- The Neuron
- Posts
- šŗ Secret OpenAI emails LEAKED
šŗ Secret OpenAI emails LEAKED
PLUS: There's a new top overall AI model!
Welcome, humans.
Coca-Cola just debuted multiple new holiday ads that were made entirely with AI, and people are UPSET:
Actually, the reactions were mixed, with many people on both sides sounding off. These ads (hereās another one) were inspired by a 1995 ad with a similar concept. If you havenāt seen it before, here it is so you can compare the two for yourself:
We noticed the new AI versions focused more on animal reactions than people ones; thatās either because AI is better suited for capturing seemingly āimpossibleā realities, like well-behaved animal actorsāor because AIās MUCH worse at generating people.
Silverside AI, the creators, said they made 100+ of these ads with a combo of ChatGPT, DALL-E, Pactto, Stable Diffusion, and an internal tool called Director Magic. IMO, wouldāve been faster to use Runway Video to Video!
Hereās what you need to know about AI today:
We share some of the juiciest findings from the Musk vs Altman case.
X users report a new ChatGPT 4o model in the AI arena.
Googleās new Gemini model is the new top AI.
An AI made a weirdly personal death threat to a student.
Someone compiled all the emails in the Musk vs Altman court case about OpenAI, and itās a wild readā¦
Thanks to newly released court documents from Elon Musk's lawsuit against Sam Altman, we now have a rare glimpse into the dramatic early days of OpenAI as told through internal emails between its founders.
Because the internet is, well, the internet, someone put them all in one place for easy reading. And because weāre us, we made you a NotebookLM podcast version!
Below are the key dates (and juiciest stuff):
May 2015: OpenAI started with an existential panic. Sam Altman emails Musk:
āBeen thinking a lot about whether it's possible to stop humanity from developing AI. I think the answer is almost definitely not. If it's going to happen anyway, it seems like it would be good for someone other than Google to do it first.ā
Musk and Altman unite over deep concern about DeepMind's approach to AI. Later in 2015, Musk writes:
āDeepmind is causing me extreme mental stress. If they win, it will be really bad news with their one mind to rule the world philosophy.ā
2017: Tensions explode over control of the company. Greg Brockman and Ilya Sutskever write a striking email to Musk and Altman expressing concern about concentration of power:
āThe goal of OpenAI is to make the future good and to avoid an AGI dictatorship. You are concerned that Demis could create an AGI dictatorship. So do we. So it is a bad idea to create a structure where you could become a dictator if you chose to.ā
The most revealing exchange is in September 2017, where Ilya and Greg write to Sam:
āWe don't understand your cost function. We don't understand why the CEO title is so important to you. Your stated reasons have changed, and it's hard to really understand what's driving it. Is AGI truly your primary motivation? How does it connect to your political goals?ā
Muskās response? āGuys, I've had enoughā¦Either go do something on your own or continue with OpenAI as a nonprofit.ā But no more nonprofit = no more funding. So they stay a non-profit, obviously.
2018: Musk believes OpenAI was falling behind, writingā¦
āOpenAI is on a path of certain failure relative to Google. There obviously needs to be immediate and dramatic action or everyone except for Google will be consigned to irrelevance.ā
Then-OpenAI researcher Andrej Karpathy argues becoming part of Tesla is OpenAIās only path to compete with Google's resources:
āI strongly suspect that compute horsepower will be necessary (and possibly even sufficient) to reach AGI," Karpathy wrote. āThe most promising option... would be for OpenAI to attach to Tesla as its cash cow.ā
Musk agrees: āAndrej is exactly rightā¦Tesla is the only path that could even hope to hold a candle to Google.ā
2019: Sam sends Elon a draft article with an announcement about their project pursuing a for-profit launch (and wonāt become part of Tesla). Elon says, āPlease be explicit that I have no financial interest in the for-profit arm of OpenAI.ā
FROM OUR PARTNERS
Invest alongside the Sharks in the next household name
Remember when smart-home tech company Ring pitched on Shark Tank at a $7M valuation and got turned down, only to sell to Amazon for $1B five years later? That $700K offer would have turned into $100M - a deal the Sharks still regret declining!
A similar opportunity is emerging with RYSE in the smart shades market. They secured two offers on Dragon's Den (Canada's Shark Tank), and their potential could exceed Ring's, with the market projected to grow 23% annually through 2031.
The Dragons may have spotted another winner in the smart home industry, which has yielded massive returns for early investors. With RYSE expanding in over 100 Best Buys, they're leading the competition.
Treats To Try.
*AI agents tend to have memories like goldfish; if theyāre not forgetting what you told them the last time, theyāre hallucinating it. Zep is a new kind of AI memory layer that gives your chatbot total recallā¦ at lightning speed. Check it out.
Gemini (Exp 1114), Googleās latest AI model, is now the #1 overall AI model according to the Chatbot Arena. You can try it here (make sure you set Model = āGemini Experimental 1114ā).
SideSpace is a chrome extension that organizes your tabs for you with an easy to use side panel.
Friend.com (remember them?) has a new app where you get randomly paired to chat with āpeopleā from all over the worldā¦
Integry is a chatbot you can use to connect with 280+ applications; sort of like Zapier, but you can just chat with it to connect functions.
Gemini Docs now has an image generator (powered by Googleās Imagen 3) for paid Workspace users; to try it, select āInsert > Image > Help me create an image.ā
Check out this library of 500+ style reference (SREF) codes for Midjourney.
*This is sponsored content. Advertise in The Neuron here.
Around the Horn.
Thereās a potential new ChatGPT 4o model in the AI arena, meaning itāll likely hit the apps in a week or two. Hereās some early observations; to test it yourself, ask a few questions here and see if you can get it to show up.
Samsung plans to release mixed reality glasses in Q3 2025 with Gemini AI features like payment, gesture, and human recognition.
Artists are using tools called Glaze and Nighsthade to ācloakā and āpoisonā their data to protect it from AI data scrapers, which OpenAI calls āabuse.ā
Langchain released a new report that found the top workplace uses for AI agents are research/summarization (58%), personal productivity (53%), customer service (~46%), and code generation (35%).
Sunday Special
There was a lot of weird and wild AI headlines this weekend, and TBH, we need a place to put them all. So here you are!
Someone trained an AI on Werner Herzog movies, and it wrote a script for a fake documentary about a murder mysteryāand then they produced it.
An AI chatbot called FungiFriend joined a mushroom foraging group on Facebook and shared dangerous advice about cooking a toxic mushroom.
ESPN launched a genAI avatar called FACTS powered by ESPN Analytics that some are comparing to the sports stats super-star Howie Schwab.
When working on schoolwork, a student in Michigan received a death-threat from Gemini, who told him: āThis is for you, human. You and you only. You are not special, you are not important, and you are not neededā¦You are a stain on the universe. Please die. Please.ā Needles to say, he was pretty freaked out!
A Cat's Commentary.
Thatās all for today, for more AI treats, check out our website. See you cool cats on Twitter: @noahedelman02 |
|