- Bot Eat Brain
- Posts
- This NYU AI model was trained with footage from a baby
This NYU AI model was trained with footage from a baby
PLUS: Don't get catfished again
TOGETHER WITH
Good morning, human brains, and welcome back to your daily munch of AI news.
Here’s what’s on the menu today:
Exploit your infant to train AI 👶 🍼
Sam, a baby, trained AI with years worth of first-person footage.
Why is this scroll worth $1 million? 📜 💸
A team of 3 won Nat Friedman’s contest to decipher ancient text.
Bumble’s anti-catfish AI 🫣 👹
It announced a new tool that shows a 45% reduction in reported spam.
MAIN COURSE
Strap AI to your baby’s head for years 👶 🍼
Back in May, we covered a Senate testimony by a NYU professor, IBM rep, and Sam Altman. They urged lawmakers on the need for AI regulations.
A couple of weeks ago, we reported on NYU and Google DeepMind’s AlphaGeometry. It’s an AI model that solves university-level math problems.
On Thursday, New York University researchers released a paper on CVCL. It’s an AI model that learns language like a human baby.
… Like a baby, baby?
CVCL (Child’s View for Contrastive Learning) was trained with video recorded by a baby named Sam, spanning ages 6 months to 2 years.
The infant… recorded video?
The researchers placed a head-mounted camera on Sam. It captured 27,500 utterances from 61 hours of video. The footage contained him playing with blocks, interacting with his parents, harassing his cat, and more.
Wow. Is the AI any good?
CVCL was able to form connections by analyzing objects and words from the videos. It correctly categorized objects into sub-categories like animals, toys, and more.
So?
The researchers claim that the data captured by Sam is more valuable than data collected in a traditional environment.
“If you took this model, and you put it on a robot, and you had it run for 61 hours, you would not get the kind of data that they got here, that will then be useful for updating a model like this.”
Andrei Barbu, MIT Research Scientist
SPONSORED BY GROWTHSCHOOL
FREE AI & ChatGPT Workshop (New Year Offer) 🤯
Join this 3-hour ChatGPT & AI Workshop (worth $99) by Growthschool to master AI tools and ChatGPT hacks at no cost– and 10x your productivity.
With AI & ChatGPT, you will be able to:
✅ Make smarter decisions based on data in seconds using AI
✅ Automate daily tasks and increase productivity & creativity
✅ Skyrocket your business growth by leveraging the power of AI
✅ Save 1000s of dollars by using ChatGPT to simplify complex problems
SIDE SALAD
Scroll Over Beethoven 📜 💸
In September, we reported on Nat Friedman’s $1 million contest. Whoever deciphered text from 2,000-year-old scrolls would win the money.
On Monday, a team won $700,000 for transcribing the scrolls. The writing is attributed to the ancient Roman philosopher and poet, Philodemus.
$1 million for a scroll?
The scroll has the potential to rewrite the history of ancient philosophy, science, law, religion, drama, and more.
Wow. What’s on this thing?
Only 5% has been unrolled and deciphered. It contains Philodemus’s philosophical musings about life’s pleasures, music, food, and more.
Who won the money?
A team of three students won the Vesuvius Challenge Grand Prize of $700,000. They used AI-powered 3D scans to visualize the papyrus sheets, detect ink, and reveal the writings. Several other teams won $50,000 for their contributions.
Recommended Reading
How do you stay up-to-date with the insane pace of AI? Join The Rundown – the world’s fastest-growing AI newsletter with over 500,000+ readers learning how to become more productive using AI every morning.
1. Our team spends all day researching and talking with industry experts.
2. We send you daily updates on the latest AI news, tools, and tutorials.
3. You learn how to become 2x more productive by leveraging AI.
A LITTLE SOMETHING EXTRA
Don’t get catfished again 🫣 👹
Back in September, we reported on Bumble’s AI for bots, spam, doxing, and more. It blocked over 8.2 million spam accounts.
On Tuesday, Bumble announced Deception Detector. It’s a new tool that boasts a 45% reduction in user-reported spam, scams, and fakes.
Who cares?
According to FTC reports, victims of romance scams lost over $1.3 billion in 2022. In a Bumble survey, almost half of female users experienced anxiety over their matches being potentially fake.
Oh, boy. So this stops that?
Bumble claims Deception Detector blocked 95% of suspicious accounts in early tests.
What about dem nasty pics?
Bumble also has an AI tool called Private Detector for that. They open-sourced it, here.
YOUR DAILY MUNCH
Tools
Galileo 1.0 — a text-to-user-interface tool.
Coze — a platform for creating no code chatbots.
Ytube AI — takes full YouTube videos and turns them into SEO articles.
UserSketch — an intuitive text-to-data-retrieval tool.
Think Pieces
How confident is Microsoft in its AI domination? Its CEO, Satya Nadella, says, “we are waiting for competition to arrive.”
Who was in charge of the AI-generated Biden calls? Turns out it’s a company called Life Corporation that has been accused of misinformation before.
How can Venture Capitalist thrive in the crowded AI market? The role of research, social proof, and more in investing.
Startup News
Google leaked info about Gemini Advanced. It’s a paid form of Gemini that will compete with OpenAI’s GPT-4.
Apple is in negotiation with Brighter AI. It’s a startup that offers more natural image anonymization than traditional methods.
HuggingFace launched an AI Assistant. It’s open-source and leverages several LLMs like Llama 2 and Mixtral.
Research
PokéLLMon — an AI agent that achieves human-parity performance in Pokémon battles and other strategic battle games.
StepCoder — a code generation framework that excels in exploration and optimization tasks.
AToM — (Amortized Text-to-Mesh) a text-to-textured mesh method that generates high-quality, fully texted meshes in under a second.
MEMES FOR DESSERT
TWEET OF THE DAY
Has AI really developed that much since last year, or do we just understand it better? Or both?
Tag us on Twitter @BotEatBrain for a chance to be featured here tomorrow.
AI ART-SHOW
Until next time 🤖😋🧠
What'd you think of today's newsletter? |