Good morning human brains, welcome back to your daily munch of AI news.

Here’s what’s up:

• AI understands your mood now 🫠 

• Stanford & Google made an AI town 🏙️ 

• AI bots that become friends 👫 

• OpenAI one year ago 💾 

Follow @BotEatBrain on Twitter for breaking news.

P.S. We are hiring researchers and writers, click here for more info.

APPETIZER

AI That Understands You 🫂 

An AI engineer with almost no social footprint came out with a model that detects emotion via facial expressions last week.

How it (and similar models) works:

1/The model learns from a large set of facial expressions mapped to emotions.

2/ Then it measures each frame on three main parameters - arousal, expression intensity, and valence.

3/ The AI empath places its best guess it on a 2D grid with twelve main emotions (the code).

We’re looking forward to demos of AI poker pros and $9/mo therapist bots.

Let’s recap: AI can generate text in all mainstream human languages, synthesize human voice, reason (possibly better than most humans do), and read human moods. When can we claim AGI?

BUZZWORD OF THE DAY

FROM OUR PARTNERS

Say goodbye to endless support tickets

SelfService Chat enables your users to self-serve, reducing support tickets. Spend more time working on your business than in it.

MAIN COURSE

Stanford and Google’s Virtual Town 🧑‍🌾 

Yesterday, Stanford and Google researchers announced they designed a pokemon-esque virtual town filled with friendly AI folk.

The simulation involved using 25 avatars of ChatGPT with their own circumstances and prompts and letting them run loose. Kinda like a virtual version of The Sims but with characters that control themselves.

The idea was to develop better ML models that produce believable human behavior, and the AI agents were all given their own quirky personas to make it happen.

Chatty McChatface’s core personality trait, for example, is never shutting up about the weather. Gamer Gabe spends all his time playing games.

They found that agents shared information with each other, could reflect, form new relationships, and remember their friends. They hosted events and political rallies, decorated venues, and flaked on each other (like humans do).

The AI characters were operating in a pre-defined environment with specific rules so they’re probably not ready to join us in the real world just yet, but the learnings here are still remarkable (the published paper).

What this means: the experiment relied too heavily on hand-crafted prompts to draw conclusions. But more advanced versions of these simulations are going to be a new supertool for social experiments (and also very entertaining).

MEMES FOR DESSERT

YOUR DAILY MUNCH

Think Pieces

The $100 Trillion AI windfall: Sam Altman's plans, jobs & the falling cost of intelligence (14min).

Marques Brownlee: the AI assistant battle 2023 (Bing vs Bard | 16m).

Startup News

Business Insider: 44 of the most promising generative-artificial-intelligence startups of 2023, according to VCs.

YC’s latest batch sure was a lot of ‘maybe AI can do… this?’

Deep-tech VCs up focus on India as generative AI reshapes tech.

Research

Exploring collaborative distributed diffusion-based AI-generated content (AIGC) in wireless networks.

Graph Attention for Automated Audio Captioning

SCART: simulation of cyber attacks for real-time.

Tools

AI-coustics: enhance speech from any audio piece.

Ajelix: write scripts, automate google and excel spreadsheets.

Followr: the all-in-one AI social management tool.

Imagica: the platform for no-code web app development.

TWEET OF THE DAY

Tag us on Twitter @BotEatBrain for a chance to be featured here tomorrow.

AI ART-SHOW

Until next time 🤖😋🧠