OpenAI will consider 16 states for its other Project Stargate data center campuses—its first one is under construction in Texas.

• Apple’s self-driving AI averaged 17.5 years of continuous driving between incidents just by practicing against itself.

• Researchers analyzed 70 years of movie dialogue with AI and found movies have grown more violent over time.

• Here is a list of 20 jobs that o3 Deep Research thinks it could replace (from this chat).

• Check out this chart that shows the progress recent AI’s have made on “Humanity’s Last Exam”… notice anything, ahem, interesting?

Bots Behaving Badly

• Anthropic CEO Dario Amodei claimed DeepSeek was “the worst of basically any model we’d ever tested” in safety tests, and specifically, it generated rare bioweapon information (full podcast).

• California lawmakers proposed a new bill (bill SB 243) that would require AI companies to remind children their chatbots aren't human.

• GhostGPT is a new uncensored AI chatbot for cybercriminals that empowers them to more easily create malware and sophisticated phishing emails.

• “AI slop” books are infiltrating public library systems through vendors like Hoopla, forcing understaffed libraries to unknowingly distribute low-quality or potentially harmful AI-generated content to patrons.

• The UK announced four new laws targeting AI-generated CSAM, becoming the first country to make it illegal to possess AI tools that enable such content.

• Tinder hopes to reverse its 10% user decline by adding AI matchmaking, because nothing says “genuine connection” like having an algorithm tell you who you should fall in love with.

Monday Meme.

A Cat's Commentary.